Pacing profiles and slowdown patterns in ultra-triathlon performance

Abstract

Pacing in long-distance triathlon has been studied primarily for cycling and running in IRONMAN triathlon and daily-format ultra-triathlons, as managing fatigue is critical for success. However, no study has analysed pacing across all three disciplines in an ultra-triathlon such as the second-longest non-stop triathlon format, the Double Deca Iron ultra-triathlon covering 76 km swimming, 3600 km cycling, and 844 km running. This study examined pacing during a Double Deca Iron ultra-triathlon by analysing split and lap times for all official male and female finishers. We assessed pacing patterns, the influence of pacing variability on performance, and whether faster performance was associated with more frequent moderate slowdowns or with fewer but more pronounced slowdowns. Official race data from the 2023 Swissultra Double Deca Iron ultra-triathlon held in Buchs, Switzerland were analysed for 9 men and 4 women. Swimming splits were recorded manually, and both cycling and running splits were recorded via RFID timing. Variables included mean speed, checkpoint speed variability (ACCS), proportions of slow-down checkpoints (25–50%, 50–75%, > 75% slower than mean speed), and magnitude of slowdown. Athletes showed negative pacing in swimming (decreasing time) but an even or slightly positive pacing in both cycling and running. Running variability did not correlate with running speed. In swimming, faster athletes tended to show fewer relative slowdowns. In cycling, faster cyclists had fewer moderate but more pronounced slowdowns, while in running, faster runners showed more ≥ 75% slowdowns. In summary, higher performers maintained a relatively steady and fast baseline pace interspersed with occasional substantial slowdowns, rather than moving continuously at a slower and more variable pace. Because the timing data do not directly identify intentional rest, these slowdown patterns should not be interpreted as confirmed rest breaks.

Introduction

Triathlon is a multi-sports discipline including the three disciplines swimming, cycling, and running. The race format can be held as sprint triathlon¹, Olympic distance triathlon², half-IRONMAN distance triathlon (IRONMAN 70.3)³, full distance IRONMAN triathlon⁴ and triathlon races covering x-times the IRONMAN distance from 3×, 5×, 10 times up to the Double Deca Iron ultra-triathlon distance of 76 km swimming, 3600 km cycling, and 844 km running⁵ as the second-longest triathlon distance and the Triple Deca Iron ultra-triathlon with distances of 114 km swimming, 5400 km cycling and 1260 km running⁶ as the longest triathlon distance held as an official race.

Pacing in an endurance sports discipline such as triathlons is of utmost importance for managing fatigue and achieving overall race success. Generally, pacing strategies can be classified as negative (the athlete started the run at a slower speed and gradually increased their pace), all-out (the athlete is running at maximum effort), positive (the athlete started the run faster and gradually slowed down), even (the athlete maintained a relatively constant speed throughout the race), parabolic-shaped (the speed follows a U-shaped pattern over the course of the race: the athlete starts fast, slows down in the middle, and then speeds up again toward the end), and variable pacing strategies⁷. In triathlon, pacing regulation is influenced by a variety of intrinsic and extrinsic factors, including race distance, drafting, environmental conditions (e.g., temperature and wind), course topography, transitions, age, experience, and sex⁸, biomechanics and physiological⁹, as well as psychological factors¹⁰ and recovery and wellbeing¹¹.

Pacing in triathlon can be considered from three perspectives: (i) the split disciplines where the most predictive split discipline can be defined¹², (ii) pacing during a specific split discipline such as cycling and running¹³, and (iii) pacing over days in multi-day triathlons covering 10 IRONMAN-distance triathlons in 10 days¹⁴, and up to 20 IRONMAN-distance triathlons in 20 days¹⁵.

Regarding the pacing in long-distance triathlon races, the longest distance that has been investigating for pacing during cycling and running was the IRONMAN distance^13,16 and distances of 2×, 3×, 5× and 10× the IRONMAN-distance in ultra-triathlon races¹⁴. In IRONMAN triathlon, positive¹⁶ as well as negative and even pacing¹³ can be found, most likely depending upon the level of the athletes. In longer race distances up to the Deca Iron ultra-triathlon, pacing is generally positive¹⁴.

For longer distances such as the Double Deca Iron ultra-triathlon⁵, no study ever investigated pacing in this race format. To date, the first Double Deca Iron ultra-triathlon was held in 1998 in Monterrey, Mexico. The second Double Deca Iron ultra-triathlon was held in 2019 in Leon, Mexico. Unfortunately, no lap times were recorded and stored in these two races. In 2023, the first Double Deca Iron ultra-triathlon was held in Europe and all split, transition, and lap times for all three disciplines were available.

In long triathlon races, such as IRONMAN and ultra-triathlons, pacing was analyzed only during cycling and running due to the fact that lap times were recorded only in these two disciplines by using an electronic chip system. In the longest triathlon held in history, the Triple Deca Iron ultra-triathlon covering 114 km of swimming, 5400 km of cycling and 1266 km of running with four men and three women officially finishing the 2024 Triple Deca Ultra Triathlon in Desenzano del Garda, Italy, pacing was analyzed for cycling and running⁶. Cycling showed the greatest pacing variability, while running exhibited a steadier pacing, with more consistent athletes performing better overall⁶.

Regarding the above mentioned, the first aim of the present study was to analyze pacing in the second-longest triathlon race considering split and lap times for all three split disciplines including swimming of all official male and female finishers. We aimed to analyze pacing and to see if the pacing variability affects results (one variable to asses pacing variability like in ultra-marathon running^17,18. The second aim of the study was to investigate whether faster performance was associated with more frequent moderate slowdowns or with fewer but more pronounced slowdowns. Based on previous findings, we hypothesized that successful finishers would adopt a positive pacing strategy (i.e., progressively longer lap times) across all split disciplines.

Methods

Ehical approval

The study examined pacing and performance in the 2023 Swissultra Double Deca Ultra Triathlon, an ultra-endurance triathlon held in Buchs, Switzerland. All athletes who started and finished the 2023 Swissultra Double Deca Ultra Triathlon were included. The sample comprised four female and nine male competitors. The Institutional Review Board of Kanton St. Gallen, Switzerland, has approved this study (EKSG 01/06/2010), with a waiver of the requirement for informed consent of the participants as the study involved the analysis of publicly available data. The study was conducted in accordance with ethical standards recognized in the Declaration of Helsinki, adopted in 1964 and revised in 2013.

The race

The Double Deca Iron ultra-triathlon distance corresponds to 20 consecutive long-distance triathlons, yielding a total race distance of 76 km swimming, 3600 km cycling, and 844 km running. Swimming was performed in a 50 m outdoor pool (Freibad Rheinau) at an average water temperature of ~ 20–23 °C, with solar-heated surface temperatures occasionally reaching ~ 25 °C. Athletes swam a total of 76 km, completed in 15 × 5 km laps, with the last lap extended to 6 km to reach the total distance. Cycling consisted of 400 laps of 9 km each (total 3600 km) on a flat, traffic-free loop along the river Rhein. Running was performed on a 1.22 km loop around Freibad Rheinau, repeated 692 times for a total of 844 km. In 2023, 13 athletes (four women, nine men) started the Double Deca Ultra Triathlon, and all finished successfully within the official time limit. Notably, the women’s winner set a new world record, improving the previous record (2019) by ⁓78 h.

Data set and data preparation

Official race results (split and lap times) were downloaded from the Swissultra website (www.swissultra.ch/results). Because the final swimming split covered 6 km whereas the preceding splits covered 5 km, all swimming pacing analyses were performed using speed (m/s) rather than raw split time. Cycling splits (9 km laps) and running splits (1.22 km laps) were recorded electronically using an RFID chip system (RaceResult, www.raceresult.com) worn by each athlete. The raw timing data were imported into Microsoft Office Excel 2024 for cleaning and calculation of lap and discipline mean speeds (m/s).

Transition times were not analyzed separately in the present study since they likely included a heterogeneous mix of clothing changes, nutrition, hygiene, medical care, and sleep.

Variables

The following variables were used to quantify performance and pacing:

• Mean speed: Average lap speed (m/s), calculated as lap distance divided by lap time. Overall mean speed for a discipline was the distance-weighted average of lap speeds.

• Checkpoint speed variability (ACCS): Defined as the “average change in checkpoint speed.” It represents the mean percentage change in speed between consecutive laps in the run, with higher values indicating greater pacing variability:

  • ACCS (%) = (100/(n − 1)) × Σ |(v_i − v_{i − 1})/v_{i − 1}|, where v_i is the speed at checkpoint i and n is the number of checkpoints in the discipline.

• Slow-down checkpoint: Slow-down checkpoint: For the purposes of the present categorical analyses, a lap classified as 25–50%, 50–75%, or > 75% slower than the athlete’s mean speed for that discipline.

• Proportion of slow-down checkpoints: The proportion of laps in each discipline classified into the three reported slowdown categories (25–50%, 50–75%, or > 75% slower than the athlete’s mean speed for that discipline. It was calculated as:

Proportion of slow-down checkpoints (%) = Number of slow-down laps/Total laps in discipline × 100.

To further characterize the magnitude of temporary speed reduction, each lap was compared with the athlete’s own mean discipline speed and classified into one of three categories: 25–50% slower, 50–75% slower, or > 75% slower than mean speed. These thresholds were defined a priori as descriptive, operational cut-offs to distinguish moderate, marked, and near-stop reductions in speed in the absence of established classifications for Double Deca ultra-triathlon pacing. Therefore, they should be interpreted as data-driven indicators of slowdown magnitude, not as externally validated physiological thresholds.

Statistical analysis

Before inferential analyses, descriptive statistics were calculated as mean and standard deviation. Because of the small sample size, distributional assumptions were assessed cautiously using visual inspection of histograms and quantile–quantile plots. To assess pacing profiles, linear regressions were applied on mean speed for each of the 15 laps in swimming, 400 laps in cycling and 692 laps in running. Furthermore, the Spearman correlation coefficient was performed to assess the correlation between the proportion of slow-down checkpoints and mean speed and pacing variability. All correlation coefficients were interpreted as small, r = 0.10–0.29; moderate, r = 0.30–0.49; and large, r = 0.50–1.0¹⁹. Given the limited sample size, all analyses were considered exploratory and descriptive. Alpha level was set at p ≤ 0.05. All statistical tests were performed using Microsoft Office Excel 2024 (Microsoft Corporation, Redmond, WA, USA) and SPSS 26 (IBM, Armonk, NY, USA).

Results

A total of 9 men and 4 women started the race, all of them officially finished within the time limit. The first women set a new world record by improving the existing world record from 2019 by ~ 78 h. Regarding the descriptive analysis, mean speed and checkpoint speed variability (ACCS) was presented for each discipline and each participant, as well as for the entire sample (Table 1).

Table 1.

Mean speed and checkpoint speed variability (ACCS) for each discipline and each participant.

No	Gender	Place	S_avg (m/s)	C_avg (m/s)	R_avg (m/s)	T_avg (m/s)	S_ACCS (%)	C_ACCS (%)	R_ACCS (%)
1	Male	1	0.40	6.19	1.63	2.38	26.37	20.37	20.40
2	Male	2	0.45	5.67	1.85	2.38	12.34	29.05	24.41
3	Male	3	0.58	5.12	1.87	2.28	15.40	24.72	20.22
4	Male	5	0.53	6.09	1.38	2.17	22.75	22.54	21.23
5	Male	6	0.59	5.76	1.69	2.13	21.42	24.38	35.46
6	Male	8	0.43	5.27	1.51	2.00	21.04	30.44	27.46
7	Male	9	0.54	6.23	1.45	1.97	28.25	23.06	28.60
8	Male	12	0.48	5.97	1.46	1.86	26.67	26.38	25.95
9	Male	13	0.31	5.01	1.19	1.74	23.82	23.37	29.29
10	Female	4	0.56	6.10	1.42	2.26	15.91	16.14	24.19
11	Female	7	0.44	5.35	1.34	2.01	19.59	24.84	27.31
12	Female	10	0.42	5.18	1.24	1.93	29.60	21.47	22.19
13	Female	11	0.42	4.68	1.42	1.87	21.59	20.98	23.16

S_avg, average swimming speed; C_avg, average cycling speed; R_avg, average running speed; T_avg, average triathlon speed; S_ACCS, average change in checkpoint speed for swimming; C_ACCS, average change in checkpoint speed for cycling; R_ACCS, average change in checkpoint speed for running.

Table 2 shows the rate of slowed-down checkpoints as a percentage of total checkpoint for each participant.

Table 2.

The rate of slowed-down checkpoints as a percentage of total checkpoint for each participant.

No	Gender	Place	S_25–50%	S_50–75%	S_75+%	C_25–50%	C_50–75%	C_75+%	R_25–50%	R_50–75%	R_75+%
1	Male	1	6.7	13.3	0.0	4.0	5.0	5.3	5.2	3.9	4.0
2	Male	2	0.0	0.0	0.0	14.5	5.0	5.0	7.5	3.6	5.9
3	Male	3	0.0	0.0	0.0	9.3	6.5	4.0	2.2	0.7	9.2
4	Male	5	0.0	13.3	0.0	5.0	6.0	5.8	4.0	3.5	5.2
5	Male	6	0.0	6.7	0.0	5.3	6.3	5.3	9.1	8.8	6.6
6	Male	8	6.7	6.7	0.0	8.8	10.0	5.0	5.3	7.9	5.8
7	Male	9	6.7	13.3	0.0	2.5	7.5	5.8	5.9	5.9	5.8
8	Male	12	0.0	13.3	0.0	3.8	8.8	6.5	3.6	5.3	7.1
9	Male	13	6.7	13.3	0.0	10.5	2.3	4.5	9.7	10.4	3.6
10	Female	4	0.0	6.7	0.0	4.0	3.3	3.8	2.7	3.2	7.9
11	Female	7	0.0	13.3	0.0	10.0	3.5	6.0	6.2	7.1	5.8
12	Female	10	20.0	6.7	0.0	11.3	2.3	3.0	8.4	4.9	3.8
13	Female	11	0.0	13.3	0.0	6.3	5.5	4.0	6.2	5.6	4.5

S_, C_, R_, = Proportion of swimming, cycling, and running checkpoints classified as 25–50%, 50–75%, and > 75% slower than the athlete’s mean discipline speed, in relation to the total number of checkpoints.

Across all athletes, the overall proportion of checkpoints classified as at least 25% slower than the athlete’s discipline mean was 12.8% in swimming, 17.7% in cycling, and 17.1% in running.

Figures 1, 2 and 3 depict linear regressions of average speed for each checkpoint for swimming (Fig. 1), cycling (Fig. 2), and running (Fig. 3), indicating negative pacing in swimming and approximately even pacing in cycling and running.

Fig. 1.

Linear regression of averaged individual checkpoint for swimming.

Fig. 2.

Linear regression of averaged individual checkpoint for cycling.

Fig. 3.

Linear regression of averaged individual checkpoint for running.

Because the primary aim was to characterize pacing at the cohort level, Figs. 1, 2 and 3 present group-averaged checkpoint-speed profiles rather than individual pacing trajectories.

Table 3 presents the exploratory associations between checkpoint speed variability (ACCS), slowdown categories, and mean discipline speed. The strongest individual association was the negative correlation between cycling mean speed and the proportion of 25–50% slowdowns (r = − 0.729, p = 0.005). The other nominally significant associations (swimming 25–50%, p = 0.030; cycling > 75%, p = 0.048; running > 75%, p = 0.016) were near the significance threshold and should be interpreted with caution, given the number of correlations examined and the small sample size.

Discussion

The aim of the present study was to analyze pacing in the second-longest non-stop triathlon in the world by examining split and lap times of all three disciplines and all official male and female finishers. Our hypothesis that finishers would adopt a positive pacing strategy across all disciplines was only partially supported. Swimming showed negative pacing, whereas cycling and running were closer to even pacing. This difference may reflect discipline-specific demands, early-race settling effects, and the structured environment of pool swimming.

Our central observation was that better cycling and running performance co-occurred with fewer moderate slowdowns and, in some cases, more pronounced slowdowns, rather than with continuously slower movement. Specifically, faster athletes showed fewer moderate slow-downs (25–50% below their mean speed) but more pronounced slow-downs (≥ 75% below mean speed) in cycling (r = − 0.729 and r = 0.506, respectively) and running (r = 0.652 for ≥ 75%) (Table 3).

Table 3.

Correlations between checkpoint speed variability (ACCS), slowdown categories, and mean discipline speed.

Mean speed	Test	ACCS	25–50%	50–75%	75+%
Swimming	Correlation Coefficient	-0.385	− 0.600*	− 0.392	–
	Sig. (2-tailed)	0.194	0.030	0.185	–
Cycling	Correlation Coefficient	-0.231	− 0.729**	0.212	0.506*
	Sig. (2-tailed)	0.448	0.005	0.487	0.048
Running	Correlation Coefficient	-0.143	− 0.352	− 0.313	0.652*
	Sig. (2-tailed)	0.642	0.238	0.297	0.016

ACCS, average change in checkpoint speed.

These findings suggest that sustaining a relatively steady, fast baseline pace while concentrating speed loss into fewer, more pronounced slowdowns may differ from accumulating numerous moderate slowdowns during continuous movement. However, because the timing data do not allow direct identification of intentional rest, this pattern should not be interpreted as confirmed rest behavior. Moreover, given the very small number of finishers, the present analyses are descriptive and exploratory and should be considered hypothesis-generating rather than confirmatory. Larger datasets from future editions of this race format will be required to verify whether the same pattern is reproducible.

These Double Deca Iron ultra-triathlon patterns align with even-pacing advantages reported in ultrarunning/trail contexts, with the nuance that occasional pronounced slowdowns may coexist with faster overall speed in multi-day triathlon²⁰. In the UTMB (Ultra-trail du Mont Blanc), a 170 km long circular trail with over 10,000 m of vertical gain, and crossing three countries (France, Italy, and Switzerland), seven valleys, 71 glaciers and ~ 400 summits along the Tour du Mont Blanc hiking path even pacing throughout the UTMB correlated with faster finishing times²¹. Berger et al.²⁰ concluded that an even pacing—in addition to heat acclimation and individualized hydration—would enhance performance. On the contrary, an uneven pacing might lead to transitions between exercise intensity levels increasing fatigue and, consequently, deteriorating performance²².

In ultra-cycling, athletes rather adopt, however, a positive pacing^23,24 where environmental factors (i.e. temperature and wind speed) seem to have an influence on cycling speed²³. This observation was in agreement with the notion of Thiel et al. suggesting that pacing depended on the locomotion mode²⁵.

Although sleep-related variables were not measured in the present study, rest and sleep remain relevant contextual factors in ultra-endurance events of this duration²⁶ since it is not possible to compete for days without rest^27,28 and sleep deprivation has a considerable influence on race outcome²⁹. In runners competing in the UTMB, the lack of sleep had marked adverse effects on cognitive and physiological performances³⁰. Sleep deprivation may perturb the balance of key neurotransmitter systems (e.g., ↑ adenosine, ↓ dopamine, altered GABA/glutamate dynamics), compromise cortical excitability and plasticity, and raise the perception-of‐effort threshold—all of which are mechanistically linked to central fatigue and thus can impair performance in ultra-endurance events^31–34. In the Race Across AMerica (RAAM), a 4856 km continuous cycle race across the United States of America, sleep efficiency was better maintained during longer rest periods³⁵. Generally, ultra-endurance athletes often prefer short naps (micro naps) lasting less than 30 min^36,37.

Limitations

Although this is the first study to investigate pacing during swimming, cycling, and running in the second-longest non-stop triathlon, this study is not free of limitations. Indeed, we were not able to include important aspects such as training^38–40, previous experience^38,39, nutrition^41,42, energy balance⁴³, environmental conditions⁴⁴, or anthropometry⁴⁵ which all have an influence on triathlon race performance. Nevertheless, the findings may help ultra-triathletes and their coaches refine pacing strategies and effort distribution during multi-day ultra-triathlon races. In addition, although the race included both women and men, the number of female finishers was too small to support robust sex-specific inferential analyses. Finally, slowdown categories were defined relative to each athlete’s own mean speed. This approach may introduce mathematical coupling when such variables are correlated with the mean speed itself. Therefore, these results should be interpreted as descriptive and exploratory rather than as independent evidence of a performance mechanism.

Conclusion

Our findings suggest that faster finishers tend to show a steadier baseline pace with intermittent, pronounced slowdowns, rather than continuous movement at a slower, more variable speed. These patterns should not be interpreted as direct evidence of intentional rest behavior, because the available timing data cannot distinguish deliberate rest from other causes of marked speed reduction. Studies such as this one provide a valuable foundation for potential research and interventions in ultra-endurance events. Future studies should test whether these pacing patterns are reproducible in larger samples and should combine pacing data with direct measures of rest, sleep, and transition behavior.

Author contributions

S.M., I.C., S.D. and B.K. drafted the manuscript, S.M., I.C., and S.D. performed the statistical analysis and prepared methods and results, B.K. obtained the data, P.F., L.B.L., M.S.A., P.T.N., K.W., and T.R. helped in drafting the final version. All authors read and approved the final manuscript.

Data availability

For this study, we have included official results and split times from the official Swissultra website (https://www.swissultra.ch/results). The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.