Journal of Science and Cycling

Cognitive performance of healthy volunteers and professional cyclists during high intensity cycling

Wim Notebaert

Are decisions in cycling influenced by the physical load? There is little consensus regarding the relationship between the intensity of acute physical exercise and the quality of decision-making. Theories predict an inverted U-shaped relationship where cognitive performance is optimal during moderate-intensity physical exercise compared to both light and heavy intensities. This aligns with the Yerkes-Dodson law (1908), which posits that the level of arousal is the underlying factor explaining the inverted U-shaped curve. We examined behavioural and computational markers of decision-making during a conflict task across three increasing cycling intensities, up to the heavy domain. Thirty-one healthy volunteers completed the flanker task at light, moderate, and heavy exercise intensities on a stationary bike. In the flanker task, participants respond to a central target (arrow left of arrow right) with a corresponding response key, while ignoring flanking congruent or incongruent arrows. The response keys in this study are mounted on the handlebars. Throughout the experiment, heart rate, oxygen uptake, and rates of perceived exertion were measured and confirmed the prescribed exercise intensities. We used the diffusion model for conflict tasks (DMC, Ulrich et al., 2015) to break down the decision process into biologically plausible and theoretically grounded parameters. DMC analyses showed that drift rate (evidence accumulation) followed the hypothesised inverted U-shaped pattern, peaking at moderate intensity, suggesting optimal evidence accumulation at this workload. In a second study, we tested six elite cyclists from Soudal Quickstep – AG insurance Soudal. They performed the same flanker task in three intensities (based on recent physiological exercise tests): 80% of VT1 (moderate), middle of VT1 and VT2 (heavy), and finally, 5% over VT2 (extreme). Results showed increased performance from moderate to high intensity and, contrary to the healthy volunteers, no overall decline in the highest (extreme) intensity. Individual analyses suggest that sprinters (n=2) show further improvement in the highest intensity. This indicates that the detrimental effects of physical load are less pronounced in professional cyclists. Additional data suggest that sprinters show low body awareness (assessed by the MAIA questionnaire) which might shield against the otherwise negative impact of extreme physical load. Our novel method allows to test cyclists’ cognitive abilities under varying levels of physical load and might in the future be used to develop cognitive profiles.

Participation and Finish Trends for North American Winter Ultra-Endurance Races (129-1,609 km)

Kyle Wehmanen

Introduction: Winter ultra-endurance races present some of the most extreme challenges for endurance athletes. For these races, competitors travel across snow-covered terrain via bike, cross-country ski, or foot (i.e., walking, running, and/or snowshoeing) while carrying survival gear (1, 2, 3). Races last multiple hours, days, and even weeks and take place in temperatures that can reach as low as -40°C. In North America, the five premiere winter ultra-endurance races consist of the Tuscobia Winter Ultra 129 km, Arrowhead Ultra 217 km, Tuscobia Ultra 257 km, and Iditarod Trail Invitational 563 km and 1,609 km. To date, a comprehensive analysis of participation and finish trends from all of these races has not been reported. It is also unclear which exercise mode gives athletes the best chance of successfully finishing the race. Accordingly, our purpose for conducting this study was to compare participation and finish trends across the three exercise modes (bike, cross-country ski, foot) for North American winter ultra-endurance races ranging from 129 to 1,609 km.

Materials and Methods: Results from the Tuscobia Winter Ultra 129 km (2011-2023), Arrowhead Ultra 217 km (2005-2025), Tuscobia Winter Ultra 257 km (2011-2023), and Iditarod Trail Invitational 563 and 1,609 km (2022-2024) races were obtained from race event websites (1, 2, 3). For each year, the number of race participants and finishers in each exercise mode (bike, cross-country ski, foot) were identified. A chi-squared test was performed to assess the dependence between exercise mode and number of participants and finishers. Finish rates were calculated by dividing the number of race finishers by the number of race participants. Additionally, odds ratios were calculated to determine the association between exercise mode and finish rate (did finish vs. did not finish). A meta-analysis was used to combine odds ratios from the individual races to estimate the pooled odds ratio. Note that for the Iditarod Trail Invitational 563 and 1,609 km races, data were only included from 2022-2024 as race results from previous years did not include the number of participants who started the race.

Results: Across all 5 race events, there were a total of 4,129 participants and 2,455 race finishers race thus representing an overall finish rate of 59%. Number of participants, finishers, and finish rates across exercise mode for each race are illustrated in Figure 1. Notably, for all races bike had the most participants (Χ² (2, 4,129) = 1,704, p < 0.001)) and finishers (Χ² (2, 2,455) = 1,261, p < 0.001)). Accordingly, bike also had higher finishing rates (68%) compared to cross-country ski (51%) and foot (49%). Forest plots illustrating odds ratios for the association between exercise mode and finish rate for each race are presented in Figure 2. Specifically, racers who traveled via bike had greater odds of successfully finishing the race compared to those who traveled via cross-country ski (OR = 2.2, 95% CI [1.9-2.5], P < 0.001) or foot (OR = 2.3, 95% CI [1.7-3.1], P < 0.001). Racers who traveled via foot had no better odds of successfully finishing the race compared to those who traveled via cross-country ski (OR = 1.0, 95% CI [0.7-1.4], P = 0.99).

Discussion: Taken together, results from North American winter ultra-endurance races ranging from 129-1,609 km indicated that traveling via bike consistently yielded the most participants and finishers and greatest odds of finishing compared to cross-country ski and foot travel. These findings imply that traveling via bike may offer distinct advantages for self-supported winter ultra-endurance races. They also suggest that bike is the more efficient exercise mode thus offering a unique tool to facilitate human locomotion on snow (4, 5). Consistent with our findings, course records for all 5 races were also achieved via bike (1, 2, 3). Interestingly, for these races there were very few cross-country skiers that participated compared to cyclists and runners. Also, finish rates and odds of finishing were similar to those for traveling on foot. These data let us speculate that the depth of the cross-country skiing pool in North America may yield fewer skiers capable of completing these races compared to other continents like Europe. It is also important to consider the extent to which race conditions (i.e., snow surface, temperature, wind, precipitation) impacted finish rates. While we did not directly investigate this, race conditions, likely varied considerably across race courses and years and thus have a form of natural control built in. Notably, our analysis included limited data for the two longest races and thus the three shorter races were weighted heavier and had a greater impact on our results. Thus, the trends reported here should be interpreted with some caution. Nevertheless, these data offer preliminary insight into participation and finish trends for winter ultra-endurance races spanning a broad range of distances.

Figure 1. Number of participants (top), finishers (middle), and finish rates (bottom) across exercise mode for the 5 winter ultra-endurance races.

Practical Applications.

Our preliminary findings have implications for cyclists, cross-country skiers, and runners/snowshoers who compete in winter ultra-endurance races. Specifically, factors such as training, fueling, equipment, as well as exercise mode, should be considered when optimizing race strategy and performance. These findings may also have implications for individuals who travel long distances over snow covered terrain, including polar explorers and military personnel.

Cadence influences exercise thresholds and peak power output from a ramp test in recreational cyclists

Matthew Van Dyck

Purpose:

The purpose of this study was to examine how cadence influences performance outcomes and physiological responses during ramp incremental (RI) exercise tests. Specifically, we investigated whether cycling at 70, 90, or 110 rpm affects time to exhaustion (TTE), peak power output (PPO), and ventilatory thresholds (VT₁ and VT₂), and whether the effect of cadence differs across these exercise intensity levels.

Methods:

Nine healthy male recreational cyclists (22.4 ± 1.2 y; 74.9 ± 8.5 kg; 182 ± 5 cm) completed three RI tests at 70, 90, and 110 rpm in randomized order. Breath‑by‑breath gas exchange was recorded. VT₁ and VT₂ were determined using standard criteria and their associated PO were corrected for dissociation in the mean response time. Cadence and intensity effects were assessed using repeated‑measures ANOVAs.

Results:

TTE for the RI tests was longer at 70 rpm (747 ± 103 s), followed by 90 rpm (727 ± 98 s) and 110 rpm (696 ± 96 s) (p < 0.001). Similarly, PPO was highest at 70 rpm (424 ± 51 W), intermediate at 90 rpm (414 ± 49 W), and lowest at 110 rpm (399 ± 48 W) (p < 0.001). Power output at VT₂ was progressively lower with increasing cadence (70 rpm: 284 ± 50 W; 90 rpm: 260 ± 47 W; 110 rpm: 239 ± 46 W; p < 0.001). PO at VT₁ was higher at 70 rpm (218 ± 46 W) compared with 110 rpm (182 ± 35 W; p = 0.014), while no significant difference was found at 90 rpm (198 ± 43 W). Linear regression analysis showed that the decline in PO with increasing cadence was greatest at VT₂ (-1.15 W·rpm^-1), followed by VT₁ (-0.96 W·rpm^-1), and PPO (-0.66 W·rpm^-1). There was no significant difference in VO_2peak between the different conditions (70 rpm: 4.2 ± 0.6 L·min^-1; 90 rpm: 4.0 ± 0.5 L·min^-1; 110 rpm: 3.9 ± 0.4 L·min^-1).

Conclusion:

Cadence significantly influences key outcomes derived from RI tests, including TTE, PPO, and the PO associated with VT₁ and VT₂. The greater decline in PO with increasing cadence observed at ventilatory thresholds, particularly at VT₂, suggests that submaximal exercise intensities are more sensitive to cadence than maximal performance. These findings highlight that exercise thresholds are cadence-dependent and suggest that prescribing training intensity based on a single threshold value without accounting for cadence may result in substantially different physiological stress.

Principal Component Analysis (PCA) Detects a Bilateral First Principal Component (PC1) Score Shift Not Systematically Captured by Conventional Metrics: A Longitudinal Single-Rider Case Study (2014–2025)

Masahiro Fukuda

Introduction

In practice, power balance (PB)—the ratio of left-to-total power output—is the most commonly used bilateral scalar for assessing pedaling asymmetry (ANT+ Alliance, 2013). PB reflects scalar force magnitude rather than the force-profile shape that characterizes each leg's coordination pattern; asymmetry indices may therefore vary substantially depending on the measurement and analysis approach (Murray et al., 2024). Most commercially available power meters report only power output, and even devices capable of measuring crank forces typically export only scalar summaries per revolution. The Pioneer Pedaling Monitor and its Shimano-branded equivalent instead record tangential and normal crank forces at 12 discrete crank-angle positions per revolution, enabling force-profile shape analysis that is not available from conventional power meters. A principal component analysis (PCA) framework applied to slope values of these pedal force vector components was evaluated previously in a multi-rider study (Fukuda & Kitawaki, 2026). PC1 was interpreted as reflecting coordinated force production between the pushing and pulling phases, representing a coordination pattern characteristic of each leg's technique and independent of power output magnitude. The present study extends this framework to independent bilateral projection using a single-rider longitudinal dataset (2014–2025) to examine whether PCA-derived left–right asymmetry detects patterns that remain undetected by conventional scalar indicators.

Methods

A single trained male competitive cyclist (age 41–52 years) was monitored across nine Tour de Okinawa (Gran Fondo 140 km) race events during a 12-year observation period (2014–2025). Data were collected using bilateral crank-arm power meters (Pioneer Pedaling Monitor, 6 events; Shimano-branded equivalent, 3 events). After quality filtering (communication-error and outlier exclusion), 1,327 training activities were included in the analysis. PCA eigenvectors derived from a multi-rider model (Fukuda et al., 2026) were used to project the left and right legs independently, yielding PC1 scores for each leg. Bilateral asymmetry was defined as ΔPC1 = PC1_Left − PC1_Right (positive values indicate a higher left-leg PC1 score). PB was used as the conventional scalar reference metric.

Results & Discussion

PB showed year-to-year variability without a systematic directional trend (range: −1.4% to +2.3%; L% − 50), which may limit its usefulness for longitudinal technique monitoring. In contrast, PC1_Left showed a substantial shift while PC1_Right showed no systematic directional trend (range: −1.44 to +0.03; Table 1). Consequently, ΔPC1 increased from 0.12 in 2014 to a peak of 2.06 in 2023 (Figure 1). This divergence was not reflected in PB and suggests greater sensitivity of PCA-derived indices for longitudinal monitoring. Race-day PC1 scores (Figure 1A) equaled or exceeded the training distribution median in eight of nine observed years, broadly consistent with enhanced coordination during peak race preparation. PC1_Left began increasing during the earlier period (2014–2019), preceding both injury events (right shin laceration, February 2020; crash with right clavicle fracture, June 2023). PC1_Right showed no notable change following either event, which is consistent with the rider's report of limited right-side function after injury. Training-period ΔPC1 showed a strong rank correlation with race-day ΔPC1 (Spearman ρ = 0.88, n = 9 race years).

Practical Applications

Cyclists equipped with force-sensing power meters (Pioneer Pedaling Monitor or the Shimano equivalent) can apply this PCA framework to monitor bilateral crank-force coordination independently for each leg. Unlike conventional power balance, PCA-derived ΔPC1 reflects force-profile shape and may reveal asymmetric technique changes or injury-related compensatory patterns that remain undetected by scalar power metrics. Coaches and sports scientists may therefore use longitudinal ΔPC1 trends to complement power-based monitoring when evaluating technique interventions or rehabilitation progress.

Conclusions

ΔPC1 (= PC1_Left − PC1_Right) captured a bilateral asymmetry pattern—characterized by increasing left-leg PC1 values with no systematic directional trend in right-leg PC1—that was not reflected in power balance throughout the observation period. This pattern is consistent with the rider's self-reported intentional left-leg technique modification, which predated both injury events, and with right-leg functional limitation following injury, although these interpretations remain speculative and qualitative. These findings provide proof of concept and warrant replication in larger independent cohorts.

Cycling Performance Optimization, A New Telemetry System to Study Tire Inflation Pressure

PABLO DÍEZ MARTÍNEZ

This study investigated the relationship between tire inflation pressure (ranging from 4 to 10 bar) and rolling resistance force to optimize cycling performance at a concrete velodrome, using a new low-cost telemetry system also the aerodynamic coefficient in real-time. Results revealed significant differences between pressures up to ±9% absolute power to maintain the same speed, while the aerodynamic coefficient remained stable. Concluding that using telemetry to determine ideal pressure settings can significantly improve mechanical efficiency and overall athlete performance

The effect of geometry simplifications on aerodynamic cyclist posture optimization through numerical simulations

Wouter Terra

Cyclist aerodynamic performance can be optimized using Virtual Skeleton Methodology, adapting the rider posture, and evaluating its performance through numerical simulations, aiming to minimize overall aerodynamic drag. The goal of the present work is to understand if the outcome of this optimization procedure relies on the presence/absence of a bike or the specific leg orientation, when changing the rider's upper body position. The Generic Cyclist Model in time-trial position is used adapting the lower arm angle. The results show that the flow downstream of the cyclist remains relatively unaltered when removing the bike. The arm position that exhibits minimum aerodynamic drag is also independent of the presence of the bike and, in addition, of the specific leg orientation. This suggests that virtual cyclist posture optimization can be conducted without a bike and considering only a single leg position, which significantly simplifies the optimization process.

Effect of carbohydrate and sodium bicarbonate co-ingestion on the dynamics of the power-duration relationship and blood acid-base regulation following 3 h of moderate-intensity cycling

Bernardo Norte

Introduction: Durability, defined as the ability to withstand deteriorations in physiological profiling variables during prolonged exercise, is increasingly recognised as a key determinant of endurance performance. While carbohydrate (CHO) ingestion may attenuate, or even negate, fatigue-induced reductions in intensity domain transitions, the extent to which CHO preserves functional declines across the power-duration curve remains largely uncharacterised. Moreover, supplementation strategies beyond CHO fueling aimed at enhancing durability are still in their infancy. Sodium bicarbonate supplementation is widely used to mitigate the effects of metabolic acidosis and improve extracellular buffering capacity during short duration high-intensity exerciselasting up to 10 mins. However, prolonged moderate-intensity exercise also induces progressive metabolic and acid-base disturbances, which may reduce tolerance to heavy- or severe-intensity exercise under fatigued conditions. It is plausible that augmenting extracellular buffering capacity could attenuate fatigue-induced reductions at the boundary between exercise intensity domains and preserve the power-duration relationship. Therefore, the aim of this study was to investigate the effects of CHO and bicarbonate co-ingestion on the durability of the heavy-to-severe domain transition (i.e., critical power), repeated sprint ability and blood-acid regulation following 3 h of moderate-intensity cycling.

Materials and Methods: Using a randomized, counterbalanced crossover design, twelve endurance-trained cyclists and triathletes (aged 18–49; V̇O2max >50 mL·kg–1·min–1) will complete two characterisation trials to establish baseline physiological parameters, including a series of repeated all-out sprints (3 x 6-second) followed by two time-trial (TT) efforts (2-min TT and 12-min TT) performed in a rested state to determine critical power (CP) and work capacity (W′) above CP. Subsequently, on two separate occasions, participants will repeat the same protocol immediately following 3 h of moderate-intensity cycling at 95% of gas exchange threshold while consuming either CHO only (80 g·h-1; glucose-to-fructose ratio of 1:0.8) or CHO plus sodium bicarbonate (80 g·h-1 and 20 g of bicarbonate over the 3 h of exercise). Each experimental trial will be preceded by a standardized 24 h CHO loading protocol (8 g·kg–1 body mass) and pre-exercise breakfast (2 g·kg–1).

Results: Pending.

Conclusion: Pending.

Dropped Out - Physiological and perceptual responses at the point of task disengagement in intermittent endurance events

Joao Henrique Falk Neto

Intermittent endurance events require athletes to perform multiple surges in intensity over the course of a race. These surges usually occur at supramaximal intensities (i.e., above the athletes’ maximal aerobic power - MAP), exacerbating the race’s physiological and perceptual demands. The ability to complete these surges and stay within the leading pack is essential for success in these events, as the race outcome is often decided in the finishing sprint. Many athletes, however, are unable to sustain the demands of these surges, and need to manage their efforts to ensure they can complete the race. Thus, this study had 2 aims: (1) to describe the physiological and perceptual responses at the point of task disengagement during an intermittent endurance event, and to determine whether athletes who skipped a sprint experienced greater physiological and perceptual demands at disengagement than those who did not, and (2) to determine if there were any differences in the determinants of endurance performance between the group that completed all sprints and the group that did not.

The study recruited 14 endurance-trained males (age: 35.5 ± 9.6y; bodyweight: 82.5 ± 10.3 kg, VO₂max: 58.9 ± 12.4 ml.kg.min^-1). The participants performed a protocol consisting of fifteen 10-second sprints separated by 50 seconds of low intensity cycling (60% maximal aerobic power – MAP). The sprints were performed at an intensity associated with 25% of the athletes’ anaerobic power reserve (APR). Seven participants completed all sprints during the protocol (COMP), while seven skipped at least one sprint (SKIP). On average, participants first skipped a sprint at sprint 9 (60% of the protocol). Therefore, the responses of the COMP group at sprint 9 were used for between group comparisons. An independent samples t-test was used to determine if there were any differences in physiological (oxygen consumption - VO₂, heart rate – HR, respiratory rate – RR, blood lactate, muscle activity – expressed as a percentage of the participants’ maximal voluntary contraction – MVC) and perceptual (rate of perceived exertion – RPE, and leg discomfort - LD) responses between groups at the point the participants decided to skip a sprint. The same test was used to assess if there were any differences in physiological and performance characteristics between the two groups (VO₂max, MAP, maximal peak power, and APR).

At the point of task disengagement, the participants in the SKIP group had a high oxygen consumption (84.1 ± 6.9% of VO₂max), respiratory rate (49.0 ± 5.9 breaths per minute), muscle activity (60.1 ± 17.5% MVC), and blood lactate level (10.4 ± 2.0 mmol.L^-1), along with near maximal heart rate (92.8 ± 2.7% HRmax), and high perceptual demands (RPE = 15.7 out of 20, LD = 6.8 out of 10). However, there were no differences in these responses between the COMP and SKIP groups. Despite the lack of statistically significant differences, a moderate to large effect size was reported for respiratory rate (d = 0.73, 95% CI = -0.36 to 1.80), blood lactate (d = 0.75, 95% CI = -0.35 to 1.82), and LD (d = 0.93, 95% CI = -0.19 to 2.03). Participants in the COMP group had significantly higher VO₂max (p = .034, mean difference = 13.6 ml.kg.min^-1, 95% CI = 1.17 to 26.10 ml.kg.min^-1) and MAP values (p = .018, mean difference = 61.4W, 95% CI = 12.5 to 110.2W) than those in the SKIP group.

The results indicate that very high levels of physiological and perceptual demands are attained at the point of task disengagement during intermittent events. However, having a higher VO₂max and MAP might allow athletes to withstand these demands during a race. As such, coaches and sport scientists involved in these events should focus on developing their athletes’ aerobic power to increase their chances of success. Lastly, the moderate to large effect sizes reported for physiological (respiratory rate and blood lactate) and perceptual measures (leg discomfort), highlights that further studies are required to determine the importance of these variables to assess exercise tolerance in intermittent endurance events.

Fractal Correlation Properties of Heart Rate Variability Differentiates Training Intensity During and After Energy-Matched Cycling Sessions

Anton Olieslagers

Introduction/purpose: Recovery optimization requires precise monitoring of physiological stress. This study examines whether detrended fluctuation analysis alpha 1 (DFAa1), a non-linear heart rate variability index, exhibits intensity-dependent responsiveness across energy-matched training sessions in cycling. Methods: Nineteen endurance-trained male participants (18-35yrs) completed three equalized workload protocols at different intensities: light (LIGHT), moderate (MOD), and high-intensity interval training (HIIT). Training sessions were matched for energy expenditure by adjusting time via the formula E=P×Δt (E=Energy (J); P=power (W); Δt=time (s)). LIGHT and MOD training were conducted at first and second lactate threshold power (i.e. LT1 and LT2), respectively. HIIT followed a 10x1min protocol at 110% V̇O₂peak power, alternated with 1min rest at 80% LT1 power. Heart Rate (HR) and DFAa1 were assessed throughout the different intensity training sessions as well as during a standardized 10min warm-up (WU) and cool-down (CD) at LT1. Results: DFAa1 was remarkably stable during LIGHT (p=0.997) and MOD sessions (p=1.000), while showing deterioration during HIIT (p<0.001). On the other hand, HR drift was apparent for all intensities (LIGHT: p=0.003; MOD: p<0.001; HIIT: p<0.001). When directly comparing DFAa1 and HR via ΔZ-scores, a significant effect was found only in MOD (p=0.008). Interestingly, DFAa1 demonstrated significant WU to CD differences for MOD and HIIT (p<0.001), but not for LIGHT (p=0.210), while HR responses showed significant WU to CD differences for all intensities (p<0.001). ΔZ-scores showed significant differences between DFAa1 and HR in MOD (p=0.039) but no significant differences in LIGHT (p=0.098) and HIIT (p=0.071). Conclusion: These findings reinforce that DFAa1 holds promise as an objective, system-wide marker of internal workload across varying exercise domains.

Beyond a Single Time Constant Evaluating the Cross-Protocol Transferability of τW′ in Cyclists

GORKEM BALCI

Abstract:

Introduction: The quantification of W' recovery kinetics (τW′) remains a critical challenge in W' balance (W'BAL) modelling (Welburn et al., 2025; Caen et al., 2021). While early models proposed generalised τW′ values, recent consensus emphasises the necessity of individualisation to enhance predictive accuracy (Pugh et al., 2025; Bartram et al., 2017). However, a critical unresolved question is whether a single individualized τW′ value functions as an intrinsic, static physiological constant for a given athlete, or if it is dynamically modulated by the specific characteristics of the exercise protocol (e.g., work-rest ratio, bout duration). From a practical standpoint, resolving this is crucial: while generalized equations are already known to produce large predictive errors for individual athletes, even an individualized τW′ will fail to optimize real-time W'BAL monitoring and pacing strategies if it cannot be reliably transferred across varying intermittent exercise modalities. That is, it remains unclear whether τW′ is protocol dependent or transferable across disparate exercise modalities. Therefore, this study aimed to test the robustness and cross-protocol transferability of the individualised τW′. We aim to determine if an individualized τW′ derived from a standard intermittent protocol (60s-30s HIIT) can accurately predict task failure across disparate exercise modalities including short HIIT (S-HIIT), long HIIT (L-HIIT) and decreasing power output HIIT (DPO-HIIT).

Methods: Level 3-4 male cyclists (n= 6), age 27.6 ± 6.8 years, body mass 70.7 ± 6.7 kg, CP 284 ± 36 W (SEE: 1.9 ± 0.9 W), W' 19.1 ± 2.1 kJ (SEE: 1.3 ± 0.5 kJ) and V̇O2max 59.1 ± 4.7 mL·kg·min−1, completed 8-9 laboratory visits. Baseline tests involved determination of V̇O2max and peak power output (PPO) using a ramp incremental cycling test, critical power (CP), W′, and IHIGH (highest intensity permitting V̇O2max attainment) using constant-load tests. Following baseline tests, participants performed a 60 s–30 s HIIT protocol (Work: 94% PPO; Recovery (Rec): 50 W) to volitional exhaustion. The specific τW′ value required to yield a W′BAL of exactly 0 kJ at task failure was retrospectively calculated for each participant and designated as the reference time constant (Ref-τW′). This Ref-τW′ was subsequently applied to three distinct HIIT protocols performed to exhaustion: S-HIIT (15 × 44 s/44 s; Work: ~110% of PPO; Rec ~10% of PPO), L-HIIT (4 × 3 min/3 min; Work: ~89% of PPO; Rec ~25% of PPO), and DPO-HIIT (6 × 115 s/115 s; Work: starting at IHIGH and decreased by 4 equal steps to the CP + 10%; Rec ~25% of PPO). Model performance at exhaustion was assessed using Bias (mean of raw errors), and Root Mean Square Error (RMSE) to quantify the overall magnitude of the prediction error. Variables were analysed using one-way repeated-measures ANOVA with Bonferroni post-hoc comparisons. Significance was accepted at p<0.05. Results: End-exercise W′BAL estimates at task failure differed significantly when Ref-τW′ was applied across the distinct HIIT protocols (F=27.70; p<0.001; ηp²=0.847). The mean Ref-τW′ values derived from the 60 s–30 s HIIT protocol was 83±8 s. When this Ref-τW′ value was applied, the lowest systematic bias and RMSE were observed in the S-HIIT protocol (Bias: 3.3 kJ; RMSE: 4.3 kJ). Conversely, application to the L-HIIT and DPO-HIIT protocols resulted in substantially larger prediction errors (Bias: 6.5 kJ and 7.0 kJ; RMSE: 6.8 kJ and 7.4 kJ, respectively). Actual τW′ values required to reach 0 J at task failure were different across protocols (F=26.24; p<0.001; ηp²=0.840). Post-hoc comparisons revealed no differences in the required τW′ between the 60 s–30 s and S-HIIT protocols (p=0.095) nor between the L-HIIT and DPO-HIIT protocols (p=0.269).

Discussion/Conclusion: Preliminary findings suggest that speed of an athlete's recovery drastically slows down during long-interval workouts compared to short-interval workouts. Translated to bioenergetic modelling, these findings suggest that τW′ is not a constant, universally transferable physiological trait, but rather a highly protocol-dependent variable. The inability of the Ref-τW′ to accurately predict task failure in L-HIIT and DPO-HIIT underscores the limitations of using a single time constant across divergent HIIT structures. The required τW′ values naturally grouped into two distinct clusters: short-interval (60 s–30 s and S-HIIT) and long-interval (L-HIIT and DPO-HIIT) modalities. This clustering aligns seamlessly with recent bi-exponential models of W′ reconstitution (Caen et al., 2021; Chorley et al., 2022). Specifically, short-interval modalities are thought to rely predominantly on the fast phase of recovery, which is closely associated with rapid phosphocreatine (PCr) resynthesis. In contrast, the prolonged work bouts in long-interval modalities likely induce greater metabolic perturbation, presumably making the overall reconstitution highly dependent on the slow phase (associated with intracellular metabolite clearance), thereby substantially lengthening the required τW′. Consequently, contemporary W′BAL models must evolve beyond treating CP, W′, and τW′ as unchanging constants and assumptions and incorporate dynamic scaling factors that account for the accumulated work duration and the progressive nature of exercise-induced metabolic stress. From an applied perspective, coaches and practitioners should recognize that real-time W′BAL monitoring and pacing strategies cannot rely on a single static recovery metric based solely on athlete characteristics. To optimize performance and prevent premature exhaustion, training prescriptions and algorithms must explicitly account for protocol characteristics, particularly the inherently slower recovery dynamics that occur during longer interval structures.

Single-Session Field-Based Evaluation of Durability in Professional Male Cyclists via 5-Minute Time Trials

Borja Martinez-Gonzalez

This study investigated endurance durability in professional male cyclists using a single-session, field-based 5-minute time trial (5min-TT) protocol under fresh and fatigued conditions. Twenty elite cyclists from a UCI ProTeam participated during a pre-season training camp. The experimental session simulated race-specific demands, combining prolonged moderate-intensity cycling with repeated high-intensity efforts on a standardized uphill road segment. Following an accumulated workload of approximately 45–50 kJ·kg⁻¹, cyclists exhibited significant declines in absolute and relative power output, torque, and PO:HR during the fatigued 5min-TT compared with the fresh condition. In contrast, heart rate and cadence remained largely unchanged. These results indicate that performance reductions were primarily driven by decreased torque rather than alterations in cadence or cardiovascular response, highlighting the role of neuromuscular mechanisms in fatigue resistance. The dissociation between power output and heart rate suggests that PO:HR decoupling may serve as a practical marker of endurance durability during high-intensity efforts. Overall, the findings demonstrate that a single-session, field-based 5min-TT is a feasible and effective method for assessing durability in professional cyclists.

Periodization and Training Context Modulate Hemoglobin Mass Responses to Altitude Training in Elite Endurance Athletes

Romain Carin

Altitude training camps are commonly used by endurance athletes to increase total hemoglobin mass (Hbmass). However, inter- and intra-individual variability in Hbmass response remains incompletely understood. We hypothesized that Hbmass responses are influenced by periodization and training context rather than by fixed “responder” characteristics. Several protocols conducted over the last three years highlight key factors to optimize Hbmass response during altitude training.

Nine elite triathletes completed a 4-week “live high–train high” (LHTH) camp during the competitive season, with an intermediate measurement at day 18 (D18). Hbmass was not increased at D18 but was significantly elevated at D28 (+3.7 ± 2.6%). The same athletes also performed an 18-day LHTH camp during a return-to-training phase, during which Hbmass increased significantly by +4.5 ± 2.8%. Notably, the magnitude of increase during the 18-day camp was comparable to that observed after 28 days in the competitive-season camp. Differences in training periodization between phases may have contributed to these distinct responses, with a faster Hbmass increase during the return-to-training phase.

To further explore the role of training context, seven elite open-water swimmers completed a 4-week sea-level block with increased training load, resulting in a +5.9 ± 3.4% increase in Hbmass. This finding indicates that substantial Hbmass expansion can occur in normoxia, suggesting that training-related factors may interact with hypoxic exposure.

Finally, repeated altitude exposure appeared to accelerate adaptation: four swimmers showed no Hbmass increase after 3 weeks during a first camp, but significant increases after 3 weeks during a second camp performed one month later.

Taken together, these observations suggest that Hbmass responses depend strongly on training phase, cumulative exposure, and overall load management. The classification of athletes as “responders” or “non-responders” may therefore be overly simplistic. Regular Hbmass monitoring may help optimize altitude strategy and training individualization.

Relative Age Effect and Performance Bias in Italian Youth Road Cycling

Andrea Colombo

1. Introduction

The Relative Age Effect (RAE) refers to the over-representation of athletes born earlier in the competition year in sport participation (Dudink, 1994) and is documented across sports, particularly in youth. RAE is generally stronger in adolescence and may progressively attenuate with age (Jakobsson et al., 2021). RAE has also been documented in performance outcomes, potentially leading to selection biases related to differential maturation. Therefore, the aims of this study were: (1) to examine the presence and magnitude of RAE among successful athletes compared with the total population of registered athletes within the Italian Cycling Federation (FCI) from 2015 to 2025; and (2) to investigate differences in seasonal points according to birth quartile across age categories.

2. Materials and Methods

A retrospective analysis was conducted using official registration data from the FCI from 2015 to 2025. The sample included 34,764 male athletes (aged 13–18 years). Birth dates were categorized into quartiles: Q1 (Jan–Mar), Q2 (Apr–Jun), Q3 (Jul–Sep), and Q4 (Oct–Dec). Athletes were classified as successful if they achieved at least one top-5 finish in road races during the competitive season, corresponding to at least one point in the official FCI ranking system. The presence of RAE was assessed separately by age (13–18 years) using chi-squared goodness-of-fit tests comparing the distribution of successful athletes with the expected distribution derived from all registered athletes within the same birth cohorts. Effect size was quantified using Cramér’s V, and standardized residuals were inspected to identify specific over- and under-represented quartiles. Performance differences across birth quartiles were examined using non-parametric Kruskal–Wallis tests on seasonal points, stratified by age. When significant, pairwise post-hoc comparisons were performed using the Dwass–Steel–Critchlow–Fligner (DSCF) test with Holm adjustment. All analyses were conducted separately for each age. Statistical significance was set at p < 0.05.

3. Results

A significant RAE was observed among male athletes at ages 13, 14, 15, and 16. The strongest deviation from the expected distribution was observed at age 13 (χ² = 191.63, p < 0.001, V = 0.11), followed by age 14 (χ² = 154.62, p < 0.001, V = 0.10), age 15 (χ² = 86.40, p < 0.001, V = 0.08), and age 16 (χ² = 55.34, p < 0.001, V = 0.06). The effect was not statistically significant at ages 17 and 18. Standardized residuals indicated a consistent over-representation of Q1 athletes and under-representation of Q4 athletes in younger categories. Effect sizes were small, with a clear attenuation across age.

Significant differences in seasonal points were observed across birth quartiles in athletes aged 13 to 18 years (all p < 0.014) (Fig. 1). Post-hoc DSCF analyses revealed that differences were primarily driven by higher performance in Q1 compared with Q3 and Q4 in younger categories. At ages 13 and 14, Q1 athletes obtained significantly higher median points than Q3 and Q4 athletes (p < 0.001). A similar pattern was observed at ages 16, 17, and 18, with Q1 athletes consistently outperforming later-born peers. Median values generally followed a descending gradient from Q1 to Q4 in younger age groups.

4. Discussion

The presence of RAE among successful athletes compared to registered athletes is significant, although the size effect is small and, as reported in the literature, tends to decrease with age (Jakobsson et al., 2021). In line with the findings of Gallo et al. (2022) and Filipas et al. (2024), no significant RAE emerged at ages 17 and 18. However, RAE was evident in Italian road cycling from ages 13 to 16. Although effect sizes were small, a clear and consistent performance advantage was observed for relatively older athletes across all age categories. The points obtained consistently favored athletes born in the first quartile across the examined period. The magnitude of the effect progressively decreased with age; however, performance disparities were still evident at age 18 and should be carefully considered in annual talent identification and selection processes, as suggested by the results shown by Voet and colleagues (2022).

5. Practical Applications.

The results of the study contribute to a better understanding of how the day of birth could influence success in young cycling categories. For talent scouting, it may be misleading to base talent selection solely on current performance, as this may favor relatively older cyclists. Federations may need to consider restructuring youth competitions according to biological rather than solely chronological age.

Differentiating Training Stress and Non-training Stress Responses Using a Custom Stress–Recovery Questionnaire in Youth Elite Athletes

Friso van der Bijl

Overreaching and overtraining affect approximately 20–30% of youth high-level athletes, highlighting the importance of effective monitoring strategies. Stress–recovery questionnaires are widely used to detect changes along the overtraining continuum. Although validated sport-specific instruments exist, applied sport settings often rely on customized questionnaires to improve monitoring adherence and contextual relevance.

According to the stress–recovery model, both training-related and non-training stressors influence an athlete’s stress-recovery balance. Knowing which stressor impairs this stress-recovery balance provides valuable information on what recovery strategies to conduct. Several stress–recovery subscales (e.g., fatigue, vigor, well-being) have demonstrated sensitivity to changes in training stress. However, it remains unclear whether custom stress–recovery subscales respond differently to periods of high training stress compared with high non-training stress. Therefore, this study aims to examine the differential responsiveness of custom stress–recovery subscales during high training stress and high non-training stress conditions in youth elite athletes.

Twenty youth speed skaters (9 males, 11 females; age 18 ± 1 years) competing at national and international level were monitored during four distinct weeks: (1) baseline, (2) high training stress week (training camp), (3) high non-training stress week caused by academic, personal, or competition-related stress, and (4) taper week. Athletes completed a daily custom stress–recovery questionnaire consisting of eight items scored on a 9-point Likert scale.

Composite scores were calculated a priori: a physical perception of recovery score (muscle soreness, physical fatigue) and a mental perception of recovery score (motivation, stress/concerns, sharpness). Higher scores reflected higher perceived recovery. Sleep duration and sleep quality were assessed separately. Internal training load was quantified using the session-RPE method (RPE × session duration). Daily values were aggregated into rolling 7-day training loads and weekly mean scores. To examine within-subject changes over time, linear mixed-effects models were specified with week as a fixed effect and athlete as a random intercept. Primary outcomes were physical and mental perception of recovery scores.

This study provides insight into whether custom stress–recovery subscales differentially reflect high training stress versus non-training stress in youth elite athletes. Understanding such differential sensitivity may improve the interpretability and practical use of monitoring tools in youth athlete development. This may support coaches in selecting a training-related or non-training-related recovery strategy.

Modeling the power-duration relationship: the physiological and practical significance of the time-constant parameter

Marton Horvath

A conference paper has been submitted without a structured abstract. For the content, please find the attached document.

The Physiological and Perceptual Effects of the Menstrual Cycle on Performance in Female Cyclists

Meg Smith

The menstrual cycle (MC) is characterised by predictable fluctuations in oestrogen and progesterone, resulting in distinct hormonal milieus across the early follicular (EF; low oestrogen and progesterone), late follicular (LF; high oestrogen, low progesterone), and mid-luteal (ML; high oestrogen and progesterone) phases. Although these hormones exert systemic effects on cardiovascular, metabolic, and neuromuscular function, evidence for meaningful MC-related changes in exercise performance remains equivocal, largely due to methodological limitations. This study aimed to comprehensively examine the physiological and perceptual effects of MC phase on high-intensity exercise performance in trained female cyclists using a longitudinal, repeated-measures design with rigorous three-step hormonal verification (calendar tracking, urinary LH testing, and serum hormone analysis).

Six eumenorrheic Tier 2 cyclists and triathletes completed laboratory testing across three consecutive cycles. Following an initial monitoring cycle, exercise trials were performed in the EF, LF, and ML phases across two experimental cycles (n = 12 observations). Trials included steady-state exercise at 80%, 100%, and 120% of gas exchange threshold (GET), followed by a fixed work rate time-to-exhaustion (TTE) test at 80% delta. Substrate oxidation, blood lactate (BLa) responses, lactate clearance, and symptom severity were assessed.

TTE differed significantly by phase (p = 0.014), with longer performance in the LF compared with EF phase. Symptom severity was also lower in LF versus EF (p = 0.016). No phase-related differences were observed for carbohydrate or fat oxidation, peak or minimum BLa, or BLa clearance (all p > 0.10).

These findings suggest that while high-intensity exercise tolerance may be modestly enhanced during the LF phase, traditional physiological markers remain largely unchanged. Individual symptom monitoring may therefore provide greater practical utility than universal phase-based training prescriptions.

The impact of fatigue accumulation on performance during stage races in professional road cyclists

Andrea Giorgi

Cumulative fatigue influences performance in professional road cyclists during stage races. This study examined the interaction between internal and external factors on performance parameters using race data from seven male professionals cyclists collected over three seasons. Maximal rolling means for power output (PO) and heart rate (HR) were calculated across durations from 1 to 3600 s, and PO– and HR–duration curves were generated for grouped race days (Days 1–3 to 19–21). Mean PO and HR for 5-, 10-, and 20-min efforts were also analysed. For each effort, average PO, relative performance (% best), mean HR, PO/HR ratio, cumulative energy expenditure, and consecutive race days were computed. Data was analysed with mixed-effects linear models (REML).Race day groups significantly affected both HR (p<0.0001) and PO (p<0.01), confirming performance declines across stages. A higher PO/HR ratio was associated with superior 5-min (coef=0.5915), 10-min (coef=0.6524), and 20-min (coef=0.7299) efforts (all p<0.0001). Importantly, cyclists with higher PO/HR ratios maintained better 10- and 20-min performance as prior mechanical work and successive race days increased. Despite overall reductions in PO and HR across stages, riders able to preserve a higher PO/HR ratio under accumulated fatigue produced superior short- to mid-duration efforts, highlighting cardiovascular efficiency as a key determinant of performance during multi-stage competitions.

Performance Demands of Completing a First Grand Tour: A Case Study from the Giro d’Italia 2025

Aitor Alberdi-Garciandia

Introduction

Grand Tours are the highest-demanding format in professional road cycling, requiring riders to compete across 21 stages over three consecutive weeks. Although previous research has described the mechanical and performance demands of individual stages in Grand Tours, limited evidence is available on the cumulative external load and intensity distribution during a first Grand Tour.

Understanding these demands is particularly relevant for debuting riders, as preparation strategies must account not only for peak performance requirements but also for durability under prolonged accumulated fatigue. Quantifying total work performed, relative intensity, and effort distribution across three weeks of competition may provide valuable applied insight for coaches and performance staff preparing athletes for stage racing at the World Tour level.

Therefore, the purpose of this study was to characterize the external load and performance demands associated with completing a first Grand Tour, using a single-case analysis of a UCI World Tour cyclist during the 2025 Giro d’Italia.

Materials and Methods

A single-case descriptive observational study was conducted on a 22-year-old male UCI World Tour cyclist during his debut at the 2025 Giro d’Italia. Anthropometric characteristics assessed in May 2025 included body height (183 cm), body mass (65.8 kg), body mass index (19.7 kg·m⁻²), and sum of eight skinfolds (40 mm). Performance characteristics from May 2025 included a functional threshold power (FTP) of 395 W (6 W·kg⁻¹). The power-duration profile comprised 5 s: 1077 W (16.4 W·kg⁻¹), 1 min: 662 W (10.1 W·kg⁻¹), 5 min: 493 W (7.5 W·kg⁻¹), 10 min: 440 W (6.7 W·kg⁻¹), 20 min: 416 W (6.3 W·kg⁻¹), 60 min: 367 W (5.6 W·kg⁻¹), and 90 min: 341 W (5.2 W·kg⁻¹). Maximal oxygen uptake (VO_2max) was estimated at 83 ml·kg⁻¹·min⁻¹ using a validated equation (Sitko et al., 2021).

General race characteristics (stage distance, duration, elevation gain, average speed, and finishing position) were obtained from the open-access platform ProCyclingStats. Stage-specific performance data, including power output, cadence, workload, and intensity metrics, were extracted from the athlete’s publicly available Strava profile. External load variables included total duration (h:mm:ss), distance (km), work (kJ), energy expenditure (kcal), elevation gain (m), normalized power (NP), average power (AP), and maximal power. Relative intensity was quantified using intensity factor (IF), percentage of FTP, and W·kg⁻¹ values. Time-in-zone distribution was calculated across seven predefined power zones. Descriptive statistics (mean ± SD) were used to characterize stage demands and variability across the three-week competition.

Results

The cyclist completed all 21 stages of the Giro d’Italia 2025, accumulating a total race duration of 88 h 59 min 53 s over 3,533.8 km, with 51,691 m of elevation gain. Total mechanical work performed during the event was 79,331 kJ, corresponding to an estimated energy expenditure of 57,596 kcal.

Mean stage duration was 4:14:47 ± 1:26:27 (h:mm:ss), with an average distance of 168.3 ± 53.5 km and elevation gain of 2,461 ± 1,456 m per stage. Mean stage speed was 40.4 ± 4.2 km·h⁻¹.

Across stages, normalized power (NP) averaged 291 ± 28 W, corresponding to 4.43 ± 0.42 W·kg⁻¹, while average power (AP) was 251 ± 36 W (3.82 ± 0.55 W·kg⁻¹). Maximal power achieved during stages was 967 ± 117 W. Relative intensity was characterized by a mean intensity factor (IF) of 81 ± 8 and a mean percentage of FTP of 64 ± 9%. Average cadence across the competition was 89 ± 5 rpm.

Time-in-zone analysis demonstrated a predominantly pyramidal intensity distribution across seven predefined power zones based on FTP (395 W). Zone 1 (0-198 W; active recovery) accounted for 36 ± 13% of total race time. Zone 2 (199-276 W; endurance) represented 14 ± 3%, Zone 3 (277-356 W; tempo) 15 ± 4%, and Zone 4 (357-415 W; threshold) 16 ± 9%. Higher-intensity efforts were less prevalent, with Zone 5 (416-474 W; VO_2max) accounting for 9 ± 3%, Zone 6 (475-593 W; anaerobic capacity) 7 ± 2%, and Zone 7 (> 593 W; neuromuscular power) 3 ± 1% of total time.

Discussion

This case study characterizes the external load and performance demands required to complete a first Grand Tour. The findings indicate that successful completion was achieved through sustained high submaximal intensity rather than prolonged supra-threshold effort. Despite the considerable cumulative workload accumulated across three consecutive weeks, relative intensity remained predominantly below FTP, with a clear pyramidal distribution across predefined power zones.

The mean intensity factor observed across stages suggests that the principal challenge of a Grand Tour debut lies in repeatedly tolerating substantial volumes of moderate-to-high-intensity submaximal work amid progressive fatigue levels. The predominance of time spent in Zones 1 to 4, combined with comparatively limited exposure to higher-intensity zones, supports the notion that durability may be more decisive for completion than maximal aerobic or anaerobic capacity alone.

Funding: This research was funded by the Spanish Ministry of Science, Innovation and Universities (FPU24 predoctoral fellowship, reference FPU24/02003) and the Basque Government (grant number IT1726-22).

Conflicts of Interest: The authors declare no conflict of interest.

References

Sitko, S., Cirer-Sastre, R., Corbi, F., & López-Laval, I. (2021). Five-minute power-based test to predict maximal oxygen consumption in road cycling. International Journal of Sports Physiology and Performance, 17(1), 9-15.

A Modular Pressure-Sensing Saddle for Personalized Ergonomic Assessment in Cycling

Claudia Delgado SIMAO

Saddle discomfort and non-traumatic pelvic injuries remain prevalent among cyclists, despite ongoing advances in saddle design. The complex interaction between individual anatomy, posture variation, pedaling technique, and saddle geometry limit’s objective ergonomic assessment using conventional methods.[1] This study presents the development and validation of a modular, low-cost pressure-sensing bicycle saddle designed to capture dynamic and individual-specific load distributions during cycling.

A scalable array of flexible force-resistive sensors was integrated into a conventional bicycle saddle and connected to a microcontroller-based data acquisition system. The sensor layout was structured into functional regions targeting ischial support, pelvic stability, and anterior saddle pressure. Dynamic cycling tests were performed under controlled posture variations (90°, 45°, and 30° trunk angles), intentional pedaling asymmetries, and altered saddle height conditions. Sensor outputs were visualized using a simplified representation termed the 'pressure-on-saddle cloud map,' enabling rapid identification of dominant load regions.

Results demonstrated reproducible and posture-dependent pressure patterns, clear detection of pedaling asymmetries, and distinct individual load ‘fingerprints’ across participants. Discomfort conditions produced characteristic centralized high-intensity pressure zones, confirming system sensitivity to ergonomic misalignment. The proposed approach provides a practical and adaptable tool for personalized saddle assessment, supporting data-informed product design, injury prevention strategies, and athlete-specific bike fitting applications.

Acute Hemodynamic Responses to Three Work-Matched HIIT Training Protocols in Trained Cyclists

Görkem Aybars Balcı

Abstract:

Background: Cardiac output (Q), stroke volume (SV), and the arteriovenous O₂ difference (a−vO_2diff) together constitute the primary determinants of maximal oxygen uptake (VO_2max) and VO_2max is one of the physiological factors that determines endurance performance (Rønnestad et al., 2020). SV and heart rate (HR), which together determine Q, represent the central component of VO_2max, whereas the a−vO_2diff constitutes the peripheral component (Faricier et al., 2025; Joyner & Dominelli, 2021). Among athletes who exhibit similar VO_2max values, individualized limitations exist; one may have a greater need for central component adaptations, whereas another may require greater peripheral improvements (Buchheit & Laursen, 2013). Accordingly, for an athlete with a larger deficit in the central component, high-intensity interval training (HIIT) protocols that elicit and sustain higher peak Q and/or peak SV for longer durations may be considerably more effective. Current literature indicates that studies comprehensively examining acute central and peripheral hemodynamic responses to different HIIT modalities in trained cyclists remain limited. Therefore, this study aims to compare the acute central (Q or SV) and peripheral (a-vO_2diff) responses elicited by three distinct HIIT models and to determine which modality provides a greater acute stimulus to the respective physiological components.

Methods: Twelve trained male cyclists (Level 3-4) participated in this study (age: 29.7 ± 5.3 years; V̇O_2max: 57.8 ± 4.1 mL·min⁻¹·kg⁻¹; body mass: 69.8 ± 6.0 kg). Familiarization sessions were conducted to adapt the participants to the electromagnetically braked cycle ergometer (Lode Excalibur Sport, Groningen, Netherlands) and cardiac measurement system (Innocor, COSMED, Italy). VO_2max was determined via a ramp incremental test. Critical power and the highest constant-intensity that permits the attainment of VO_2max (I_HIGH) were determined using multiple constant-work-rate testing sessions. Acute hemodynamic responses were evaluated with three work- and effort-matched (RPE ≥ 19) HIIT models administered in a randomized order: Long HIT (4×3 min - 3 min; WR ~88% of PPO; Rec ~25% of PPO), Short-HIIT (1 series of 15×43.5 sec - 43.5 sec; work rate [WR] ~110% of PPO; recovery load [Rec] ~10% of PPO) and Decreasing Power Output HIIT (DPO-HIIT, 6×114.5 sec - 114,5 sec; WR starting at I_HIGH and decreased by 4 equal steps to the CP + 10%; Rec ~25% of PPO). Q, SV, and a−vO_2diff were measured using a valid and reliable non-invasive inert gas rebreathing method (N₂O_RB). Throughout the sessions, variables were recorded at four time points (ΔTime: 25%, 50%, 75%, and 100% of each bout). Data were analyzed using a two-way (Protocol × ΔTime) repeated-measures ANOVA with Greenhouse-Geisser corrections where appropriate. Bonferroni post-hoc analysis applied for pairwise comparisons.

Results: Two-way repeated-measures ANOVA revealed no significant main effect of the protocol for Q (F(2.000, 22.000)=1.966, p=0.178, ηp²=0.152), SV (F(2.000, 22.000)=0.790, p=0.446, ηp²=0.067), HR (F(2.000, 22.000)=0.620, p=0.547, ηp²=0.053), or a–vO_2diff(F(2.000, 22.000)=1.586, p=0.227, ηp²=0.126). The Protocol × ΔTime interaction was significant only for HR (F(2.467, 27.132)=4.561, p=0.014, ηp²=0.293). The interactions for Q (F(2.476, 27.232)=2.448, p=0.095, ηp²=0.182), SV (F(2.330, 25.629)=0.620, p=0.569, ηp²=0.053), and a–vO_2diff(F(3.043, 33.477)=2.314, p=0.093, ηp²=0.174) were not significant. Consistent with these effects, HR increased from Δ25 to Δ100 across all protocols, with a more pronounced mid-bout rise in Long-HIIT than in DPO-HIIT and Short-HIIT (p<0.001). Furthermore, Q exhibited a significant temporal (ΔTime) increase in each protocol (Long-HIIT: 22.56 ± 2.78 to 25.28 ± 3.58 L·min⁻¹ (p<0.05); DPO-HIIT: 22.51 ± 2.73 to 24.97 ± 3.73 L·min⁻¹ (p<0.05); Short-HIIT: 22.78 ± 3.03 to 24.34±3.42 L·min⁻¹, p<0.05). Although slight numerical increases were observed, SV remained statistically stable from Δ25 to Δ100 across all protocols (Long-HIIT: 131.1 ± 17.0 to 137.7 ± 18.3 mL·beat⁻¹, p>0.05; DPO-HIIT: 130.9 ± 14.6 to 137.5 ± 19.4 mL·beat⁻¹ (p>0.05); Short-HIIT: 130.4 ± 17.2 to 133.1 ± 16.0 mL·beat⁻¹, p>0.05), while a–vO_2diff remained statistically stable from Δ25 to Δ100 across all protocols (Long-HIIT: 17.26 ± 1.86 to 16.29 ± 1.95 mL·O₂·dL⁻¹ (p>0.05); DPO-HIIT: 17.08 ± 1.89 to 16.54 ± 1.40 mL·O₂·dL⁻¹, p>0.05); Short-HIIT: 16.37 ± 1.92 to 16.30 ± 1.76 mL·O₂·dL⁻¹, p>0.05).

Discussion/Conclusion: The absence of a main protocol effect for Q, SV, and a–vO_2diff aligns with the observation that, despite differences in interval structure, the three work-matched models elicited broadly comparable average central and peripheral responses. Notably, while the Protocol × ΔTime interactions for Q and a–vO_2diff did not strictly cross the p<0.05 threshold, they exhibited moderate-to-large effect sizes (ηp² = 0.182 and 0.174, respectively). In the present sample of trained athletes (n=12), these moderate to large effect sizes can be of practical importance. These findings indicate that peak-versus-mean hemodynamic responses diverge according to the interval format, where prolonged work intervals (i.e., Long-HIIT and DPO-HIIT) likely induce distinct temporal accumulations of central blood flow and peripheral extraction compared to micro-intervals (Short-HIIT). Consequently, while overall systemic stress remains matched, it may be beneficial for coaches and sports scientists to manipulate HIIT structure to precisely target either the central or the peripheral oxygen extraction kinetics, depending on the athlete's individualized physiological deficits.

Time Spent Above 90% and 95% of VO2max During Three High-Intensity Interval Cycling Protocols

Görkem Aybars Balcı

Abstract: Background: Maximal oxygen uptake (VO2max), together with the fractional utilization of VO2max and efficiency, are among the primary determinants of cycling endurance performance (Joyner & Coyle, 2007). Maximizing the accumulated time spent above 90% or 95% of VO2max appears to be particularly important when the objective is to enhance VO2max (Rønnestad et al., 2022; Midgley et al., 2006; Turnes et al., 2016). Additionally, exercising at intensities near VO2max is thought to impose substantial stress on the body’s oxygen transport and utilization systems. Therefore, time spent at a high percentage of VO2max is a key determinant of the adaptive response to interval training. It has long been recognized that training programs aimed at increasing VO2max and the time spent at near-VO2max intensities should include high-intensity interval training (HIIT) (Almquist et al., 2019; Storen at al., 2012; Hawley et al., 1997). However, the optimal composition and precise nature of HIIT remain disputed and continue to evolve. This study examines the effects of Short HIIT (S-HIIT), Long HIIT (L-HIIT), and Decreasing Power Output HIIT (DPO-HIIT) protocols on time spent at ≥ 90% of VO2max (t90%VO2max) and time spent at ≥ 95% of VO2max (t95%VO2max). Method: Level 3-4 male cyclists (n= 12), age 29.7 ± 5.3 years; body mass 69.8 ± 6.0 kg; V̇O2max 57.8 ± 4.1 mL·kg·min-1 volunteered for this study. Participants completed 8-9 laboratory visits. VO2max and peak power output (PPO) were measured using a ramp incremental cycling test. Critical power (CP) and anaerobic work capacity W′ were determined using three to four constant-load tests performed at 80–110% of PPO, each designed to elicit task failure within 2–15 min. The highest intensity permitting VO2max attainment (IHIGH) was determined from 2–4 constant work-rate time-to-exhaustion trials, with power output adjusted (±5%) between visits to identify the highest workload at which V̇O2max could be attained or maintained. IHIGH was defined as the highest workload yielding V̇O2 values consistent with the participant’s VO2max. After baseline tests, in a randomized crossover design, participants completed three HIIT exercise sessions: S-HIIT (1 series of 15 × 43.5 sec - 43.5 sec; work rate [WR] ~110% of PPO; recovery load [Rec] ~10% of PPO), L-HIIT (4 × 3 min - 3 min; WR ~88% of PPO; Rec ~25% of PPO), and DPO-HIIT (6 × 114,5 sec - 114,5 sec; WR starting at IHIGH and decreased by 4 equal steps to the CP + 10%; Rec ~25% of PPO). All HIIT sessions were work-matched (similar total work completed) and effort-matched (rate of perceived exertion: RPE ≥19). During all baseline tests and HIIT sessions, breath-by-breath pulmonary gas exchange (Innocor, COSMED, Italy), heart rate (HR) (Polar H10, Kempele, Finland), and RPE were recorded (Borg scale, 6-20). Blood lactate [bLa−] was collected at the end of the sessions (Biosen C-Line Lactate analyzer, EKF Diagnostics, Barleben, Germany). Across all work periods, VO2 values were calculated as 5-sec averages, and subsequently, t90%VO2max and t95%VO2max were quantified for each session. Variables were analyzed using a one-way repeated-measures ANOVA, with Bonferroni post-hoc analysis applied for pairwise comparisons. Effect sizes were calculated using Cohen’s d. Results: A one-way repeated-measures ANOVA revealed significant differences in accumulated t90%VO2max across the protocols (F=14.75, p<0.001, ηp²=0.573). L-HIIT elicited significantly greater t90%VO2max (580.4±61.8 s) compared to DPO-HIIT (511.7±77 s, p=0.025, d=0.891) and S-HIIT (421.7±90 s, p<0.001, d=2.058). There was no significant difference between DPO-HIIT and S-HIIT (p=0.070, d=1.167). Similarly, t95%VO2max varied significantly across protocols also varied across protocols (F=12.36; p<0.001, ηp²=0.529), with L-HIIT producing higher t95%VO2max (404.2±125.9 s) than DPO-HIIT (308.3±131.1 s; p=0.05; d=0.854) and S-HIIT (193.2±68.8 s; p=0.002; d=1.881), while DPO-HIIT did not differ strictly from S-HIIT (p=0.098; d=1.026). There was no significant difference in peak [bLa−] concentrations across HIIT modalities (p>0.05). Discussion/Conclusion: The L-HIIT protocol produced the greatest near-maximal oxygen uptake response, demonstrating the highest accumulated both t90%VO2max and t95%VO2max across the three HIIT sessions. This pattern indicates that, within the intervals performed with a 1:1 work-to-rest ratio, longer continuous work bouts provide the most potent acute stimulus for sustaining high fractions of VO2max. Accordingly, during the present conditions L-HIIT appears to be the most effective strategy for maximize time at near-maximal oxygen uptake during the present. Keywords: High intensity interval training, Cycling performance, Near-maximal oxygen uptake, Time-spent at VO2max

Effect of Cyclist Formation Size on Wind Speed and Direction

Chia-Hsiang CHEN

Purpose: To examine how different cyclist formation sizes influence wind speed and direction. Methods: A female national-level cyclist (24 years; 159 cm; 51 kg) rode on a 100-m indoor track at 20 ± 1 km/h. Wind speed and direction were measured at the lead and last riders using a custom three-hole Pitot tube with an anemometer (1 Hz). Four straight-line single-file formations (2–5 riders) with 1-m inter-rider spacing were tested, each repeated twice. Results: Increasing formation size progressively reduced wind speed for the last rider: ~30% for 2 riders, ~38% for 3 riders, and ~41% for 4–5 riders. Wind direction angles also shifted with formation size (1–8°). The four-rider formation offered the best aerodynamic balance. Conclusions: Formation size significantly affects the aerodynamic conditions for trailing cyclists, with four-rider formations providing optimal drafting benefits.

Acute Hemodynamic Responses to Three Work-Matched HIIT Training Protocols in Trained Cyclists

HAKAN ARSLAN

Abstract:

Current state of Safety Tracking and Race Monitoring Devices in Cycling

Simon Perneel

Recent race incidents in professional roadch cycling have renewed concerns regarding rider safety, with multiple high-profile crashes resulting in severe or long-term injuries. Data from ProCyclingStats show an increased frequency in reported rider injuries over the past years. These trends coincide with rising average race speeds, denser peloton dynamics, and increasingly complex road infrastructures, all of which contribute to elevated crash risk and more serious injury outcomes. As a result, the demand for effective technological solutions that support real-time monitoring, rapid incident detection, and coordinated race response has intensified.

In response, several safety tracking and race monitoring systems have been introduced in recent years, aiming to improve situational awareness for race organizers and enhance post-incident response times. However, despite notable technological progress, the current ecosystem of cycling safety trackers presents significant limitations that hinder widespread adoption and operational effectiveness. This abstract outlines the current landscape of safety tracking technologies in cycling, highlighting technical challenges, deployment constraints, and gaps relative to the practical demands of race environments.

One of the primary challenges is coverage reliability. Professional cycling races frequently traverse remote, rural, or mountainous regions where cellular connectivity is unstable or absent. Continuous, low-latency data transmission remains difficult to guarantee in such environments, limiting the effectiveness of real-time monitoring and automated alerting systems. Closely related is the issue of reliable crash or event detection. Many existing race monitoring solutions including systems developed by Live Insiders, 3BO, Swiss Timing and GeoDynamics prioritize rider positioning and performance-related data . While valuable for race management and broadcast applications, these systems are not primarily designed for robust incident detection, often relying on indirect indicators such as abrupt speed changes or signal loss, which can generate false positives or missed events.

Hardware design represents another critical consideration. Safety tracking devices must be unobtrusive, lightweight, and ergonomically integrated into existing equipment to avoid interfering with rider performance or comfort. Feedback from professional riders has consistently emphasized the importance of minimizing additional physical or cognitive burden during competition. Pilot testing conducted in collaboration with teams such as Soudal-Quick-Step underscored the necessity of incorporating cyclist feedback early in the design and validation process, particularly with regard to wearability, device placement, and perceived intrusiveness.

Finally, technical and operational requirements from race organizers play a decisive role in system viability. Solutions must integrate seamlessly with existing race infrastructure, comply with regulatory constraints, and offer clear value in terms of safety decision-making and emergency response coordination. Without standardized requirements and interoperable frameworks, the deployment of safety tracking technologies risks remaining fragmented and inconsistent across events.

In conclusion, while current safety tracking and race monitoring devices represent an important step toward improving rider safety in professional cycling, substantial gaps remain in coverage reliability, incident detection accuracy, hardware integration, and organizational alignment. Addressing these challenges will be essential to realizing the full potential of in-race safety technologies and reducing the human cost of competitive cycling.

Keywords: Cycling safety; Race monitoring; Real-time tracking; Incident detection

Timing of Sodium Bicarbonate Uptake During Exercise and Rest in Highly Trained Cyclists

Tilmann Strepp

Sodium bicarbonate (SB) has been extensively studied for its ergogenic effects in short-duration, high-intensity exercise (Grgic et al., 2021). More recently, SB has been increasingly used in endurance disciplines such as triathlon, road cycling, and cross-country skiing, to enhance performance during decisive, high-intensity efforts (Dalle et al., 2021). However, the influence of prolonged low-intensity exercise following SB ingestion on blood bicarbonate (HCO₃) uptake kinetics remains unclear. Therefore, this study aimed to assess the effects of low-intensity cycling compared with passive rest following SB ingestion on HCO₃ kinetics, including the timing and magnitude of peak HCO₃.

Fourteen highly trained cyclists (VO_2max: females n = 5, 64.7 ± 3.7 mL·min⁻¹·kg⁻¹; males n = 9, 74.5 ± 7.3 mL·min⁻¹·kg⁻¹) completed three laboratory visits separated by ≥48 h. Visit one involved an incremental step and ramp test to determine lactate threshold 1 (LT1) and VO_2max. On visits two and three, participants completed the exercise (EXC) and resting (REST) conditions, with identical time points for data collection. In both conditions, participants consumed a standardized breakfast at 08:00 AM, followed by SB ingestion (0.3 g·kg⁻¹) at 08:45 AM. During EXC, participants cycled indoors for three hours at LT1 (3.0 ± 0.3 W·kg⁻¹) starting at 10:00 AM, with 50 g·h⁻¹ carbohydrate intake. During REST, participants refrained from exercise and food intake. In both conditions, HCO₃ was measured every 25 min from 08:45 to 10:00, and every 30 min from 10:00 AM to 1:00 PM using a blood gas analyzer (i-STAT Alinity, Abbott, USA). All data were assessed for normality using the Shapiro–Wilk test. Differences in HCO₃ time course were analyzed with repeated measures ANOVA, correcting for sphericity violations with Greenhouse–Geisser when necessary. Post-hoc pairwise comparisons were Bonferroni-adjusted. Effect sizes were calculated using partial eta squared (_pη²). Peaks in HCO₃ were identified as the first point within ±0.6 mmol·L⁻¹ of the local maximum, corresponding to a 2% technical error of measurement at 30 mmol·L⁻¹ (Larcher et al., 2023), thereby capturing the initial plateau while minimizing measurement noise. Differences between peak and baseline (08:45 AM) concentrations (ΔHCO₃) and time to peak were compared between conditions using paired t-tests.

All participants completed the protocol without adverse events, such as gastrointestinal problems. Data were normally distributed. A significant condition × time interaction was observed (F(9,117) = 5.36, p < 0.001, _pη² = 0.29), indicating that the temporal pattern of HCO₃ differed between EXC and REST. Pairwise comparisons revealed higher HCO₃ concentrations during REST compared with EXC at 30 min (mean diff. = 1.2 ± 1.2 mmol·L⁻¹, p = 0.003) and 60 min (mean diff. = 2.9 ± 2.6 mmol·L⁻¹, p = 0.001) after 10:00 AM (exercise onset in EXC). No differences were observed at any other time point (all p > 0.05). A significant main effect of time was detected (F(9,117) = 197, p < 0.001, _pη² = 0.94), reflecting the expected increase in HCO₃ following SB ingestion. However, there was no significant main effect of condition (F(1,13) = 3.95, p = 0.068, _pη² = 0.23), indicating similar overall mean HCO₃ concentrations between EXC and REST. ΔHCO₃ did not differ between conditions (EXC: 9.0 ± 1.8 mmol·L⁻¹, CV 20%; REST: 9.3 ± 1.4 mmol·L⁻¹, CV 15%; p = 0.32). In contrast, time to peak HCO₃ was significantly longer during EXC (208 ± 42 min, CV 20%) compared with REST (146 ± 22 min, CV 15%; p < 0.001).

In summary, SB ingestion effectively elevated HCO₃, with peak concentrations being comparable between EXC and REST. However, low-intensity cycling delayed the time to peak by ~1 hour, emphasizing the importance of accounting for exercise during the absorption period when planning nutritional strategies for prolonged endurance events. These findings suggest that athletes and practitioners should adjust the timing of SB intake to align peak HCO₃ with critical performance periods during prolonged exercise.

Machine Learning for UCI Point predictions

Louis Van Gaelen

Prior to 2020, the UCI point system primarily served as a metric for prestige. The Pro-Continental Teams fought for automatic wildcards and the nations for extra spots at the World Championships or the Olympic Road Race. However, the 2020 regulatory overhaul transformed these points into a promotion and relegation system between the WorldTour and ProTeam levels, through three year cycles. Despite the rise of cycling analytics over the last decade, the prediction of team-wide or single-rider point accumulation remains under-represented in peer-reviewed literature. This study leverages public data and machine learning to predict the season total of UCI points a rider will collect. Our feature set incorporates rider history, team context and rider-race performance features of 19,359 riders across 10 seasons. These features are then used to fit a regression model, which can provide point prediction based on rider schedules. Our results suggest that predictive modeling can provide teams with critical strategic advantages in roster management and race selection within the current three-year cycle. They also open the door to result-driven transfer market strategies, including a points-per-cost analysis of possible transfers.

The Role of Athlete Profiles in Repeated-Sprint Training in Hypoxia for Elite Track Endurance Cyclists

Georgina Barratt

Repeated-sprint training in hypoxia (RSH) is increasingly used by elite athletes to improve repeated-sprint ability and power production^[1]. RSH involves performing a series of short “all-out” sprints with brief recovery periods^[2] in a hypoxic environment^[3]. This training stimulus was used in preparation for track endurance cycling events, specific the team pursuit, where cyclists require the ability to produce and sustain high absolute power and torque outputs^[4] across successive lead turns and laps. Despite its widespread use across the literature, limited research has examined how athlete phenotype influences session responses. This study investigates how sprint performance in RSH differ across athlete profiles.

Nineteen Australian track endurance cyclists completed field and laboratory tests to establish power-duration profiles and calculate anaerobic power reserve (APR). Athletes subsequently undertook a two-week RSH training block, comprising four sessions, each built around two key protocols: RSH-6 (4×5×6 s) and RSH-10 (3×5×10 s) in a normobaric hypoxia (FiO₂: 14.5%). APR z‑scores were used to classify athletes along an endurance-to-sprint spectrum within each sex, enabling individualised, phenotype-specific interpretation of sprint performance. Peak power output (PPO), sprint decrement score and within‑athlete variability (mean ± SD) were evaluated against each athlete’s APR z‑score for both protocols.

Male hybrid- and sprint-profiled athletes produced higher absolute PPO than endurance-profiled males; however, sprint-profiled males relied on a smaller percentage of their APR to achieve these PPO values. This suggests that athletes with larger anaerobic reserves or capabilities may not fully access or express that reserve during repeated-sprints and in hypoxic conditions, likely influenced by recruitment thresholds or premature neuromuscular fatigue. In females, absolute PPO did not differ meaningfully across profiles, yet sprint-profiled athletes produced lower relative PPO (%APR) than endurance- and hybrid-profiled teammates, suggesting they did not fully utilise their anaerobic capabilities during RSH sprints. Across sexes, sprint-profiled athlete demonstrated higher within-athlete variability, with standard deviations of 118–219 W for males and 69–73 W for females. In comparison endurance-profiled athletes demonstrated standard deviations of 22–47 W for males and 25–28 W for females. This suggests that sprint-profiled cyclists had a much great drop of in power over the session, while the endurance-profiled cyclists demonstrated low individual variation and therefore better repeatability of sprint power across each session. APR z-score showed a small positive correlation with sprint decrement score (males=0.32; females=0.46), supporting the observation that sprint‑profiled athletes experience greater fatigue accumulation and less repeatability across efforts.

Collectively, these findings suggest that athlete phenotype significantly affects sprint execution in RSH. Individualised training prescriptions, such as adjusting sprint volume, recovery and duration based on the athlete profile may enhance RSH effectiveness.

Collectively, these findings demonstrate that athlete phenotype strongly influences sprint execution and repeatability during RSH. Sprint‑profiled cyclists produce higher absolute PPO however, are likely to experience higher sprint decrements across repeated sprints, whereas endurance-profiled cyclists express power more consistently despite lower absolute PPO, supporting their greater repeated-sprint ability and their capabilities to recover effectively between sprints and sets. As a result, applying uniform RSH prescriptions across a team is unlikely to produce optimal adaptations for all athletes. Tailoring training variables such as sprint duration, number of repetitions, recovery intensity, and between‑set rest to the individual athlete’s phenotype may significantly enhance the effectiveness of RSH interventions.

In high-performance environments, it is not always possible to provide individualised session prescriptions and psycho-social impact of group training should not be underestimated. However, understanding that in team-pursuit cyclists, variations phenotypes and athlete physiology is required for different roles within the team. Considering these factors as part of the interpretation and to assess at an individual level is essential. Tracking within‑session power and torque variability offers sports scientists and coaches a sensitive indicator of readiness, especially for sprint‑profiled athletes who naturally display greater fluctuation in RSH PPO. This variability also gives athlete‑specific context for interpreting overall training load, enabling more targeted decision‑making for upcoming training sessions. These phenotype‑specific strategies may also assist in assigning roles within the team pursuit, as understanding who can sustain high repeated power and torque, who can deliver the greatest absolute power, and who best utilises their anaerobic capacity can inform optimal wheel order and lap contributions during racing. Overall, incorporating phenotype‑specific adjustments into RSH offers a practical framework to enhance individual development and optimise team performance in the team pursuit.

Bone Mineral Density in Elite Male Road Cyclists

Yan Brave

This cross-sectional study examined bone mineral density and cumulative racing exposure in 26 professional male cyclists. Osteopenia was identified in 27% of athletes. Cyclists with osteopenia demonstrated significantly longer professional careers and greater cumulative race days and total race distance, while body composition did not differ between groups. Years of professional exposure were moderately inversely correlated with BMD (r = −0.53, p = 0.005). These findings suggest an exposure-related association between prolonged participation in non–weight-bearing endurance cycling and reduced bone mass independent of global body composition.

A multiomic approach to characterize individual fatigue trajectories in professional cyclists during the Volta a Catalunya 2025

Laura Isus

This study integrated longitudinal urinary metabolomics (67 metabolites, 8 time points), ECG-derived autonomic markers, power meter data, and polygenic risk scores in seven professional cyclists during the Volta a Catalunya 2025. Cumulative analysis revealed 21 significantly altered metabolites (FDR < 0.05), predominantly cortisol-axis activation and tryptophan depletion. Anticipatory stress (baseline to pre-Stage 1) produced larger metabolic shifts than cumulative racing effects, while acute steroidogenic responses progressively attenuated across stages. PRS profiling revealed inter-individual heterogeneity in stress resilience, recovery, and energy metabolism. These findings support a multiomic approach to individualized fatigue monitoring in elite stage racing.

Training Load Control During Training Camp Using Biomarkers and Digital Monitoring

Yan Brave

This study examined hematological and cytokine responses during a 10-day triathlon training camp with individualized, threshold-based intensity prescription in amateur athletes. Eighty-four participants were enrolled, and a biomarker subcohort (n = 59) underwent serial blood sampling at baseline, mid-camp, and post-camp. Despite a substantial increase in training duration, perceived exertion remained stable. Leukocyte and neutrophil counts increased moderately but remained within reference ranges, with transient lymphocyte redistribution. Cytokine profiling showed reduced CXCL10 and interleukin-8 concentrations alongside increased interleukin-1 receptor antagonist and tissue-repair–associated mediators. Overall, the findings indicate adaptive immune modulation without excessive inflammatory activation, supporting the effectiveness of individualized load supervision during intensified training in amateur athletes.

Modeling Delayed and Non-Linear Training Load Effects on HRV: A Distributed Lag Non-Linear Modeling Approach in Endurance Athletes

Juan Ramon Gonzalez

Introduction:
Short-term heart rate variability (HRV) monitoring, often based on 7-day rolling averages of RMSSD, is widely used to assess recovery status in endurance athletes. However, this approach implicitly assumes immediate and linear load–response dynamics, potentially overlooking delayed and cumulative autonomic effects of training. Distributed Lag Non-Linear Models (DLNM) allow simultaneous estimation of non-linear exposure–response relationships and their temporal lag structure, yet have not been applied to individual HRV forecasting in sport science. This study evaluated the use of DLNM to model and predict daily RMSSD responses to training load.

Methods:
Daily RMSSD (outcome) and Training Stress Score (TSS; exposure) were analyzed in competitive endurance athletes. For each athlete, a DLNM with a maximum lag of 21 days was fitted to capture delayed and non-linear load effects. To assess practical monitoring requirements, time series were progressively truncated (15–240 days) and refitted using a 70/30 chronological split. The first 7 days of the test segment were used for out-of-sample prediction. Predictive performance was quantified using RMSE, MAE, and Pearson correlation between predicted and observed RMSSD.

Results:
DLNM identified structured lag-response patterns, with peak negative autonomic responses occurring several days after high training loads rather than on the same day. Prediction accuracy improved substantially once ≥120 days of monitoring were available (mean MAE ≈ 44 ms), with further stabilization beyond ~150–180 days (MAE ≈ 15–18 ms). Shorter monitoring windows (≤90 days) yielded unstable and imprecise forecasts. These findings contrast with common 7-day rolling approaches, which do not account for distributed lag structures and may underestimate cumulative load effects.

Discussion:
DLNM provides a physiologically coherent framework to model delayed autonomic responses to training, challenging the widespread reliance on short rolling averages. Reliable individual HRV forecasting appears to require several months of longitudinal data to adequately characterize the athlete-specific lag structure. Incorporating distributed lag modeling into sport analytics may enhance individualized recovery monitoring and decision-support systems.

Enacting Coach–Athlete Relationships in Remote Endurance Coaching: A Longitudinal Study

DOUGLAS STEWART

Conference Abstract for Science and Cycling 2026 Conference

Data-Driven Computational Framework for Personalized Training Plans in Cycling

Tom Van Deuren

Context: The widespread adoption of mobile power meters and heart rate monitors has transformed professional and amateur cycling into a highly data-driven discipline. Despite this massive collection of longitudinal sensor data, day-to-day training prescription still relies heavily on manual expert judgment rather than formalized mathematical optimization. While the predictive modeling of athletic performance has made significant advances, the transition from descriptive monitoring to automated, data-driven training prescription remains an unrealized opportunity in endurance sports. Methods: To address this gap, this study introduces a modular computational framework designed to automate the generation of personalized cycling training plans. First, we extract a continuous performance metric, PO@HR_target, utilizing longitudinal submaximal power output and heart rate data. Specifically, this metric filters historical training data for efforts within a personalized target heart rate range and employs linear interpolation to normalize the power output to this target heart rate, bypassing the need for disruptive and infrequent maximal laboratory testing. Second, we extract cycling-specific features from daily training data to capture the dose-response relationship, quantifying training load and intensity distributions through metrics like Training Stress Score, Acute and Chronic Training Loads, Chronic Intensity Load, and the Polarization Index. Third, we model the relationship between these training inputs and performance outputs using four diverse regression architectures: Lasso, Random Forest, XGBoost, and a novel tree-based model called Random Forest Featuring Linear Extensions (RaFFLE). These models are rigorously evaluated using 10-fold cross-validation, incorporating a seven-day buffer to combat data leakage. Finally, we apply a differential evolution algorithm to invert these predictive models, generating optimized training prescriptions, defining weekly durations, training load increases, and time-in-zone distributions, designed to maximize future athletic performance. Results: Validation against laboratory testing confirmed that the submaximal PO@HR_target metric effectively tracks true physiological trends in the first lactate threshold. During the predictive modeling phase, the models achieved an average out-of-sample Mean Absolute Error ranging between 5 and 10 W, which aligns closely with the intrinsic ±1.5% measurement uncertainty of commercial power meters. Furthermore, feature importance analysis revealed that chronic training load (e.g., CTL₅₆) consistently emerged as the dominant predictor of performance across all model architectures. Our analysis also revealed significant inter-subject variability across cyclists; while linear models like Lasso accurately predicted performance and provided the lowest test error on average (7.17 ± 2.70 W), other athletes required complex non-linear tree-based ensembles to accurately capture the dose-response relationship. During the prescriptive phase, the optimization algorithm demonstrated a potential average performance increase of 5.5% when generating individualized training schedules targeting the PO@HR_target metric. Conclusion: This framework successfully bridges the gap between predictive analytics and prescriptive decision-making in endurance cycling. By combining robust feature extraction from daily training data with advanced optimization techniques, it supports the automated generation of training plans that adapt to an athlete's evolving physiology. Ultimately, this methodology offers a scalable, sport-agnostic pathway for coaches and athletes to transition from retrospective data analysis to individualized performance optimization based on historical training data. Furthermore, future iterations of this framework should examine the relationship between this performance metric and daily recovery indicators, such as heart rate variability (HRV), sleep quality, and fatigue, to better account for daily fluctuations in an athlete's capacity to execute a prescribed plan.

Effects of small cadence variations on joint mechanics in highly trained track cyclists

Chantelle du Plessis

A track bicycle’s gear ratio (and consequently, cadence) can influence muscle force production, shortening speed, and activation rate, ultimately affecting joint-specific power transfer to the crank, energy cost and fatigue. However, the impact of small, practical cadence adjustments on joint mechanics during fatiguing, race-specific intensities in the aerodynamic position remains unknown. Understanding how joint power is distributed during pursuit-specific workloads may reveal strategies cyclists use to sustain high power under fatigue. This study examined the effects of small cadence variations (preferred, ±5 rpm) on joint mechanics (power, angular velocity, and moment) using statistical parametric mapping to identify where within the pedal stroke differences occurred, and to explore the interaction between joint torque and angular velocity in generating joint power. Thirteen highly trained track cyclists completed three 2-min constant work rate trials at individual pursuit race intensity. Higher cadences (+5 rpm) increased hip and knee joint moments concentrated around the crank cycle’s transition phases (i.e. top and bottom dead center) where higher angular velocities demanded greater neuromuscular control. Ankle moment differences were larger than hypothesized, decreasing during the downstroke and increased during upstroke, suggesting a key role in managing faster proximal-segment angular velocities through stabilization and transmitting forces to the crank. These findings highlight the value of time-series analysis for revealing cadence-specific biomechanical adaptations that are typically overlooked by discrete metrics, highlighting the importance of individualized cadence selection in pursuit cycling.

The Spinal column protection system for cycling races

Francesco Enrico Ragazzini

Keywords: spinal column protection, safety, cycling

1. Introduction

In Italy in 2024, a study published in Quotidiano Nazionale, which analyzed Hospital

Discharge Records (SDO - Schede di Dimissione Ospedaliera) from the Ministry of

Health, reported that in a sample of approximately 12,000 spinal trauma accidents, 81% of the

incidents were recorded in cyclists. Of these, around 1,500 cases involved the cervical spine.

Spinal injuries in cyclists can lead to severe and permanent consequences. The mechanisms of

bicycle-related trauma - high-speed falls, frontal or rear impacts, or torsional forces - can result

in flexion - compression fractures, burst fractures, distraction injuries, or rotational injuries,

with varying degrees of instability and potential neurological involvement.

Aim of study: to present an advanced spinal column protection system specifically designed for

professional road cycling and MTB.

2. Materials and Methods

Engineer + Medical

Subject: professional riders (ROAD + MTB)

3. Results

4. Discussion

Crash in the races and safety

5. Practical Applications

Races

6. Conclusions

Working progress

Predicting self-reported race-readiness and self-efficacy on race day: A machine learning approach

Felix Imbery

It is assumed that the training load in the week before a race is linked to an athlete's readiness and self-efficacy. In order to induce performance adaptations, training must provide an overload stimulus. However, without adequate recovery time this can lead to accumulated fatigue and temporary performance decrements. Balancing training stress with recovery therefore is central to optimising performance on race day.

Objective training load metrics are routinely used to inform training plan decisions and tapering strategies, but performance on race day is not solely determined by physiological load alone. Little is known about the extent to which objective training load metrics can predict self-reported race-readiness and self-efficacy on race day in professional cyclists.

Therefore, this study aimed to examine whether objective training load metrics and short-term load dynamics can predict self-reported race-readiness (pre-race) and self-efficacy (post-race) in professional cyclists.

Twelve professional male cyclists of a UCI ProTeam participated over a total of X race days. The power output of the riders has been measured during training and racing using a crank-based power meter. Information about self-reported race-readiness and self-efficacy was collected on race day using two questionnaires, each comprising two items, derived from the Recovery-Stress Training Questionnaire and the Short Recovery and Stress Scale. The pre-race questionnaire was presented to the participants prior to the race, and the post-race questionnaire was presented to them after the race. Scoring of the questions was on a 7-point Likert scale, with higher scores reflecting higher race-readiness and self-efficacy.

All training and race data were processed using a structured, Python-based workflow (Python 3.11.7), which integrated multiple scientific computing libraries. Objective training load metrics were calculated and rolling aggregates (1-, 2-, 3-, and 7-day windows) were derived in order to quantify short-term load dynamics. High-load days (>2000 kJ) and race participation (yes/no) during the seven days preceding race day were included as binary features. For each rider a chronic 28-day pre-race baseline has been defined to contextualise acute load relative to habitual training. Statistical analyses were conducted to enable both linear mixed-effects modelling on group and individual level with the composed feature set as fixed effect and rider ID as random effect. Primary outcomes were the questionnaire scores. Further, two machine learning approaches have been applied on the same features.

This study advances understanding of the relationship between objective training load metrics and race-ray psychological states in high-performance cyclists, informs tapering decisions and supports coach decision-making.

Accurate video tracking of indoor cycling races

Philippe Odier

We present a computer vision method combining accurate field calibration and innovative localization algorithm, enabling to provide automatically the precise trajectory of cyclists on a track race hence allowing quantitative analyses, including power production estimates based on the measured trajectory combined with a physical model.

Automated workflow for 4D scanning cyclists in motion

Kobe Hermans

Recent advances in human body modeling have enabled high-resolution reconstruction of articulating subjects in motion. Systems such as MOVE4D (Instituto de Biomecánica de Valencia) demonstrate the capability to generate watertight 3D meshes of moving humans. However, when a subject interacts with an object—as in cycling—conventional homologous registration approaches fail to produce accurate watertight models due to geometric interference between the body and equipment (Van Gastel & Verwulgen, 2022). This limitation has significantly constrained the study of object-assisted movement and dynamic aerodynamics.

In this study, we present a fully automated 4D scanning workflow specifically designed to address this challenge for cyclists in motion. The system captures high-resolution, time-resolved point clouds (±1 mm, up to 178 Hz) over complete pedal cycles and synchronizes them with multiple external force sensors. The principal advancement is a dedicated pointcloud processing pipeline that isolates the cyclist from the bicycle without requiring intrusive physical modifications or restrictive scanning conditions.

The processing workflow comprises four main steps: (1) detection of the bicycle’s position and orientation using three colored reference markers; (2) alignment and insertion of a CAD-generated bicycle model into the captured point-cloud scene; (3) removal of all points within a defined proximity of the virtual bicycle model; and (4) cluster-based filtering to retain only the largest remaining geometry, corresponding to the human subject. The resulting clean 4D point clouds are subsequently converted into watertight models, eliminating geometric interference between body and equipment while preserving temporal consistency across frames. Synchronization with pedal and saddle force measurements enables precise temporal coupling between external loading and body surface kinematics.

The developed workflow transforms dynamic cyclist reconstruction from a constrained experimental setup into a scalable, repeatable methodology. It enables rapid generation of watertight, force-synchronized 4D cyclist models suitable for quantitative biomechanical analysis (De Rosario et al., 2023). This technological foundation opens opportunities for bike fitting, movement stability assessment, fatigue monitoring, equipment and garment development (Yang et al., 2022), and time-resolved CFD simulations for the instability of cyclists (Winter et al., 2025). Moreover, the workflow is extendable to other forms of object-assisted human movement beyond cycling.

Comparing Absolute (kJ) and Relative (TSS) Work Load Metrics for Assessing Durability - Relationships to Real World Cycling Performance

George Evans

Introduction

The ability to avoid or minimise the downward shift in the power–duration relationship has been termed an athlete’s durability and is increasingly recognised as an important determinant of performance. In cyclists, durability is typically assessed by comparing parameters derived from the power–duration relationship in a fresh versus fatigued state. To standardise the fatiguing protocol between participants, studies generally prescribe either absolute loads (such as set kilojoule targets e.g. Mateo-March et al. (2022)) or relative loads (such as a set duration at the same relative intensity e.g. Clarke et al. (2018)). Matching absolute work results in athletes of different abilities completing the task in different durations, whereas matching time and relative intensity leads to different amounts of mechanical work being accumulated. The differences in how the protocol is standardised may have implications for how each approach is best applied. For example, absolute load approaches may better reflect the pressures of racing, where athletes must tolerate a given workload regardless of how long it takes or the relative intensity, making them potentially more useful for talent identification; whereas relative load approaches anchored to cyclists thresholds, such as training stress score (TSS) level the playing field and allow tighter experimental control, making them potentially more suitable for tracking individual durability over time. Despite these conceptual differences, studies have yet to directly compare the two methods and examine how the resulting durability estimates relate to real world race performance. Therefore, the aims of this study are to determine whether durability estimates differ when cyclists’ power profile parameters are compared before and after completing a range of relative and absolute loads, secondly to evaluate the extent to which each durability measure is associated with cycling performance.

2. Materials and Methods

2.1 Participants

Cyclists attending the North Wales Horseshoe Pass Road Race (n=80) will be invited to take part in the study. All participants must have at least 12 months of power meter verified training data available for analysis. This study received ethical approval by the Health, Science, Engineering and Technology Ethics Committee with Delegated Authority at the University of Hertfordshire.

2.2 Methodology

CP and W` will be estimated for each cyclist under fresh conditions and after exposure to graded absolute and relative workloads. CP, W` and workload bins (kJ, kJ•kg^-1 & TSS) will be derived retrospectively from all training rides and races completed in the 365 days preceding the event.

Durability will be quantified as the change in CP and W′ between the fresh condition and each workload condition. Race performance will be defined by their finishing time.

2.3 Data analysis

Maximum mean power (MMP) data from the preceding 365 days will be extracted using commercially available software (Training Peaks and WKO5) to model CP and W′ under each workload condition.

A repeated‑measures ANOVA will be used to assess differences in CP, W` and durability across workload conditions. Where significant effects are detected (p ≤ 0.05), Bonferroni‑corrected paired‑samples t‑tests will identify specific between‑condition differences. Associations between CP, W`, durability metrics (change in CP and W` under each workload condition) and race performance will be examined using linear regression models. Race performance will be entered as the dependent variable, with CP, W` and durability metrics included as predictors. Standardised regression coefficients (β), confidence intervals and model fit statistics will be reported to quantify the strength and direction of relationships.

3. Results

The race will take place on 17 May 2026. Data processing and statistical analysis will begin immediately following the event once all performance and training history data have been collated.

Conclusions

This study aims to clarify whether durability estimates derived from absolute (kJ’s) and relative workload (TSS) conditions produce meaningfully different insights into cyclists’ performance capacities. By examining how each durability measure relates to race outcomes, the findings may help determine whether one approach offers greater ecological validity or predictive value. Such evidence could inform both research practice and applied talent identification by indicating which method better captures performance relevant fatigue responses.

Intra-Unit Test–Retest Reliability of the COSMED K5 During Submaximal and Maximal Cycling Exercise: Implications for Longitudinal Monitoring

Tim Lutz Kracht

Abstract: The purpose of this study was to quantify the intra-unit test–retest reliability of the COSMED K5 system during repeated submaximal and maximal cycling exercise under applied laboratory conditions. Thirty-three endurance-trained adults completed three identical cycling tests within 10 days (Figure 1), consisting of three consecutive 5-minute submaximal stages at fixed relative intensities of 2.0, 2.5, and 3.0 W·kg⁻¹, followed by a 5-minute recovery at 1.0 W·kg⁻¹ and a maximal incremental ramp test increasing at 25 W·min⁻¹ until volitional exhaustion (Figure 2).


Figure 1. Overall study design with three identical test sessions separated by at least 48 hours and completed within 10 days.


Figure 2. Exercise protocol illustrated for a 70 kg participant, consisting of three submaximal stages (2.0, 2.5, and 3.0 W·kg⁻¹; 5 min each), a 5-min recovery at 1.0 W·kg⁻¹, and a maximal incremental ramp test (increasing by 25 W·min⁻¹).

Oxygen uptake, carbon dioxide production, and ventilation were measured continuously in mixing chamber mode using the same K5 unit across all sessions. Reliability was assessed using typical error (TE), coefficient of variation (CV), intraclass correlation coefficient (ICC), smallest detectable change (SDC), mean absolute error (MAE), mean absolute percentage error (MAPE), and Bland–Altman analysis.
Oxygen uptake and carbon dioxide production demonstrated good to excellent reliability across all workloads, with CV below 5% and ICC ranging from 0.92 to 0.96. Ventilation showed moderate to good reliability, with CV between 5.0% and 6.3% and ICC between 0.83 and 0.92. Reliability improved consistently with increasing exercise intensity across all three variables (Figure 3).


Figure 3. Intensity-dependent changes in coefficient of variation (CV) for oxygen uptake (VO₂), carbon dioxide production (VCO₂), and ventilation (VE) across submaximal stages and peak exercise.

For peak oxygen uptake, mean values were 4108 ± 707 mL·min⁻¹ (~58 mL·kg⁻¹·min⁻¹), with excellent relative reliability (ICC = 0.96, 95% CI: 0.92–0.98), a CV of 3.34%, and low absolute error (MAE = 86 mL·min⁻¹; MAPE = 2.1%). No systematic bias or time-dependent drift were observed across repeated measurements. However, the SDC for peak oxygen uptake was 397 mL·min⁻¹ (95% CI: 319–499 mL·min⁻¹).
These findings indicate that the portable metabolic system provides stable and reproducible gas-exchange measurements under standardized conditions. Nevertheless, despite high relative reliability, the magnitude of the SDC limits the sensitivity for detecting small individual changes in peak oxygen uptake. Consequently, small longitudinal changes should be interpreted cautiously when using peak oxygen uptake for monitoring performance in applied endurance sport settings.

A simple and accurate wearable device to measure ventilatory threshold: principles and validation

Pierre Jallon

Training zones are key for training and yet, there are only limited tools on the market to make accurate measurements. Coggan and CPxx tests lack accuracy, lactate tests require blood samples, are sensitive to food intake and require protocols with long steps, and ventilatory threshold require expensive, bulky and complex laboratory machines. In this paper, we present a novel device to make measurement of training zones based on ventilatory threshold easy and accessible. The measurement principle of the device is presented as well as a validation study.

Cognitive performance of healthy volunteers and professional cyclists during high intensity cycling

Wim Notebaert

Are an elite pursuit cyclist’s anthropometrics, flexibility and standing posture associated with their cycling position?

Daniel Williams

N/A

A mobilization index for intermittent work: combining breathing frequency and heart rate for training intensity monitoring.

Thomas Lagoute

Background: Heart rate (HR) remains the most widely used indicator for monitoring athletes’ internal load during training sessions. However, limitations appear during short work-to-rest intervals. First, HR kinetic is inherently slow, with peak values occurring several seconds after the onset of exercise. Second, HR often reaches a ceiling effect: once athletes approach an average HR, further increases in exertion are no longer reflected in the HR signal. Breathing frequency (BF), on the other hand, is closely associated with rating of perceived exertion (RPE) and is regulated by rapid mechanisms, including type III/IV muscle afferent feedback and central command, making it highly responsive. These characteristics suggest that BF may be more sensitive than HR to the dynamic stress experienced during interval training. Combining information from HR and BF could therefore provide a more complete picture of the athlete’s internal response. The aim of this study was twofold: (1) to compare HR and BF responses during a high intensity interval session, and (2) to explore the potential of combining the two metrics into a single index — termed the mobilization index — to better reflect perceived training stress. Methods: Thirteen male athletes trained in cycling and running participated in the study, comprising four visits to the laboratory. During the first two visits, participants completed maximal incremental running and cycling tests in a same session. The first visit included a facemask for gas exchange measurement, whereas the second did not. Third and fourth visits consisted of sessions of a 5 minute warm up, followed by 40 minutes of low intensity work at 90% of the first ventilatory threshold (VT1), then a ramp protocol with one-minute steps to maximal exertion. Finally, participants performed two sets of ten 40-second work intervals interspersed with 20 seconds of passive recovery. A 5-minute recovery was observed between each part of the session. Throughout all sessions, HR was recorded via ECG and BF via a strain-measuring belt. Heart rate reserve (HRR) and breathing frequency reserve (BfR) were calculated using Karvonen’s formula. A mobilization ratio was computed for each interval as BfR/HRR. To describe the kinetics of maximal BF and HR values reached throughout the intervals, three models were tested: linear, exponential, and exponential with a drift constant. A linear mixed model was fitted to the mobilization ratio. The final value of the mobilization ratio in each set was tested using a one sided hypothesis t-test, aligned with the expectation that the ratio would exceed 1 at the end of the session. Results : Maximal BF during the ramp did not differ significantly between running and cycling (Δ_{cycling-running}= 2 bpm; p=.43) but did during submaximal effort (Δ_{cycling-running} = -10 bpm; p=.03). None of the participants reached maximal BF during mask on trials. In contrast, eight athletes reached maximal HR during the masked ramp test. Linear models yielded the lowest Akaike Information Criteria for both HRR and BfR, suggesting that a linear description best captured their progression across intervals. The effects of interval number and set number were greater for BfR (β_interval=0.067; β_set =0.028) than for HRR (β_interval =0.010 ; β_set =0.013). In exponential models, asymptote was lower for HRR (A_estimate=92%) than for BfR (A_estimate=150%), and growth rates were higher for BfR. In the drift adjusted exponential model, the rate of increase was not significant when modelling BfR (p = .58). Modelling results suggest that both HR and BF tend to increase linearly throughout repeated high intensity intervals, yet BF displays a steeper and more variable increase than HR. While HR frequently plateaus, BF continues to rise, matching the progressive nature of rate of perceived exertion (RPE) during such sessions. This supports the interpretation of BF as a sensitive marker of real time internal load. Across interval sessions, athletes exercised at approximately 139 ± 5% of power at VT2. No difference in blood lactate (p = .91) or RPE (p = .19) was observed between the two interval sets. The mobilization ratio exceeded 1 at the end of the second set (M = 1.05, p = .003), but not at the end of the first (M = 0.90, p = .95), supporting the idea that the ratio could reflect the accumulated stress throughout the session. Conclusion: Breathing related metrics are often examined to derive metabolic thresholds; however, BF also offers valuable insights into internal load monitoring, particularly during short intervals where HR may not fully reflect rapid changes in exertion. Practitioners should calibrate BF training zones separately for each modality (running vs. cycling), and these zones should be based on maximal BF values obtained without a mask, assuming a reliable BF monitor is used. A mobilization ratio (BfR /HRR) exceeding 1 may indicate that the athlete has reached a meaningful level of physiological stress, providing coaches with a simple yet informative marker.

Do Women and Men Fail Differently? DNF Risk Prediction in Self-Routed Ultracycling (TCRNo10–No11)

Juan Francisco Garamendi Bragado

We analyse whether women and men differ in Do-Not-Finish (DNF) risk in self-routed ultracycling using Transcontinental Race (TCR) No10–No11 replay tracking. Men rode faster when moving (~10%) but DNF survival curves showed minimal separation by sex. Stopping strategies did not differ. In post-48 h landmark Cox models, early stopped time (within 48h) strongly predicted later DNF while higher moving speed was protective; sex had no effect. Route-aware distance analysis supported these conclusions. Early pacing and stop management (regardless the sex) dominate DNF risk.

Validation of unsteady aerodynamic effects under real-world apparent flow conditions using high-frequency wind sensing and cycling performance data

Christoph Feichtinger

Aerodynamic drag is the dominant resistance term in road cycling at moderate to high speeds, yet real-world assessment remains challenging because the flow incident on the rider–bike system is disturbed, unsteady, and strongly route dependent. In controlled environments, cycling aerodynamics is typically characterized through steady wind-tunnel measurements and steady Computational Fluid Dynamics (CFD). However, agreement between CFD and wind-tunnel data does not necessarily translate into accurate performance prediction outdoors, where riders experience rapidly varying apparent wind due to meteorological variability, terrain effects, traffic-induced disturbances, and yaw-angle changes from steering and cornering. This motivates the need for a validation approach that explicitly represents the outdoor wind state and its unsteadiness.

We present a test and validation toolchain designed to bridge this validation gap by correlating steady-state aerodynamic maps with time-resolved measurements recorded in real riding conditions. The central contribution is demonstrating that comparison-ready aerodynamic descriptors can be reconstructed from standard cycling performance signals when these are combined with enhanced, high-frequency apparent wind sensing. Apparent flow is measured and expressed through apparent wind magnitude $v_{\infty}$ and apparent yaw angle $\beta_{app}$. Inflow unsteadiness is quantified using segment-based Gust Intensity (GI) and Turbulence Intensity (TI), enabling outdoor datasets to be compared on the basis of both mean flow state and fluctuation level.

A representative validation use case combines Windpuls apparent wind sensor logging with common bike-computer data (e.g., Garmin, Wahoo), including power, speed, cadence, and altitude/grade information, alongside air-density inputs derived from temperature, pressure, and humidity. Data streams are time-aligned and analyzed in route- or lap-based segments. Measured mechanical power is converted to a conservative resistive-power signal by removing gravity and inertia contributions, and is then normalized using dynamic pressure based on measured apparent wind to compute a coefficient-like metric suitable for cross-condition comparison. This normalization yields a time-resolved, comparison-ready resistance descriptor that can be aggregated within yaw bins to reconstruct an outdoor yaw-sensitivity curve.

We demonstrate the pipeline for two operating environments: (i) road riding, which captures realistic variability in wind and terrain, and (ii) an outdoor velodrome, which provides repeatable geometry to isolate and interpret wind-driven effects under controlled lap structure. Across repeated passes, reconstructed yaw sensitivity remains consistent over a wide apparent yaw-angle range. Increased scatter at higher yaw angles is primarily attributable to quantified inflow unsteadiness (higher GI/TI) and reduced sample density at large yaw, rather than to a breakdown of the reconstruction approach. By reporting yaw distribution together with GI and TI, the framework provides a principled basis for selecting comparable subsets of outdoor data and for avoiding over-interpretation of day-to-day variability.

Overall, the pipeline supports rapid equipment or position comparisons in field tests while transparently documenting the wind-state context. The approach offers a scalable pathway to evaluate performance-relevant cycling aerodynamics under realistic, dynamic flow environments and to connect outdoor findings back to steady-state maps from CFD and wind-tunnel testing.

Validation of unsteady aerodynamic effects under real-world apparent flow conditions using high-frequency wind sensing and cycling performance data

Christoph Feichtinger

Real-world cycling aerodynamics are governed by disturbed and unsteady apparent wind fields that are only partially represented in controlled environments such as wind tunnels and steady Computational Fluid Dynamics (CFD). As a result, agreement between CFD and wind-tunnel measurements does not necessarily translate to accurate performance prediction outdoors. This work presents the simonaero test and validation toolchain, developed to bridge this validation gap by correlating steady-state aerodynamic maps with time-resolved measurements recorded in real riding conditions.

The central contribution is demonstrating that comparison-ready aerodynamic descriptors can be reconstructed from readily available cycling performance signals when combined with enhanced, high-frequency apparent wind measurements. Apparent flow is quantified through wind magnitude and yaw angle, and inflow unsteadiness is quantified using segment-based Gust Intensity (GI) and Turbulence Intensity (TI). A validation use case combines Windpuls sensor “airdrop” logging with common bike-computer data (e.g., Garmin, Wahoo), including power, speed, cadence, altitude/grade, and air-density inputs (temperature, pressure, humidity). The methodology is additionally demonstrated on an outdoor velodrome to validate the pipeline under repeatable lap conditions.

Results show that reconstructed yaw sensitivity is consistent over a wide apparent yaw angle range. Additional scatter at higher yaw angles can be explained by quantified inflow unsteadiness. The presented framework supports rapid comparison of equipment or position changes in real-world tests and provides a scalable pathway to evaluate performance-relevant aerodynamics under realistic, dynamic flow environments.

Building an End-to-End Incident Management Platform for Elite Level UCI Races

Robbe Decorte

Elite-level road cycling races operate in complex, dynamic, and high-risk environments. With race convoys spanning multiple kilometers, high rider speeds, variable terrain, and dense roadside infrastructure, effective incident management is critical for long-term safety improvement. Incident data has historically been fragmented across commissaire reports, broadcast footage, team communications, and post-race medical summaries. In this research, we present the design and implementation of an end-to-end incident management platform used by UCI and SafeR officials that integrates operational, technical, medical, and public data streams into a unified system to support decision-making, safety analytics, and preventive screening for future race editions. For instance in the 2025 season, 640 incidents were recorded involving 802 distinct riders. 88% originate from uci officials vs. 12% sourced from social media. At the core of the platform is structured incident generation by UCI race commissaires operating in-race vehicles. Using Veloviewer as a geospatial interface, commissaires log incidents in real time, tagging GPS coordinates, and inferring race context (stage, sector, kilometer marker), riders or vehicles involved, and preliminary categorization (e.g., crash, obstruction, rule infringement, infrastructure hazard). The in-car workflow is optimized for speed and minimal input, recognizing that the commissaires primary concern remains race supervision. Incident records generated through Veloviewer automatically populate the incident management database, creating a location-specific tag which is further extended after the race. Where available, Video Assistant Referee (VAR) teams provide a second input stream. Utilizing broadcast feeds from fixed cameras, motorcycle and helicopter footage, VAR officials attach video clips to the incident entry. The integration of VAR data reduces reliance on narrative-only descriptions and enables downstream (safety) analysis through computer vision techniques. Following the race, SafeR analysts further validate and extend the information included in the incident (involved riders, causes, etc.). Their role is to reconcile discrepancies between commissaire entries, VAR footage, and social media reports (from verified accounts), or in some cases, add missed incidents. Note that most incident inputs consist of standardized taxonomies to ensure cross comparability across events and seasons. Within the UCI World Tour, team medical representatives contribute further medical data (Fallon et. al, 2026). These include anatomical injury location, and a provisional severity indicator expressed as estimated days out of competition/training. By linking medical outcomes to the original incident record, the platform supports severity-weighted risk analysis rather than relying solely on incident counts. This distinction is important: not all crashes are equal in consequence, and safety interventions should be focused and prioritized based on potential severity. As part of ongoing platform development, we are piloting the integration of AI-enabled tracking devices capable of automated crash detection. These wearable or bike-mounted devices combine sensor and GPS values with machine learning models to identify crash-like events in real time. Beyond crash detection, aggregated tracking data provide contextual variables such as rider density, relative speed differences, speed entering technical sectors, and peloton configuration at the moment of impact. While experimental, these data streams offer an evidence base to better inform regulatory interventions, including targeted speed-reduction measures and other safety-focused rules introduced by the UCI. Beyond management and post-event analysis, a key innovation of the platform is its reuse of incident data for proactive safety screening in future race editions or other races using overlapping routes. Geospatial clustering algorithms identify high-incidence zones, such as specific descents, roundabouts, road narrowings, or finish straight configurations, and generate risk flags when these segments appear in proposed race itineraries. Historical incident profiles are presented to organizers and safety delegates during pre-race screening. This allows incident data to be used in a predictive planning tool and provides a data driven way of tagging dangerous points in combination with the machine learning based video analysis output on pre race parkour recordings (traversed with a camera on a car). The implementation of this end-to-end incident management system demonstrates that improved safety insights in elite cycling is achievable through structured data, captured from multi-source integration design. By connecting in-race officiating, video review, digital public signals, expert analysis, medical outcomes, and forward-looking route screening, the platform provides a holistic view of race risk. As elite road cycling continues to evolve in speed, technology, and global reach, such integrated systems will be essential for safeguarding athletes while preserving the competitive integrity and spectacle of the sport.

Optimisation of wire spoke Bicycle wheel - Aerodynamics

Brian-Francis Corrigan

Optimisation of the Wire-spoked Bicycle Wheel

Aerodynamics

Brian-Francis Corrigan^1,2

¹Head of Engineering, xtreme carbon Pty Ltd

²brian.xtremecarbon@gmail.com

Abstract:

This manuscript outlines the development of a ‘Prestress cable / strand’ structural model for the wire-spoked bicycle wheel that uses ‘Flexural’ structural analysis. During the independent research, it was verified by the results and data from this new Prestress cable / strand structural model that a simple, yet paradoxical discovery was made. A wire-spoked bicycle wheel design, that is based in ‘triangle or polygon theory’ as expressed structurally in the traditional ‘simplified triangular truss model,’ underestimates the actual Lateral force per spoke couple, and the Total Lateral force per side, for the wire-spoked bicycle wheel by a factor of ‘approximately 2’, when compared to the equivalent bicycle wheel that is structurally designed by using a ‘Prestress cable / strand structural model’. For Design engineers, this underestimation by a factor of ‘approximately 2’ in the Lateral direction plane, evidences that a more aerodynamic bicycle wheel can be designed by using the ‘Prestress cable / strand structural model’. Using a Prestress cable / strand structural model that replaces the traditional ‘simplified triangular truss’ structural model, an optimisation of the centreline-to-centreline flange separation was achieved. The reduction in the frontal area aerodynamic drag of the humble wire-spoked bicycle wheel is the direct empirical result; one which maintains the exact same lateral strength, stability, deformation under load, and safety, as the bicycle wheel designed using the traditional ‘simplified triangular truss model.’

Assessing performance fatiguability: reliability of a short severe-intensity cycling test

Chantelle du Plessis

Reliable and time-efficient testing protocols are required to assess acute performance interventions and their effects on athlete durability, as prolonged, fatiguing tests across multiple days are often unfeasible in high-performance environments. This study evaluated the within-session reliability of a short-duration (2 min), severe-intensity, constant-work rate (CWR) cycling test with pre-and-post-test maximal sprints to estimate performance fatigability in highly trained cyclists. The study involves a repeated measures reliability design conducted within a single testing session using linear mixed models and reliability statistics to assess differences in physiological and mechanical variables. Sixteen male and female cyclists completed three testing blocks, each separated by 30 min. Each block included a 2-min CWR test and 10-s maximal sprints performed before and immediately after. Two pre-CWR sprints (rolling and stationary starts) determined the start condition effects. Performance fatigability was estimated as the difference in power output between the pre-and post-CWR sprints. V̇O_2peak and heart rate showed excellent reliability (ICC>0.96, CV>1.2%), while mean and peak sprint powers (ICC>0.87, CV<5.5%) showed good reliability across repetitions. Sprints from a stationary start demonstrated higher reliability than rolling starts (CV=1.9 % vs. 3.0%) while remaining strongly correlated (r=0.77). Fatigue estimates (pre-CWR from a stationary start vs. post-CWR) were consistent across repetitions (~31%) and showed good reliability (ICC=0.75, TE=0.42 W∙kg^-1). This short-duration, severe-intensity cycling protocol reliably detects acute performance fatigue (a proxy for durability) within a single session. It offers a practical diagnostic and training tool for monitoring fatigue and evaluating different intervention strategies.