HYROX Running Study: Lessons From 39,696 PRO & ELITE Results Reveal

Running Wins the Clock—But It Does Not Win HYROX Alone: What 39,696 Race Results Reveal

Executive takeaway

HYROX is a hybrid race, but the clock is endurance-dominated. In the first large longitudinal analysis of PRO and ELITE competition, running occupied about half of total race time, running rank most closely tracked overall finishing rank, and faster running accounted for the largest share of performance improvement across seven seasons. That makes running the broadest opportunity for time gain—but not a license to neglect strength, stations, or transitions.

The most useful coaching interpretation is a two-part model: build the aerobic speed, economy, threshold, and durability required to protect all eight runs; then develop enough strength, technique, and local muscular endurance that stations do not create disproportionate losses or destroy the next run. For slower athletes especially, weak station performance amplified total time loss.

Prioritize running capacity as the largest performance lever. Protect it with a sufficient strength floor, efficient station execution, and repeated exposure to running after muscular work. Test the individual athlete before deciding where the next training block belongs.

The paper at a glance

Why this study matters

Early HYROX guidance often borrowed from CrossFit, obstacle racing, and endurance sports because direct evidence was scarce. The standardized HYROX format creates a valuable natural dataset: the same broad sequence of eight 1-km runs and eight stations can be compared across thousands of performances and multiple seasons. Rappelt and colleagues used that scale to move the conversation beyond small laboratory samples and anecdote.

The paper also captures the professionalization of the sport. Median PRO performances improved most clearly from Season 4 onward, Top-100 performances became dramatically faster, and ELITE fields became more tightly clustered. The practical consequence is uncomfortable but useful: general fitness is becoming less sufficient. As competitive density rises, precise limiter identification and long-term specialization matter more.

How the study was designed

Data source and extraction

Researchers downloaded publicly available individual PRO and ELITE results from the official HYROX results website. Custom Python and R scripts first identified events, retrieved event result pages, extracted links to individual results, compiled split data, and merged the records by sex. Collection occurred over five weeks from September 29 to October 30, 2025, with throttled requests and no bypass of access controls.

From 51,469 records to 39,696 analyzable results

The starting dataset contained 15,754 female and 35,715 male records. Researchers removed duplicate listings and excluded performances with a penalty, time bonus, or at least one missing discipline split. The final dataset contained 11,842 female results (11,605 PRO; 237 ELITE) and 27,854 male results (27,618 PRO; 236 ELITE). A result is not necessarily a unique athlete: some athletes raced more than once.

IMPORTANT INTERPRETATION This is a study of recorded performances, not a prospective study that measured each athlete’s training, physiology, injury status, nutrition, or equipment. It can show strong patterns and associations; it cannot prove that one training intervention caused the improvement.

What counted as running

Running time was the cumulative time assigned to the fixed 8-km running distance and excluded Rox Zone time, which was analyzed separately. Because venue layouts historically created differences in individual lap and Rox Zone configuration, the authors did not analyze pacing for each 1-km segment. Segment distances became standardized in the 2024/2025 season, which should support stronger pacing research in future seasons.

The statistical strategy

Percentile curves. The authors mapped entire PRO finishing-time distributions by season, then compared the 25th, 50th, 75th, and 90th percentiles using 5,000-resample bootstrapping. Holm–Bonferroni adjustment reduced false-positive risk from multiple comparisons.

Rank reshuffling. For each discipline, they compared discipline rank with overall rank and normalized the difference to field size. A low value means the discipline ranking closely resembles the finishing ranking.

Quantile regression. They modeled the association between each discipline and the rest of the race at several performance levels. Importantly, the discipline’s own time was removed from total time to avoid a built-in part–whole correlation.

Top-100 mixed models. The top 100 performances per season were compared for total, running, station, and Rox Zone time. Models included a participant random intercept, and sensitivity analyses addressed repeat performances and event clustering.

ELITE convergence. Means, medians, coefficients of variation, bootstrapped confidence intervals, Top-5 results, and retention were used to describe whether elite fields became faster and denser.

What the study found

The sport became much faster

Between Seasons 4 and 7, male and female PRO median performances improved by roughly 8–10 minutes. Across Seasons 1 to 7, the Top-100 mean improved from 70:13 to 56:56 in men (13:17; about 19%) and from 80:43 to 63:11 in women (17:32; about 21%). Season 3 was an outlier, coinciding with the COVID-era reduction in events and a much smaller available field.

Running produced the largest absolute gains

Top-100 cumulative run time fell from 35:44 to 27:38 in men—an 8:06 gain—and from 40:20 to 30:17 in women—a 10:03 gain. Those changes represented about 61% of the total male improvement and 57% of the total female improvement. Running still occupied roughly half the race: 48.5% of men’s and 47.9% of women’s Top-100 time in Season 7.

Figure 1. Original summary calculated from Rappelt et al. (2026), Top-100 Season 1 and Season 7 mean times.

Running rank tracked finish rank most closely

Running consistently produced the lowest rank reshuffling. By Season 7, the normalized rank-difference standard deviation was 12.7% for men and 12.5% for women—lower than every station. Put simply, an athlete’s place in the running ranking tended to resemble the athlete’s final place more closely than any single station ranking did.

Figure 2. Season 7 rank reshuffling by discipline. Lower values indicate closer alignment with final ranking. Data: Rappelt et al. (2026).

But “running was the strongest relationship” needs precision

Running was the largest contributor to total time and the discipline most closely aligned with finishing rank. However, in quantile-regression models that removed each discipline’s own seconds from total time, running had only intermediate independent explanatory value. For men, rowing showed the highest pseudo-R² range (0.25–0.46), while running ranged from 0.14–0.35. For women, rowing (0.22–0.55), lunges (0.17–0.46), and burpee broad jumps (0.17–0.40) were among the strongest, while running ranged from 0.14–0.37.

WHY THE TWO FINDINGS CAN COEXIST Total running is long, so it has the largest direct footprint on the clock. The adjusted models ask a different question: does being good at one discipline also identify athletes who are good during the rest of the race? Rowing, lunges, and burpees may capture broader aerobic durability, muscular endurance, technique, and fatigue resistance. Neither analysis means rowing should receive more total training time than running.

Weak stations punished slower athletes more

Discipline times were positively associated with remaining race time across nearly all modeled performance quantiles. Coefficients generally became larger toward the slower quantiles, particularly for strength-oriented stations. The practical message is not that stronger athletes always win; it is that an inadequate strength-and-skill floor can create nonlinear damage—long station time, excessive local fatigue, and slower running afterward.

ELITE competition became faster and denser

Male ELITE median time improved from 1:06:24 in Season 2 to 57:17 in Season 7, while variation fell from 10.4% to 4.7%. Female ELITE median time improved from 1:11:09 to 1:03:22, with variability falling to 4.8% by Season 7. Top-five performance improved markedly as well. The field is not only faster—it is less forgiving.

What the study does—and does not—prove

MAJOR STRENGTHS Real-world official race data; exceptional scale; both sexes and competitive levels; complementary statistical approaches; protection against part–whole correlation; sensitivity analyses; transparent public data and disclosures.

Strengths worth keeping in view

• Exceptional ecological validity: official real-world race performances rather than isolated fitness tests.
• Large dataset spanning both sexes, PRO and ELITE categories, and seven seasons.
• Multiple complementary analyses rather than reliance on one simple correlation.
• Thoughtful protection against part–whole correlation in quantile models.
• Sensitivity analyses addressing repeat performances, event clustering, and one performance per athlete per season.
• Transparent public data source, supplementary scripts, funding statement, and conflict-of-interest disclosures.

Limitations that constrain the claim

• Observational association cannot establish that more running training caused faster race times.
• Results are performances rather than unique people; repeated races can enter the dataset.
• PRO entry does not require a universal performance standard, creating a heterogeneous population.
• Excluding penalties, bonuses, and missing splits improves comparability but may remove meaningful execution failures and introduce selection bias.
• The dataset lacks training history, VO₂max, thresholds, VLamax, running economy, body composition, injury status, footwear, temperature, travel, and fueling.
• Changing participation and competitive density can make later seasons look faster even without within-athlete improvement.
• Venue and Rox Zone differences complicate historical lap-level and transition comparisons.
• Top-100 comparisons are informative but are selected extremes, not a fixed cohort followed prospectively.

What this means physiologically

The authors reasonably propose that faster running may reflect improvements in maximal aerobic power, lactate-threshold-related speed, running economy, and physiological resilience. Race-result data cannot identify which mechanism changed. Still, the interpretation aligns with the first HYROX physiological study: in 11 recreational athletes completing a simulated OPEN race, higher VO₂max and greater endurance-training volume were associated with faster performance, while higher body-fat percentage was associated with slower performance (Brandt et al., 2025).

For a coach, the word resilience is especially useful. A fresh 5-km or threshold result is not the entire HYROX running problem. Athletes must repeatedly restore effective mechanics and sustainable metabolic control after sleds, burpees, rowing, carries, lunges, and wall balls. Trunk and upper-body fatigue can also worsen running economy, reinforcing why station preparation and core capacity support—not compete with—running performance (Drum et al., 2019; Jones, 2024).

HYROX running performance = fresh running capacity × station-cost control × the ability to reproduce pace eight times. Improving only the first factor leaves race-specific leakage untreated.

The actionable plan: Test → Analyze → Train → Retest

Step 1 — Test the athlete, not the headline

• Fresh running: 3-km or 5-km time trial, plus a threshold-oriented field or laboratory assessment where appropriate.
• Aerobic profile: VO₂max/velocity at VO₂max, LT1, LT2 or maximal lactate steady-state proxy, and running economy when available.
• HYROX durability: repeated 1-km running with standardized station inserts; compare early versus late run pace, heart rate, RPE, and mechanics.
• Strength floor: race-load sled push/pull, farmers carry, lunges, and wall-ball capacity with valid technique—not gym maxes alone.
• Station efficiency: split time, cadence, breaks, no-rep risk, transition time, and post-station running cost.
• Context: training age, body mass/composition, injury history, available weekly volume, heat, sleep, and nutrition tolerance.

Step 2 — Classify the primary limiter

Step 3 — Build the week around the limiter

A sensible default is to make running the largest endurance exposure while preserving two strength touches and one controlled race-specific session. Exact volume must reflect training age, injury history, division, race date, and recovery capacity.

Do not turn every run into a compromised metcon. Fresh running develops qualities that fatigue can obscure; compromised running teaches transfer. Most athletes need both, with intensity distributed deliberately rather than accumulated randomly.

Step 4 — Progress from capacity to specificity

Step 5 — Retest what should have changed

Retest after 4–8 weeks depending on the target adaptation. Use the same protocol, conditions, equipment, surface, warm-up, and scoring rules whenever possible. The decision is not simply whether a number improved; it is whether the intended limiter improved without unacceptable trade-offs elsewhere.

• Fresh run time or threshold speed improves while station competency is maintained.
• Compromised-run decay shrinks at the same station load and similar RPE.
• Station breaks, no-rep risk, and transition time decline without overpacing the next run.
• Total race-model time improves with a smaller final-quarter slowdown.
• Pain, excessive fatigue, or loss of strength does not increase in the pursuit of more run volume.

Race-day implications

• Protect the run average, not the fastest first kilometer. The study analyzed cumulative running, and HYROX rewards repeatability.
• Treat stations as cost-management problems. The fastest possible station split is not always the fastest total-race decision.
• Earn the right to push late. A controlled opening preserves technique and reduces the chance that lunges and wall balls become catastrophic time sinks.
• Practice transitions. Rox Zone time was substantial—about 7% of Top-100 Season 7 time—and is trainable without improving VO₂max.
• Fuel and hydrate to preserve the limiter you trained. Strategies should be practiced and individualized; avoid introducing a new plan on race day.

Five mistakes this paper should prevent

• Programming HYROX as eight random metcons and assuming that race specificity automatically develops aerobic speed.
• Reacting to “running is king” by removing heavy strength work and losing the force reserve needed for PRO loads.
• Chasing faster SkiErg or row splits when the effort increases the cost of the next run more than it saves on the machine.
• Using one elite template for every athlete instead of testing whether the current leak is fresh speed, durability, strength, skill, pacing, or fueling.
• Judging progress only by a full simulation, which is fatiguing and can hide the mechanism that changed.

A coach’s decision rule

IF RUNNING IS WEAK FRESH Build running capacity first while maintaining the strength floor.

IF RUNNING COLLAPSES ONLY AFTER STATIONS Target durability, station cost, and compromised mechanics—not simply more high-intensity running.

IF STATIONS ARE SURVIVAL EVENTS Raise strength and local muscular endurance until loads are manageable, then redirect more capacity toward running.

IF EVERYTHING IS ADEQUATE BUT TOTAL TIME STALLS Audit pacing, Rox Zone execution, no-reps, body composition, fueling, heat preparation, and recovery before adding volume.

Conclusion

Rappelt and colleagues provide the strongest large-scale description of competitive HYROX performance to date. Across 39,696 PRO and ELITE results, running was the largest component of the race, the biggest source of longitudinal time improvement, and the discipline whose ranking most closely resembled overall finishing position. That evidence should change how HYROX programs allocate attention.

But the highest-quality interpretation is not “just run more.” Running explained the clock broadly; strength-endurance and technical stations still reshuffled positions and disproportionately punished slower athletes. The winning model is therefore endurance-led, strength-protected, technically efficient, and individually tested. Coaches should use the study to set priorities—then use athlete-specific data to decide the prescription.

References

Brandt, T., Ebel, C., Lebahn, C., & Schmidt, A. (2025). Acute physiological responses and performance determinants in HYROX©—A new running-focused high intensity functional fitness trend. Frontiers in Physiology, 16, Article 1519240. https://doi.org/10.3389/fphys.2025.1519240

Buchheit, M., & Laursen, P. B. (2013a). High-intensity interval training, solutions to the programming puzzle: Part I: Cardiopulmonary emphasis. Sports Medicine, 43, 313–338. https://doi.org/10.1007/s40279-013-0029-x

Buchheit, M., & Laursen, P. B. (2013b). High-intensity interval training, solutions to the programming puzzle: Part II: Anaerobic energy, neuromuscular load and practical applications. Sports Medicine, 43, 927–954. https://doi.org/10.1007/s40279-013-0066-5

Davids, C. J. (2026). A performance analysis of HYROX: A review of the physiologic, mechanical, and technical demands. Strength & Conditioning Journal, 48(1), 88–100. https://doi.org/10.1519/SSC.0000000000000913

Drum, S. N., Rappelt, L., & Donath, L. (2019). Trunk and upper body fatigue adversely affect running economy: A three-armed randomized controlled crossover pilot trial. Sports, 7(8), Article 195. https://doi.org/10.3390/sports7080195

Fernández-Navarrete, I., Ruiz-Alias, S. A., Gómez-García, M., Riquelme-Sebastián, L., & García-Pinillos, F. (2026). HYROX©: A descriptive analysis of participation trends among over 275,000 athletes in the new fitness modality. Research Quarterly for Exercise and Sport, 97(2), 266–272. https://doi.org/10.1080/02701367.2025.2584226

Jones, A. M. (2024). The fourth dimension: Physiological resilience as an independent determinant of endurance exercise performance. The Journal of Physiology, 602, 4113–4128. https://doi.org/10.1113/JP284205

Joyner, M. J., & Coyle, E. F. (2008). Endurance exercise performance: The physiology of champions. The Journal of Physiology, 586, 35–44. https://doi.org/10.1113/jphysiol.2007.143834

Laursen, P. B., & Jenkins, D. G. (2002). The scientific basis for high-intensity interval training: Optimising training programmes and maximising performance in highly trained endurance athletes. Sports Medicine, 32, 53–73. https://doi.org/10.2165/00007256-200232010-00003

Lundby, C., Montero, D., & Joyner, M. (2017). Biology of VO₂max: Looking under the physiology lamp. Acta Physiologica, 220, 218–228. https://doi.org/10.1111/apha.12827

Muniz-Pumares, D., Hunter, B., Meyler, S., Maunder, E., & Smyth, B. (2025). The training intensity distribution of marathon runners across performance levels. Sports Medicine, 55, 1023–1035. https://doi.org/10.1007/s40279-024-02137-7

Rappelt, L., Wiedenmann, T., Held, S., Heinke, L., Micke, F., Wicker, P., & Donath, L. (2026). Longitudinal performance development in PRO and ELITE HYROX competitions across the first seven competitive seasons. Frontiers in Physiology, 17, Article 1847569. https://doi.org/10.3389/fphys.2026.1847569

Editorial note

This article is an independent educational review of published research. It does not imply endorsement by HYROX or the study authors and is not medical advice. Training recommendations should be individualized to athlete history, capacity, division, injury status, and available supervision.

Original study access: https://doi.org/10.3389/fphys.2026.1847569