Skip to content
/ tri_stats Public

Analyse data of the World Triathlon (formerly ITU) to try to answer key questions about elite triathlon. ๐ŸŠ๐Ÿšด๐Ÿƒ

0 stars 1 fork Branches Tags Activity
Star
Notifications You must be signed in to change notification settings

chauvinSimon/tri_stats

BranchesTags

Repository files navigation

This document analyses data of World Triathlon, formerly International Triathlon Union (ITU), to try to answer questions such as:

  • ๐Ÿง Should bad swimmers be happy when the wetsuit is allowed?
  • ๐Ÿฉฑ Can the benefit of swim wetsuits be estimated?
  • ๐Ÿ‘Ÿ Are runs faster since carbon plates shoes?
  • 3๏ธโƒฃ Does each sport (swim - bike - run) accounts for 1/3 of race durations? What about transitions?
  • ๐Ÿ‘ซ How much faster are men over women? In which sport is the gap larger/smaller? How has the gap evolved over the years?
  • โณ How much faster are athletes in sprint over olympic distance?
  • ๐Ÿ“ˆ Does the level increase over years?
  • ๐ŸŽฏ How often does an athlete win from a bike breakaway?
  • ๐Ÿƒ How often does the best runner win?
  • ๐Ÿš€ How often is the win decided with a sprint finish?
  • ๐ŸŒก๏ธ Do water and air temperatures affect swimming and running performance?
  • ๐Ÿณ๏ธ What are the most represented nations? Which nations have serious problems for their Olympics selection?
  • ๐Ÿ’ช What is the typical age of performing athletes and how has it evolved over years?
  • โญ๏ธ How old are athletes when they stop racing?
  • ๐Ÿ‹๏ธ What is the body mass index of performing triathletes?
  • ๐ŸŽ‚ Are two kids, born the same year but on two different months, equally likely to become professional triathletes?

๐Ÿ“š DATA

Data are collected from the Triathlon.org API.

"The Triathlon.org API Platform allows access to the entire Triathlon.org infrastructure and data"

Race results are processed as followed:

  • Year: from 2009 (start of the world-series) to mid-2024, just after the Paris Olympics.
  • Event: only world-cups, world-series (called WTCS) and games related events (Commonwealth, Olympics and Olympic test events).
  • Distance: only sprint (750 - 20 - 5) and olympic (also called "standard": 1500 - 40 - 10) formats.
  • Minimum number of finishers: 25.
  • โš ๏ธ IMPORTANT: How to summarize all split-times of one race?
    • For each leg (swim, t1, bike, t2, run), an average of 5 times is computed.
    • Specifically, the 5-th to 9-th best times in each sport are used to compute this average.
    • This choice is arbitrary but, as explained below, relevant for my goal: "Analysing the general competitive field in each sport".
    • Other settings, e.g. top-1, top-3, top-10 and top-50, could also yield valuable insights, by adjusting a few parameters in the provided scripts.
    • Notably, the PACES and LEVEL OVER THE YEARS sections include top-3 and top-10 analyses as well. ๐Ÿ…

Why consider split-times based on the ranking in each sport?

  • The split times of the finish-top-10 athletes can vary greatly depending on the race scenario:
    • For instance, a race might feature a strong breakaway group of swimmers reaching T2 with a significant lead, filling the top-10 positions.
    • In the same race, a large pack might arrive at T2 together, with the strongest runners then dominating the overall finish.
    • See this dedicated section for an analysis of race scenarios.
  • On the other hand, considering the ranking in each sport allows for more consistent and reliable comparisons across races.

Why consider the 5-th to 9-th best times?

  • To mitigate outliers:
    • Outstanding swimmers, riders, or runners might miss a race due to injury, scheduling conflicts, or other reasons, or they might have unusually good or bad performances.
    • Conversely, the 5-th to 9-th times are usually denser, providing a more robust representation of the general competitive field.
  • Against strategic variability among top performers:
    • The top-4 athletes might engage in strategic tactics on the run, such as testing each other or waiting for a final sprint, leading to varying performances.
    • Conversely, I believe athletes ranked 5-th to 9-th are more likely to give their all without strategic calculations, resulting in more robust and consistent comparative times across races.

What about the data quality?

  • World Triathlon puts efforts to uniform/standardize race reports and race timings.
  • But some manual cleaning is required.
    • Some information, such as the permission of wearing wetsuit, is often missing.
    • Also, I could not find any way to access the rankings of past years via the API.
  • Obviously, variations in distances, weather conditions, athlete levels and scenarios make comparisons between race difficult.
    • Fortunately, the number of races is large enough to smooth out these variations and provide interesting insights.
  • Comparing swim-times can be tricky, not just because the distances vary between events, but also because the positions of timing mats are not consistent.
    • Most mats are placed directly at the exit of the water, while others are located at the entrance of transition area, which can be hundreds of meters further.
    • Future results should be more consistent: World Triathlon is currently working on standardising placement of timing mats for all to be at swim exit as well as T-In.

๐Ÿ—‚๏ธ CONTENT

My favourite sections are marked with โญ.

โ“ THREE SPORTS

This section:

  • Investigates the statement: "swim, bike and run account each for one third of the total duration".
  • Analyses some transition data: T1 and T2.
sport_proportion.png
Proportion of swim, t1, bike, t2 and run in the overall race duration. For women/men and sprint/olympic formats.

Clearly not 1/3 + 1/3 + 1/3!

  • During a sprint format, women spend twice as much time running (run+t1+t2) as they do swimming.
  • Athletes spend more than three times longer on their bikes than in the water.
  • The proportion of the three sports remains similar between sprint and olympic formats.
    • The proportion of transition time is almost halved: transitions are fixed durations while the race time doubles from sprint to olympic.
  • Women spend less relative time on the swim but more on the run.

Why 1.5+40+10 as a format?

  • The format was allegedly introduced by the producers of the U.S. Triathlon Series (USTS) ( ๐Ÿ‡บ๐Ÿ‡ธ ) in the mid-1980s
  • According to this article by the Bass Lake triathlon:
  • "A need was seen to standardize the distances and make them more in sync with each individual sport. USTS is credited with inventing the distances of the modern day Olympic distance triathlon. For the swim, 1500 meters was chosen because it is the standard long distance competitive swimming event. For the bike, USTS chose 40 kilometers because it was the standard solo time trial distance in bike racing. And the choice for the run was the standard road distance of 10 km. Note that the distances were not chosen to be symmetrical nor were they in direct ratio to Ironman distances."

Drafting is allowed on bike.

  • Otherwise, gaps would probably be much larger, and probably the bike skills would become much more decisive.
  • I could not find when the drafting was first allowed on elite races.
    • In the first ITU world championship in 1989 in Avignon ( ๐Ÿ‡ซ๐Ÿ‡ท ), drafting seems to be banned.
    • Banning drafting seems nowadays complicated, considering the density of the swim level. Individual starts, e.g. every minute or so like during cycling time-trials, would be an option, and would give the bike section a much higher importance.

T1 and T2 represent a tiny part of the overall racing time, yet they are crucial!

  • E.g. to catch a good bunch at bike.

Comparing to other triathlon formats:

  • Gustav Iden won the IRONMAN world championships in 2022 with the following times.
  • "00:48:23" (3.8k), "4:11:06" (180k), "2:36:15" (42k), which represents:
    • 10.6% swim
    • 55.1% bike
    • 34.3% run
  • I.e. More run and less swim, which was expected from the distances: 3.8-180-42 compared to 1.5-40-10.
Click to expand - ๐Ÿ‡๐Ÿข Some of the shortest and longest transitions.

(using Men data)

T1

The duration of T1 depends the distance between the water and the transition area, as well as on the position of the timing mats.

  • E.g. when the swim exit happens on a sand beach (Mooloolaba ( ๐Ÿ‡ฆ๐Ÿ‡บ ) , Huatulco ( ๐Ÿ‡ฒ๐Ÿ‡ฝ ), Hy-Vee ( ๐Ÿ‡บ๐Ÿ‡ธ )), timing mats are rarely placed directly on the sand, but instead further, close to the transition area.
  • Short T1 are often related to unusually long swim times: the best swim time 19:14 at Hy-Vee ( ๐Ÿ‡บ๐Ÿ‡ธ ) 2010, and 22 min at Mooloolaba ( ๐Ÿ‡ฆ๐Ÿ‡บ ) 2012!
    • That is a shame: in such cases swim split times are incorrect!
    • Mooloolaba ( ๐Ÿ‡ฆ๐Ÿ‡บ ) has apparently corrected the position of the timing mats: the fastest T1 in 2015 was 01:15 (against 00:13 in 2012).
T1 EVENT
00:13 2012 Mooloolaba ITU Triathlon World Cup ( ๐Ÿ‡ฆ๐Ÿ‡บ )
00:16 2010 Huatulco ITU Triathlon World Cup ( ๐Ÿ‡ฒ๐Ÿ‡ฝ )
00:16 2009 Hy-Vee ITU Triathlon Elite Cup ( ๐Ÿ‡บ๐Ÿ‡ธ )
... ...
01:25 2019 Madrid ITU Triathlon World Cup ( ๐Ÿ‡ช๐Ÿ‡ธ )
01:33 2017 ITU World Triathlon Grand Final Rotterdam ( ๐Ÿ‡ณ๐Ÿ‡ฑ )
01:40 2023 World Triathlon Championship Series Montreal ( ๐Ÿ‡จ๐Ÿ‡ฆ )
02:35 2011 Guatape ITU Triathlon World Cup ( ๐Ÿ‡จ๐Ÿ‡ด )

T2

The duration of T2 is mainly related to the size of the transition area (impacted by the number of participants) and to the position of the timing mats.

  • Variations are smaller than for T1.
T2 EVENT
00:14 2011 Mooloolaba ITU Triathlon World Cup
00:14 2012 Mooloolaba ITU Triathlon World Cup
00:15 2010 Monterrey ITU Triathlon World Cup
... ...
00:35 2019 Banyoles ITU Triathlon World Cup
00:36 2016 Montreal ITU Triathlon World Cup
00:41 2017 Salinas ITU Triathlon World Cup

T1+T2

On average, T1+T2 takes 01:11 (men) and 01:18 (women).

  • As mentioned for T1, very short times are mainly due to wrong positions of the timing mats after the swim: they are placed at the entrance of the transition area instead of at the water exit.
  • Longer T1+T2 means athletes must run more distance to reach their bikes / shoes.
    • In Montreal ( ๐Ÿ‡จ๐Ÿ‡ฆ ) 2023, it was 02:13: more than one minute than usual.
    • For a sprint format, this is substantial: compared to the 15:05 average men's 5k, this long transition makes the run 7% longer!
T1+T2 EVENT DISTANCe
00:27 2012 Mooloolaba ITU Triathlon World Cup OLYMPIC
00:34 2009 Hy-Vee ITU Triathlon Elite Cup OLYMPIC
00:37 2010 Huatulco ITU Triathlon World Cup OLYMPIC
00:44 2022 World Triathlon Cup Vina del Mar SPRINT
... ... ...
01:48 2014 ITU World Triathlon Stockholm SPRINT
01:56 2017 ITU World Triathlon Grand Final Rotterdam OLYMPIC
02:13 2023 World Triathlon Championship Series Montreal SPRINT
02:57 2011 Guatape ITU Triathlon World Cup SPRINT

โฑ๏ธ PACES

It is worth recalling the data settings:

  • Year: from 2009 to mid-2024.
  • Event: only world-cups, world-series (called WTCS) and games related events.
  • Distance: only sprint and olympic formats.
  • At least 25 finishers.
  • โš ๏ธ IMPORTANT: for each leg of a race (swim, bike, run), an average of 5 times is computed, using the 5-th to 9-th best times of the leg.
sports_paces.png
Distributions of times and paces.

Note: The distribution of swim times includes races both with- and without wetsuit. A subsequent section does the distinction (see its "second method" subsection).

Click to expand - ๐Ÿ… Same plots for the Top-3.
sports_paces_top3.png
Times and paces, considering the Top-3 in each leg.
Click to expand - ๐Ÿ Same plots for the Top-10.
sports_paces_top10.png
Times and paces, considering the Top-10 in each leg.
Click to expand - โฑ๏ธ Pace/speed/5k/10k conversion for the run.
Run Pace (M:SS) Speed (km/h) 5k Time (MM:SS) 10k Time (MM:SS)
2:55 20.6 14:35 29:10
2:56 20.5 14:40 29:20
2:57 20.3 14:45 29:30
2:58 20.2 14:50 29:40
2:59 20.1 14:55 29:50
3:00 20.0 15:00 30:00
3:01 19.9 15:05 30:10
3:02 19.8 15:10 30:20
3:03 19.7 15:15 30:30
3:04 19.6 15:20 30:40
3:05 19.5 15:25 30:50
3:06 19.4 15:30 31:00
3:07 19.3 15:35 31:10
3:08 19.1 15:40 31:20
3:09 19.0 15:45 31:30
3:10 18.9 15:50 31:40
3:11 18.8 15:55 31:50
3:12 18.8 16:00 32:00
3:13 18.7 16:05 32:10
3:14 18.6 16:10 32:20
3:15 18.5 16:15 32:30
3:16 18.4 16:20 32:40
3:17 18.3 16:25 32:50
3:18 18.2 16:30 33:00
3:19 18.1 16:35 33:10
3:20 18.0 16:40 33:20
3:21 17.9 16:45 33:30
3:22 17.8 16:50 33:40
3:23 17.7 16:55 33:50
3:24 17.6 17:00 34:00
3:25 17.6 17:05 34:10
3:26 17.5 17:10 34:20
3:27 17.4 17:15 34:30
3:28 17.3 17:20 34:40
3:29 17.2 17:25 34:50
3:30 17.1 17:30 35:00
3:31 17.1 17:35 35:10
3:32 17.0 17:40 35:20
3:33 16.9 17:45 35:30
3:34 16.8 17:50 35:40
3:35 16.7 17:55 35:50
3:36 16.7 18:00 36:00
3:37 16.6 18:05 36:10
3:38 16.5 18:10 36:20
3:39 16.4 18:15 36:30
3:40 16.4 18:20 36:40

Some outliers have been excluded:

Sprint vs olympic format:

  • ๐ŸŠ Swim paces are almost identical for 750m and 1500m: about 1s / 100m difference.
  • ๐Ÿšด There is less than 1km/h difference between the 20k and 40k bike.
  • ๐Ÿƒ The 10k run requires 7s/km more than the 5k.

The next section analyses the time differences between women's and men's performance.

๐Ÿ‘ซ WOMEN VS MEN

The difference in percent is computed with:

diff_in_percent = (time_w - time_m) / time_m

I.e.
time_w = (1 + diff_in_percent) * time_m

This percentage says "by how much are women slower than men".

  • To know "by how much are men faster", use diff_in_percent / (1 + diff_in_percent).
wm.png
By how much are women slower?

Notes about swim data:

wm_over_years.png
By how much are women slower? Evolution over years.
Click to expand - ๐Ÿ“ˆ Evolution over years, considering only WTCS and games-related events.
wm_over_years_no_world_cup.png
By how much are women slower? Evolution over years. Only WTCS and games-related events.

Findings:

  • ๐ŸŠ The swim is the sport where the relative difference between women and men is the smallest.
    • Swimming is highly technique-oriented.
    • Women often excel in technical sports because these rely less on raw strength and more on skill, coordination, and efficiency.
    • Women and men have different buoyancy, as explained by Maria Francesca Piacentini, in this episode (at 19:00) of the triathlon show podcast.
    • As reported by this 2019 article by Romuald Lepers: "Elite female athletes generally have 7โ€“12% more body fat than males (Fleck, 1983; Heydenreich et al., 2017). As fat is buoyant in water, women are less penalized than men in swimming than they are within terrestrial events such as cycling and running. Male triathletes also possess a larger muscle mass, greater muscular strength and lower relative body fat than female triathletes (Knechtle et al., 2010a)."
  • ๐Ÿƒ The run is where the difference is the largest.
    • Men typically have greater muscle mass and aerobic capacity, which can provide an advantage in endurance activities like running.
  • ๐Ÿ“ The standard deviation is higher for swim and lower for run.
    • Because swim conditions (wind, current, temperature) can vary and athletes may follow non-straight swim lines leading to larger swim distances?
  • ๐Ÿ“‰ The w/m difference has not significantly reduced on the years, except for the run leg of the sprint-format races (-0.13 % / year).
    • Note: Probably some data processing should be applied to the line fitting.
      • E.g. to account for the very low number of points during the covid pandemic?
    • In WTCS and games-related events, the w/m swim gap has reduced (-0.11 % / year) as well.

These percentages can be compared with those of swim, bike or run competitions:

Click to expand - ๐ŸŠ๐Ÿšด๐Ÿƒ Comparisons to individual swim/bike/run.

diff_percent = (time_w - time_m) / time_m is computed for a couple of events taken individually.

  • It would be statistically more significant to consider many events and compute averages.
  • I have not taken the time to do that.

๐ŸŠ SWIM

  • ๐Ÿ‡ฏ๐Ÿ‡ต 2021 Tokyo 800m: diff_percent = 7.7% considering the 4th to 8th men and women:
    • times_m = [7:42.68, 7:45.00, 7:45.11, 7:49.14, 7:53.31]
    • times_w = [8:19.38, 8:21.93, 8:22.25, 8:24.56, 8:26.30]
    • In previous olympic games, there was no men's 800m.
      • In Rio, the 4th-8th women average was less than 1s different to Tokyo.
  • ๐Ÿ‡ฏ๐Ÿ‡ต 2021 Tokyo 10k: diff_percent = 8.2% considering the 5th to 9th men and women:
    • times_m = [1:49:29, 1:50:23, 1:51:30, 1:51:32, 1:51:37]
    • times_w = [1:59:35, 1:59:36, 1:59:37, 2:00:10, 2:00:57]
    • Open-water conditions are closer to the one of triathlon, but the 10k race involves more strategy.

๐Ÿšด BIKE

  • Which race format?
    • Non-TT cycling race often involve strategy, making time comparison between women and men irrelevant.
    • ICU time-trial (TT) world championships uses different distances for women and men, making the comparison difficult.
    • Fortunately regional and national TT championships use the same distance and can therefore give relevant examples of diff_percent.
  • ๐Ÿ‡บ๐Ÿ‡ธ 2024 USA TT: diff_percent = 11.1 % between Taylor Knibb (41:54) and Brandon McNulty (37:42)
  • ๐Ÿ‡ช๐Ÿ‡บ 2023 European TT: diff_percent = 12.3% considering the 5th to 9th men and women:
    • times_m = [32:43.77, 32:45.91, 32:52.03, 32:55.19, 32:55.22]
    • times_w = [36:42.01, 36:49.02, 36:53.75, 37:02.27, 37:05.13]
  • ๐Ÿ‡ช๐Ÿ‡บ 2022 European TT: diff_percent = 14.8% considering the 5th to 9th men and women:
    • times_m = [27:42.81, 28:01.56, 28:17.61, 28:18.47, 28:27.19]
    • times_w = [32:00.87, 32:01.10, 32:30.76, 32:32.76, 32:37.85]

๐Ÿƒ RUN

  • diff_percent is computed in the same way.
    • But keep in mind that running competitions are often very strategic: the time matters often less than the ranking.
      • For instance, on the athletic track the pace can be kept low until the last lap.
    • Comparing world/continental/national records for women vs men is an option, but the level of one outstanding person says nothing about the average level.
  • ๐Ÿ‡ช๐Ÿ‡บ 2024 European 10k: diff_percent = 13.8% considering the 5th to 9th men and women:
    • times_m = [31:34.90, 31:38.45, 32:15.91, 32:16.85, 32:17.24]
    • times_w = [28:01.42, 28:04.43, 28:09.87, 28:10.97, 28:11.61]
  • ๐Ÿ‡ช๐Ÿ‡บ 2024 European 5k: diff_percent = 11.6% considering the 5th to 9th men and women:
    • times_m = [14:44.72, 14:58.28, 15:00.05, 15:02.56, 15:05.66]
    • times_w = [13:23.26, 13:24.54, 13:24.80, 13:25.08, 13:25.65]
  • ๐Ÿ‡ซ๐Ÿ‡ท 2024 Paris marathon: diff_percent = 13.7% considering the 5th to 9th men and women:
    • times_m = [2:24:48, 2:26:00, 2:26:01, 2:26:03, 2:26:08]
    • times_w = [2:07:37, 2:07:39, 2:07:44, 2:08:41, 2:09:04]

Focus on the swim: ๐ŸŠ

wm_swim.png
By how much swim women slower, with / without wetsuit? ๐ŸŠ ๐Ÿฉฑ

๐Ÿ’ก FINDING:

  • The women/men time difference can be considered constant, regardless of the distance (sprint / olympic) and the equipment (wetsuit or not): women swim ~8.8% slower.
  • This finding is consistent with Vleck et al., 2011.
  • This result will be used in a subsequent section to determine the benefit provided by the wetsuit.

๐ŸŠ SWIM GAPS

This section tries to answer the question "Should worse swimmers be happy when the wetsuit is allowed?".

  • My a priori reflexion was:
    • The wetsuit makes everyone swim faster.
    • => The swim takes less time.
    • => => Gaps are smaller.
    • => => => Worse swimmers are happy.

Approach:

  • For each race, the gap between the 5-9th first swimmers and 5-9th last swimmers is computed.
  • These gaps are split into two groups: with-wetsuit and without-wetsuit.
  • The averages of both group are compared to determine if the wetsuit makes swim gaps smaller or larger.
swim_gaps.png
Comparing average gaps between good and "bad" swimmers with and without wetsuit.

๐Ÿ’ก FINDINGS:

  • Variations in the swim-pack length between event with-wetsuit and without-wetsuit are very small.
  • There is no evidence that worse swimmers should be happy about the wetsuit.
    • The wetsuit even tends to stretch the swim pack, especially for women.
    • To be honest, I would have expected the opposite!

POSSIBLE INTERPRETATION #1:

  • Wetsuits are typically worn in cold waters, often in seas and oceans, where waves can make swimming more challenging, potentially spreading out the pack.
  • However, there are many examples of sea and ocean swims that occur without wetsuits!
  • Therefore, I would dismiss this hypothesis.

POSSIBLE INTERPRETATION #2:

  • True, the swim is shorter in time.
  • But gaps do not significantly reduce because the benefit provided by the wetsuit differs between good and worse swimmers:
    • For 5-9th top swimmer: 5.4%.
    • For 20-24th top swimmer: 5.3%.
    • For last 9-5th top swimmer: 4.9%.
    • More details on this derivation can be found in the dedicated section.
  • In other words, despite the shorter swim duration, gaps do not reduce because top swimmers gain more benefits from the wetsuit.

QUESTION:

  • Would the wetsuit enable the slowest swimmers (last 9-5th) to catch the good swimmers (first 5-9th)?
    • On the olympic format, the gap is about 50s and 45s without wetsuit, while the fast women and men swim on average in 19:30 and 17:57.
    • A 4.9% improvement for the slowest swimmers represent 0.049 * (19:30 + 0:50) = 60s, and 0.049 * (17:57 + 0:45) = 55s.
    • Conclusion: the slowest swimmers with the benefit of the wetsuit would be ~10s faster that the good swimmers without.
Click to expand - Other comparisons.

Considering the "front-to-middle" distance (using the 20-24th swimming times instead of the last 5-9th), results looks similar: No significant gap reduction.

  • men+sprint+no_wetsuit may suffer from outliers: gaps at Tiszaรบjvรกros ( ๐Ÿ‡ญ๐Ÿ‡บ ) 2013 and 2016 were larger than 33s.
swim_gaps_20-24.png
Same computation as above, this time with gaps between 5-9th to 20-24th swimmers.

When considering world-series events only, the opposite trend occurs: the wetsuit tends to reduce the swim gaps.

  • However, as noted earlier, variations are very small, indicating that no significant effect of the wetsuit on swim gaps can be concluded.
swim_gaps-wcs.png
With world-series events only.

๐Ÿง WETSUIT BENEFIT

This section tries to estimate the benefit (in percent) offered by the wetsuit, defined as

improve_percent = (time_no_wetsuit - time_wetsuit) / time_no_wetsuit

I.e.
time_with_wetsuit = (1 - improve_percent) * time_without_wetsuit

Reminder:

  • The scope here is elite triathletes, going 5-9th out of water on top World Triathlon events.
  • Results would certainly differ for beginners and hobby triathletes.

The idea of the derivation is as follows:

  • Women have been found to swim on average ~8.8% slower than men, with the same equipment.

  • With examples where women had the wetsuit, but men did not, one can:

    • 1- Estimate, from the men's time, the time women WOULD HAVE done without the wetsuit (thanks to the ~8.8% rule).
    • 2- Compare the women's time with wetsuit (measured) with the women's time without (computed in 1-).
    • 3- Deduce the advantage provided by the wetsuit for the women.
    • 4- Note that improve_percent should be the same for women and men (because of the constant ~8.8% difference given the same equipment).

  • Advantages of this method:

    • Proper environment: data comes from real races, as opposed to studies in 50m or even 25m pools.
    • No need to know the exact swim distance: what matters is that men and women swim the exact same course. This assumption is reasonable because the buoys should not move between the two races.
    • Data-based: the 8.8% should be robust since it leverages results from many races (230 events).
    • Low cost: all the data is available online for free.

  • Limitations:

    • The swim conditions are not guaranteed to be equal.
    • The limited number of events where the "women-with / men-without" scenario occurs.
      • Only five races fit this scenario, but the variability is low: std = 0.5%.
Click to expand - โœ๏ธ Derivation of the formula. 
# VARIABLES
  swim_w:             time of women. no wetsuit.
  swim_m:             time of men. no wetsuit.
  swim_w_wet:         time of women. with wetsuit.
  
  improve_percent:    advantate (in percent) brought by the wetsuit. UNKNOWN.
  wm_percent:         relative delay of women over men assuming same equipment.
  wm_percent_w_fast:  relative delay when women have wetsuit, but men do not.

# FORMULA
  would W have not had wetsuit:
    (1)  `swim_w * (1 - improve_percent) = swim_w_wet`  =>  `swim_w = swim_w_wet / (1 - improve_percent)` 
  and this time would be related to swim_m:
    (2)  `swim_w = swim_m * (1 + wm_percent)`
  using (1) == (2):
    (3)  `swim_w_wet = swim_m * (1 + wm_percent) * (1 - improve_percent)`
  
  on the other hand:
    (4)  `swim_w_wet = swim_m * (1 + wm_percent_w_fast)`
  
  using (4) == (3):
    (5)  `(1 + wm_percent) * (1 - improve_percent) = (1 + wm_percent_w_fast)`
  re-written:
         `(1 - improve_percent) = (1 + wm_percent_w_fast) / (1 + wm_percent)`
  hence
         `improve_percent = 1 - (1 + wm_percent_w_fast) / (1 + wm_percent)`

# EXAMPLE
  [wm_percent = 8.8%]
  [wm_percent_w_fast = 2.9%]
  => improve_percent = 1 - (1+0.029)/(1+0.088) = 5.4%
wetsuit.png
Estimating the benefit brought by the wetsuit, using results of races where women swam with but men without.

๐Ÿ’ก FINDING:

  • The wetsuit brings an advantage of ~5.4% to top swimmers (5th-9th).
  • Put differently, top swimmers (top 5-9) swim ~5.7% slower without wetsuit.
    • 0.054 / (1-0.054) = 0.057.

โš ๏ธ CRITICISMS AND IDEAS FOR IMPROVEMENT:

  • 1) Uncertainty:
    • How to account for uncertainties in wm_percent_w_fast and wm_percent in the improve_percent = 1 - (1 + wm_percent_w_fast) / (1 + wm_percent) formula?
      • So far, the standard deviations were computed (ยฑ 3.0% and ยฑ 0.5%), telling how spread out the observed w/m-swim-diff percentages are:
        • wm_percent = 8.8% ยฑ 3.0%.
        • wm_percent_w_fast = 2.9% ยฑ 0.5%.
        • Concretely, in the case of wm_percent, ยฑ 3.0% can be interpreted as: "~68% of the observed w/m-swim-diff percentages lie between 8.8%-3% and 8.8%+3%".
      • I am not sure, but from what I understood, in order to produce a confidence interval for improve_percent, the standard errors (SE) should be used instead:
        • SE(wm_percent) = 3.0% / sqrt(230) = 0.2%.
        • SE(wm_percent_w_fast) = 0.5% / sqrt(5) = 0.2%.
    • In statistics, this question is known as Propagation of Uncertainty.
      • Approach #1 (simple): perform calculations using the extremes of the error intervals to see where improve_percent falls.
        • Here, applying the combinations (-0.2%, -0.2%), (-0.2%, +0.2%), (+0.2%, -0.2%) and (+0.2%, +0.2%) to (wm_percent = 8.8%, wm_percent_w_fast = 2.9%).
        • This results in the interval improve_percent = 5.4% with 0.4% standard error.
      • Approach #2 (using partial derivatives):
        • Using this tool, I obtain improve_percent = 5.4% with 0.3% standard error.
  • 2) Events consistency:
    • wm_percent_w_fast = 2.9% was computed from five "women-with-wetsuit, men-without" examples that all have the following properties: WTCS and olympic-format.
      • The five venues are: Yokohama ( ๐Ÿ‡ฏ๐Ÿ‡ต ) (twice), Cagliari ( ๐Ÿ‡ฎ๐Ÿ‡น ), Stockholm ( ๐Ÿ‡ธ๐Ÿ‡ช ) and Edmonton ( ๐Ÿ‡จ๐Ÿ‡ฆ ).
    • In contrast, wm_percent = 8.8% was obtained by considering all the sprint- and olympic-format WCTS, world-cups and games-related events since 2009, totaling 230 events.
      • This is inconsistent.
    • Instead, wm_percent should be estimated considering events with similar swim conditions as those for wm_percent_w_fast.
      • Idea #1: Only consider events with the same format (olympic) and event-category (WTCS).
        • This leads to a slightly lower estimate (wm_percent = 8.1% ยฑ 2.4%), resulting in an improve_percent closer to 4.8%.
      • Idea #2: Further restrict Idea #1 (same format and event-category), by considering only with the same venues.
        • The four venues mentioned have hosted multiple olympic-WTCS: 20 times, from which 15 had women and men sharing the same equipment for the swim.
        • This gives wm_percent = 7.4% ยฑ 1.6% (SE = 1.6/sqrt(15) = 0.4%), leading to improve_percent = 4.2% with 0.3% standard error.
  • 3) Additional examples:
Click to expand - ๐ŸŒ Events used for the derivation.
YEAR EVENT DIFF (%) WOMEN-WITH vs MEN-without BENEFIT (%) DISTANCE EVENT CATEGORY
2024 Cagliari ( ๐Ÿ‡ฎ๐Ÿ‡น ) 2.3% 4.7% olympic WTCS
2021 Edmonton ( ๐Ÿ‡จ๐Ÿ‡ฆ ) 2.4% 4.6% olympic WTCS
2022 Yokohama ( ๐Ÿ‡ฏ๐Ÿ‡ต ) 2.9% 4.2% olympic WTCS
2015 Stockholm ( ๐Ÿ‡ธ๐Ÿ‡ช ) 3.3% 3.8% olympic WTCS
2024 Yokohama ( ๐Ÿ‡ฏ๐Ÿ‡ต ) 3.5% 3.7% olympic WTCS

๐ŸŽ“ Literature review

Here are some findings of related scientific works:

  • This 2019 research by Gay et al. asked 33 swimmers to perform 2ร—400-m maximal front crawl in a 25-m swimming pool, with wetsuit and with swimsuit. Participants were good swimmers, but not as fast as ITU elite athletes: 1:27 / 100m average on the 400m with swimsuit.
    • The wetsuit allows for a ~6% improvement.
    • Interestingly: "Swimmers reduced stroke rate and increased stroke length (by 4%) to benefit from the hydrodynamic reduction of the wetsuit and increase their swimming efficiency."
  • This 2022 meta study by Gay et al. concludes from 26 studies, a "3.2โ€“12.9% velocity increments in distances ranging from 25 to 1500 m" for the full-body wetsuit.
    • The range is broad: it depends on many factors such as the profile of the swimmer (age, level, triathlete or swimmer), the swimming conditions (temperature, 25m pool vs open water) and the wetsuit itself.
    • This interview of Ana Gay in triathlete.com gives a good introduction to her study.
  • In this episode by scientifictriathlon.com, Maria Francesca Piacentini mentions some of her research on wetsuit, and claims 6% to 11% improvements.
    • Apart from the time reduction, her team found that wetsuit usage can make athletes feel less fatigued: during the 2x7x200m tests, the stroke index and the stroke length significantly decreased in the swimsuit condition, whereas they remained relatively stable in the wetsuit condition.
  • This article from sports-performance-bulletin reports improvements by 3% to 7%.
    • Sources are not explicitly referenced, but the article probably mentions this 1995 research by Chatard et al. which apart from computing improvements, shows that the impact of the wetsuit is very different for competitive swimmers than for competitive triathletes.

The 5.4% improvement for top swimmer (5th-9th) derived from the World Triathlon data seems compatible with the findings of these research publications.

Is the wetsuit worth for 300m?

When they can choose, pro athletes decide to use the wetsuit for ~300m swim, e.g. during mixed team relay. Examples:

Is this decision sound?

Click to expand - ๐Ÿง Answering the question "wetsuit for 300m?"

Over 300m a wetsuit should save ~12s (derivation below).

  • Most of the time, athletes spend just a bit less than 6s in transition to remove their wetsuit, and put it correctly into their box.
    • Maybe 4s or 5s when everything goes well? But up to 10s when one leg does not want to get out.
  • Conclusion: the overall time gain of wearing the wetsuit for 300m seems to be positive: around 6s.
    • Apart from time saving, wetsuits offer additional benefits: temperature comfort and probably saving some energy in the legs. Also, athletes may a couple of seconds to breath while removing it at T1.
  • Conversely, assuming an overall 6s spent for removal, wearing a wetsuit is beneficial above ~150m for top athletes. Under ~150m, time is lost overall.

Derivation: The time for a 300m swim is probably around 3:30 - 4:00.

Benefit 03:30 03:35 03:40 03:45 03:50 03:55 04:00 04:05 04:10
4.0% 8.4 8.6 8.8 9.0 9.2 9.4 9.6 9.8 10.0
4.2% 8.8 9.0 9.2 9.5 9.7 9.9 10.1 10.3 10.5
4.4% 9.2 9.5 9.7 9.9 10.1 10.3 10.6 10.8 11.0
4.6% 9.7 9.9 10.1 10.4 10.6 10.8 11.0 11.3 11.5
4.8% 10.1 10.3 10.6 10.8 11.0 11.3 11.5 11.8 12.0
5.0% 10.5 10.8 11.0 11.2 11.5 11.8 12.0 12.2 12.5
5.2% 10.9 11.2 11.4 11.7 12.0 12.2 12.5 12.7 13.0
5.4% 11.3 11.6 11.9 12.2 12.4 12.7 13.0 13.2 13.5
5.6% 11.8 12.0 12.3 12.6 12.9 13.2 13.4 13.7 14.0
5.8% 12.2 12.5 12.8 13.1 13.3 13.6 13.9 14.2 14.5
6.0% 12.6 12.9 13.2 13.5 13.8 14.1 14.4 14.7 15.0
6.2% 13.0 13.3 13.6 14.0 14.3 14.6 14.9 15.2 15.5
6.4% 13.4 13.8 14.1 14.4 14.7 15.0 15.4 15.7 16.0
6.6% 13.9 14.2 14.5 14.9 15.2 15.5 15.8 16.2 16.5
6.8% 14.3 14.6 15.0 15.3 15.6 16.0 16.3 16.7 17.0
7.0% 14.7 15.1 15.4 15.8 16.1 16.5 16.8 17.2 17.5

The benefit of the wetsuit could be estimated with a second method (less reliable):

Click to expand - Second approach.
wetsuit_2.png
Comparing duration of swims with and without wetsuit.

The idea is to compare the average swim durations, with- and without wetsuit:

  • improve_percent = (time_no_wetsuit - time_wetsuit) / time_no_wetsuit.
  • This formula is applied to four cases: W-sprint, W-olympic, M-sprint, M-olympic. And results are written in the title of the above figure.
  • Note how the swim histograms of the โฑ๏ธ PACES section reveals two distinct modes: with- (in violet, with lower times) and without wetsuit (larger times).

Outliers:

โš ๏ธ CRITICISMS.

  • This method incorporates more data, but yields less reliable results to the following factors:
    • ๐Ÿ“ Swim course distances vary between races.
    • โฑ๏ธ Timing methods (e.g., timing mat locations) are not always consistent.
  • For exemple:
    • The difference between the fastest and the slowest sprint+men+same-equipment swim exceeds 2 minutes.
      • Specifically, swim times at Mooloolaba ( ๐Ÿ‡ฆ๐Ÿ‡บ ) are unusually fast considering the absence of wetsuit.
    • For olympic+men+no-wetsuit, the difference is 5:45.
  • Clearly, swim distances or timing methods vary between races, even when they are labeled with the same format.

๐ŸŽฏ RACE SCENARIO

This section looks at the race dynamics of WCTS and games-related events (no world-cup).

  • Based on the information "does a bunch manage to breakaway on the bike?" ๐Ÿฆ

To obtain this information, the size of the front pack at the end of the bike is estimated as follows:

  • Compute, for each athlete, the cumulative times after the bike:
    • start_to_t2 = swim + t1 + bike.
  • Identify which athlete enters T2 first: min(start_to_t2).
  • Count how many athletes enter T2 pack_duration_s, e.g. 10s, or less after this first athlete.
  • This gives the size of the front pack at the end of the bike.

Note: A breakaway on the bike can happen:

  • Either by attack and escape from the main bike pack (rare).
  • Or directly from the swim (most common).
  • It should be possible to automatically retrieve the breakaway type among the two, using the swim rank of athletes composing the breakaway. (I have not done it).

Two additional pieces of information are retrieved:

  • ๐Ÿ† winner_in_front_pack: was the winner already in the leading group at the end of the bike, or did she/he come back on the run?
  • ๐Ÿ‘Ÿ is_best_runner_in_front_pack.
scenarios.png
Size of the front pack at the end of the bike.

About the number of finishers:

  • ~4.5 more finishers in men's races.
    • The standard deviation is higher for women's race.
  • ~3.5 more** finishers in sprint format races.
    • Because the olympic format is longer, weaker athletes are more likely to be caught and eliminated by being lapped?
  • It would be interesting to know the number of starters as well.

On the olympic format, men are more likely to break away.

  • Because they are stronger on the bike?
    • Georgia Taylor-Brown ( ๐Ÿ‡ฌ๐Ÿ‡ง ), one of the strongest rider in the field, admits in this video: "I would love to be able to attack and stay away, but I do not have that power."
    • Probably bike is the sport among three where top athletes will progress the most in the future.
  • At the same time, the probability to win for athletes from the front group is much lower for men than for women.

Small front packs, i.e. small breakaways, are more likely on the olympic format.

  • Possibly because a longer swim leads to larger gaps at T1?
  • On the other hand the bike is longer than for the sprint format, which should give more time for the other packs to catch up.

Some very large front groups at T2 (London ( ๐Ÿ‡ฌ๐Ÿ‡ง ) at the top ๐Ÿ’‚ ):

PACK_SIZE YEAR WINNER DISTANCE CAT EVENT
53 2011 Helen Jenkins ( ๐Ÿ‡ฌ๐Ÿ‡ง ) OLYMPIC WTCS 2011 Dextro Energy Triathlon - ITU World Championship Series London ( ๐Ÿ‡ฌ๐Ÿ‡ง )
46 2012 Erin Densham ( ๐Ÿ‡ฆ๐Ÿ‡บ ) OLYMPIC WTCS 2012 Dextro Energy World Triathlon Sydney ( ๐Ÿ‡ฆ๐Ÿ‡บ )
39 2011 Paula Findlay ( ๐Ÿ‡จ๐Ÿ‡ฆ ) OLYMPIC WTCS 2011 Dextro Energy Triathlon - ITU World Championship Series Sydney ( ๐Ÿ‡ฆ๐Ÿ‡บ )
39 2011 Andrea Hansen ( ๐Ÿ‡ณ๐Ÿ‡ฟ ) OLYMPIC WTCS 2011 Dextro Energy Triathlon - ITU World Championship Series Yokohama ( ๐Ÿ‡ฏ๐Ÿ‡ต )
37 2012 Nicola Spirig ( ๐Ÿ‡จ๐Ÿ‡ญ ) OLYMPIC WTCS 2012 ITU World Triathlon Madrid ( ๐Ÿ‡ช๐Ÿ‡ธ )
... ... ... ... ... ...
52 2014 Mario Mola ( ๐Ÿ‡ช๐Ÿ‡ธ ) SPRINT WTCS 2014 ITU World Triathlon London ( ๐Ÿ‡ฌ๐Ÿ‡ง )
51 2023 Alex Yee ( ๐Ÿ‡ฌ๐Ÿ‡ง ) OLYMPIC GAMES 2023 World Triathlon Olympic Games Test Event Paris ( ๐Ÿ‡ซ๐Ÿ‡ท )
49 2023 Alex Yee ( ๐Ÿ‡ฌ๐Ÿ‡ง ) SPRINT WTCS 2023 World Triathlon Championship Series Abu Dhabi ( ๐Ÿ‡ฆ๐Ÿ‡ช )
46 2016 Mario Mola ( ๐Ÿ‡ช๐Ÿ‡ธ ) OLYMPIC WTCS 2016 ITU World Triathlon Yokohama ( ๐Ÿ‡ฏ๐Ÿ‡ต )
45 2011 Brad Kahlefeldt ( ๐Ÿ‡ฆ๐Ÿ‡บ ) OLYMPIC WTCS 2011 Dextro Energy Triathlon - ITU World Championship Series Hamburg ( ๐Ÿ‡ฉ๐Ÿ‡ช )
Click to expand - ๐Ÿ’ช Some wins via breakaway (with front-pack-size <= 3).
PACK_SIZE YEAR WINNER DISTANCE CAT EVENT
1 2011 Sarah Haskins ( ๐Ÿ‡บ๐Ÿ‡ธ ) OLYMPIC WORLD-CUP 2011 Monterrey ITU Triathlon World Cup ( ๐Ÿ‡ฒ๐Ÿ‡ฝ )
1 2016 Flora Duffy ( ๐Ÿ‡ง๐Ÿ‡ฒ ) OLYMPIC WTCS 2016 ITU World Triathlon Stockholm ( ๐Ÿ‡ธ๐Ÿ‡ช )
1 2018 Flora Duffy ( ๐Ÿ‡ง๐Ÿ‡ฒ ) OLYMPIC WTCS 2018 ITU World Triathlon Bermuda ( ๐Ÿ‡ง๐Ÿ‡ฒ )
1 2021 Taylor Knibb ( ๐Ÿ‡บ๐Ÿ‡ธ ) OLYMPIC WTCS 2021 World Triathlon Championship Finals Edmonton ( ๐Ÿ‡จ๐Ÿ‡ฆ )
2 2011 Nicky Samuels ( ๐Ÿ‡ณ๐Ÿ‡ฟ ) OLYMPIC WORLD-CUP 2011 Mooloolaba ITU Triathlon World Cup ( ๐Ÿ‡ฆ๐Ÿ‡บ )
2 2017 Flora Duffy ( ๐Ÿ‡ง๐Ÿ‡ฒ ) OLYMPIC WTCS 2017 ITU World Triathlon Yokohama ( ๐Ÿ‡ฏ๐Ÿ‡ต )
2 2021 Taylor Knibb ( ๐Ÿ‡บ๐Ÿ‡ธ ) OLYMPIC WTCS 2021 World Triathlon Championship Series Yokohama ( ๐Ÿ‡ฏ๐Ÿ‡ต )
2 2022 Flora Duffy ( ๐Ÿ‡ง๐Ÿ‡ฒ ) OLYMPIC WTCS 2022 World Triathlon Championship Series Bermuda ( ๐Ÿ‡ง๐Ÿ‡ฒ )
3 2009 Emma Moffatt ( ๐Ÿ‡ฆ๐Ÿ‡บ ) OLYMPIC WTCS 2009 Dextro Energy Triathlon - ITU World Championship Series Hamburg ( ๐Ÿ‡ฉ๐Ÿ‡ช )
3 2014 Jodie Stimpson ( ๐Ÿ‡ฌ๐Ÿ‡ง ) OLYMPIC WTCS 2014 ITU World Triathlon Auckland ( ๐Ÿ‡ณ๐Ÿ‡ฟ )
3 2016 Helen Jenkins ( ๐Ÿ‡ฌ๐Ÿ‡ง ) OLYMPIC WTCS 2016 ITU World Triathlon Gold Coast ( ๐Ÿ‡ฆ๐Ÿ‡บ )
3 2016 Flora Duffy ( ๐Ÿ‡ง๐Ÿ‡ฒ ) OLYMPIC WTCS 2016 ITU World Triathlon Grand Final Cozumel ( ๐Ÿ‡ฒ๐Ÿ‡ฝ )
3 2017 Flora Duffy ( ๐Ÿ‡ง๐Ÿ‡ฒ ) OLYMPIC WTCS 2017 ITU World Triathlon Grand Final Rotterdam ( ๐Ÿ‡ณ๐Ÿ‡ฑ )
3 2019 Katie Zaferes ( ๐Ÿ‡บ๐Ÿ‡ธ ) OLYMPIC WTCS 2019 MS Amlin World Triathlon Bermuda ( ๐Ÿ‡ง๐Ÿ‡ฒ )
3 2019 Julie Derron ( ๐Ÿ‡จ๐Ÿ‡ญ ) OLYMPIC WORLD-CUP 2019 Weihai ITU Triathlon World Cup ( ๐Ÿ‡จ๐Ÿ‡ณ )
3 2021 Maya Kingma ( ๐Ÿ‡ณ๐Ÿ‡ฑ ) OLYMPIC WTCS AJ Bell 2021 World Triathlon Championship Series Leeds ( ๐Ÿ‡ฌ๐Ÿ‡ง )
3 2024 Lena MeiรŸner ( ๐Ÿ‡ฉ๐Ÿ‡ช ) OLYMPIC WORLD-CUP 2024 World Triathlon Cup Samarkand ( ๐Ÿ‡บ๐Ÿ‡ฟ )
... ... ... ... ... ...
1 2013 Alistair Brownlee ( ๐Ÿ‡ฌ๐Ÿ‡ง ) OLYMPIC WTCS 2013 ITU World Triathlon Stockholm ( ๐Ÿ‡ธ๐Ÿ‡ช )
1 2018 Casper Stornes ( ๐Ÿ‡ณ๐Ÿ‡ด ) OLYMPIC WTCS 2018 ITU World Triathlon Bermuda ( ๐Ÿ‡ง๐Ÿ‡ฒ )
1 2023 Morgan Pearson ( ๐Ÿ‡บ๐Ÿ‡ธ ) OLYMPIC WORLD-CUP 2023 World Triathlon Cup Karlovy Vary ( ๐Ÿ‡จ๐Ÿ‡ฟ )
2 2011 Kris Gemmell ( ๐Ÿ‡ณ๐Ÿ‡ฟ ) OLYMPIC WORLD-CUP 2011 Auckland ITU Triathlon World Cup ( ๐Ÿ‡ณ๐Ÿ‡ฟ )
2 2016 Rodrigo Gonzalez Lopez ( ๐Ÿ‡ฒ๐Ÿ‡ฝ ) OLYMPIC WORLD-CUP 2016 Chengdu ITU Triathlon World Cup ( ๐Ÿ‡จ๐Ÿ‡ณ )
3 2009 Jan Frodeno ( ๐Ÿ‡ฉ๐Ÿ‡ช ) OLYMPIC WTCS 2009 Dextro Energy Triathlon - ITU World Championship Series Yokohama ( ๐Ÿ‡ฏ๐Ÿ‡ต )
3 2011 Alistair Brownlee ( ๐Ÿ‡ฌ๐Ÿ‡ง ) OLYMPIC WTCS 2011 Dextro Energy Triathlon - ITU World Championship Series Kitzbuehel ( ๐Ÿ‡ฆ๐Ÿ‡น )
3 2014 Alistair Brownlee ( ๐Ÿ‡ฌ๐Ÿ‡ง ) OLYMPIC WTCS 2014 ITU World Triathlon Grand Final Edmonton ( ๐Ÿ‡จ๐Ÿ‡ฆ )
3 2015 Jonathan Brownlee ( ๐Ÿ‡ฌ๐Ÿ‡ง ) OLYMPIC WTCS 2015 ITU World Triathlon Gold Coast ( ๐Ÿ‡ฆ๐Ÿ‡บ )
3 2020 Vincent Luis ( ๐Ÿ‡ซ๐Ÿ‡ท ) OLYMPIC WORLD-CUP 2020 Karlovy Vary ITU Triathlon World Cup ( ๐Ÿ‡จ๐Ÿ‡ฟ )

Strong bikers and very versatile triathletes! ๐Ÿ‘

  • Especially Flora Duffy ( ๐Ÿ‡ง๐Ÿ‡ฒ ) and Alistair Brownlee ( ๐Ÿ‡ฌ๐Ÿ‡ง ).

Below is a more complicated figure: the bars show the evolution of the average first-pack size over the years.

scenarios_over_years.png
Size of front bike pack, over the years.

Run-comebacks, i.e. winning after not being in the front group after bike, are rare.

  • Apart from 2013-2016 (the era of Gwen Jorgensen ( ๐Ÿ‡บ๐Ÿ‡ธ )), no comeback has happened on women's olympic races and only a few on women's sprint format.
    • At the same time, the size of front-pack had reduced until 2022 and was even very small for some recent years (2017, 2019, 2021, 2022).
    • Now, top-women-swimmer can ride hard and run fast.
  • Helen Jenkins ( ๐Ÿ‡ฌ๐Ÿ‡ง ) explains in this 2024 video: "Women's races have definitely changed over the past few years. (...) 2021 was that breakaway era. It definitely comes back to that larger front group over the last couple of years."
    • That statement is perfectly consistent with the women's olympic bar plot.
  • The men's olympic races follow a similar same trend: the front group at T2 has, on average, never been as large, as in 2023 and 2024.

๐Ÿƒ How often does the best runner win? ๐Ÿ†

best_runner_wins.png
How often does the best runner win?

More than 2/3 races are won by the best runner.

  • This percentage is higher on the sprint than on the olympic format.
    • Probably because the swim is shorter: "good-runner-but-bad-swimmer"s are less likely to miss the front group on the bike.
  • This percentage could be:
  • The percentage drops to 50% when considering world-cup events only. Why?
    • Because athletes are not as complete as on WTCS?
    • There are more "good-runner-but-bad-swimmer"s who miss the front group on the bike, while good swimmers are not top runners?
Click to expand - Same plots for world-cup events only.
scenarios.png
Size of the front pack at the end of the bike.
scenarios_over_years.png
Size of front bike pack, over the years.
best_runner_wins_wc.png
How often does the best runner wins?

Why are there not more breakaways on the bike?

  • The mini areo TT bars have been banned in 2023.
  • Bike courses are mostly flat.
    • Events often take place on flat coasts, near to a sea, or in flat big cities.
    • Joel Filliol ( ๐Ÿ‡จ๐Ÿ‡ฆ ๐Ÿด๓ ง๓ ข๓ ณ๓ ฃ๓ ด๓ ฟ ) shares insights in this TTS podcast (around 43:00) about bike courses that allow for a break to stay away.
  • Because of U-turns.
    • As Michel Hidalgo ( ๐Ÿ‡ง๐Ÿ‡ท ) explains in this video, U-turns can be detrimental for breakaways.
  • Athletes may prefer to conserve energy for the run segment.
    • Valid of strong runners.
    • But for others? Maybe it is worth more to conserve energy by drafting and try to make top-20 with a good run, rather than risk a breakaway that might be caught, be burnt and come back home without any prize money or qualification points?
  • Not enough bike power?
    • Possibly, given these athletes need to be strong in swim and run as well, they cannot afford having massive legs.

๐Ÿš€ SPRINT FINISH

This section studies the time gap between the winner and the second.

sprint_finish.png
Time gap between the winner and the second.
  • In men's races, 17% (sprint format) and 10% (olympic) are won by a sprint finish, occurring 50% more often than in women's races.
  • Women's races offer examples of wins by very large margins.
sprint_finish_over_years.png
Time gap between the winner and the second, over years. And the proportion of events with a contested win finish (less than 3s difference between first and second).

Gaps between the winner and the second are on average:

  • ~Twice as large in olympic formats compared to sprint formats.
  • ~Twice as large for women compared to men.

It has been a long time since a women's olympic race was won by a sprint.

Click to expand - ๐Ÿ“ธ Some of the most contested finishes on the blue carpet.
WOMEN
YEAR VENUE DIST. RACE CATEGORY FIRST ( ๐Ÿฅ‡ ) SECOND ( ๐Ÿฅˆ )
2009 Madrid ( ๐Ÿ‡ช๐Ÿ‡ธ ) olympic WTCS Andrea Hansen ( ๐Ÿ‡ณ๐Ÿ‡ฟ ) Lisa Norden ( ๐Ÿ‡ธ๐Ÿ‡ช )
2010 Hamburg ( ๐Ÿ‡ฉ๐Ÿ‡ช ) olympic WTCS Lisa Norden ( ๐Ÿ‡ธ๐Ÿ‡ช ) Emma Moffatt ( ๐Ÿ‡ฆ๐Ÿ‡บ )
2010 Sydney ( ๐Ÿ‡ฆ๐Ÿ‡บ ) olympic WTCS Barbara Riveros ( ๐Ÿ‡จ๐Ÿ‡ฑ ) Andrea Hansen ( ๐Ÿ‡ณ๐Ÿ‡ฟ )
2010 Tiszaujvaros ( ๐Ÿ‡ญ๐Ÿ‡บ ) olympic world-cup Yuliya Yelistratova ( ๐Ÿ‡บ๐Ÿ‡ฆ ) Jodie Swallow ( ๐Ÿ‡ฌ๐Ÿ‡ง )
2011 Lausanne ( ๐Ÿ‡จ๐Ÿ‡ญ ) sprint WTCS Barbara Riveros ( ๐Ÿ‡จ๐Ÿ‡ฑ ) Emma Jackson ( ๐Ÿ‡ฆ๐Ÿ‡บ )
2012 Yokohama ( ๐Ÿ‡ฏ๐Ÿ‡ต ) olympic WTCS Lisa Norden ( ๐Ÿ‡ธ๐Ÿ‡ช ) Anne Haug ( ๐Ÿ‡ฉ๐Ÿ‡ช )
2012 London ( ๐Ÿ‡ฌ๐Ÿ‡ง ) olympic games Nicola Spirig ( ๐Ÿ‡จ๐Ÿ‡ญ ) Lisa Norden ( ๐Ÿ‡ธ๐Ÿ‡ช )
2015 Tiszaujvaros ( ๐Ÿ‡ญ๐Ÿ‡บ ) sprint world-cup Felicity Sheedy-Ryan ( ๐Ÿ‡ฆ๐Ÿ‡บ ) Audrey Merle ( ๐Ÿ‡ซ๐Ÿ‡ท )
2017 Cape Town ( ๐Ÿ‡ฟ๐Ÿ‡ฆ ) sprint world-cup Lucy Buckingham ( ๐Ÿ‡ฌ๐Ÿ‡ง ) Jessica Learmonth ( ๐Ÿ‡ฌ๐Ÿ‡ง )
2017 Abu Dhabi ( ๐Ÿ‡ฆ๐Ÿ‡ช ) olympic WTCS Andrea Hansen ( ๐Ÿ‡ณ๐Ÿ‡ฟ ) Jodie Stimpson ( ๐Ÿ‡ฌ๐Ÿ‡ง )
2017 New Plymouth ( ๐Ÿ‡ณ๐Ÿ‡ฟ ) sprint world-cup Katie Zaferes ( ๐Ÿ‡บ๐Ÿ‡ธ ) Joanna Brown ( ๐Ÿ‡จ๐Ÿ‡ฆ )
2018 Cagliari ( ๐Ÿ‡ฎ๐Ÿ‡น ) sprint world-cup Lisa Perterer ( ๐Ÿ‡ฆ๐Ÿ‡น ) Taylor Spivey ( ๐Ÿ‡บ๐Ÿ‡ธ )
2018 Karlovy Vary ( ๐Ÿ‡จ๐Ÿ‡ฟ ) olympic world-cup Vendula Frintova ( ๐Ÿ‡จ๐Ÿ‡ฟ ) Kaidi Kivioja ( ๐Ÿ‡ช๐Ÿ‡ช )
2019 Madrid ( ๐Ÿ‡ช๐Ÿ‡ธ ) sprint world-cup Emilie Morier ( ๐Ÿ‡ซ๐Ÿ‡ท ) Sandra Dodet ( ๐Ÿ‡ซ๐Ÿ‡ท )
2019 Tiszaujvaros ( ๐Ÿ‡ญ๐Ÿ‡บ ) sprint world-cup Emma Jeffcoat ( ๐Ÿ‡ฆ๐Ÿ‡บ ) Sara Vilic ( ๐Ÿ‡ฆ๐Ÿ‡น )
2022 Bergen ( ๐Ÿ‡ณ๐Ÿ‡ด ) sprint world-cup Tilda Mรฅnsson ( ๐Ÿ‡ธ๐Ÿ‡ช ) Jolien Vermeylen ( ๐Ÿ‡ง๐Ÿ‡ช )
2023 Tiszaujvaros ( ๐Ÿ‡ญ๐Ÿ‡บ ) sprint world-cup Tilda Mรฅnsson ( ๐Ÿ‡ธ๐Ÿ‡ช ) Noelia Juan ( ๐Ÿ‡ช๐Ÿ‡ธ )
2024 Huatulco ( ๐Ÿ‡ฒ๐Ÿ‡ฝ ) sprint world-cup Alberte Kjรฆr Pedersen ( ๐Ÿ‡ฉ๐Ÿ‡ฐ ) Rachel Klamer ( ๐Ÿ‡ณ๐Ÿ‡ฑ )
2024 Wollongong ( ๐Ÿ‡ฆ๐Ÿ‡บ ) sprint world-cup Tilda Mรฅnsson ( ๐Ÿ‡ธ๐Ÿ‡ช ) Maria Carolina Velasquez Soto ( ๐Ÿ‡จ๐Ÿ‡ด )
MEN
YEAR VENUE DIST. RACE CATEGORY FIRST ( ๐Ÿฅ‡ ) SECOND ( ๐Ÿฅˆ )
2009 Tongyeong ( ๐Ÿ‡ฐ๐Ÿ‡ท ) olympic WTCS Bevan Docherty ( ๐Ÿ‡ณ๐Ÿ‡ฟ ) Brad Kahlefeldt ( ๐Ÿ‡ฆ๐Ÿ‡บ )
2009 Des Moines ( ๐Ÿ‡บ๐Ÿ‡ธ ) olympic world-cup Simon Whitfield ( ๐Ÿ‡จ๐Ÿ‡ฆ ) Brad Kahlefeldt ( ๐Ÿ‡ฆ๐Ÿ‡บ )
2010 Seoul ( ๐Ÿ‡ฐ๐Ÿ‡ท ) olympic WTCS Jan Frodeno ( ๐Ÿ‡ฉ๐Ÿ‡ช ) Courtney Atkinson ( ๐Ÿ‡ฆ๐Ÿ‡บ )
2011 Hamburg ( ๐Ÿ‡ฉ๐Ÿ‡ช ) olympic WTCS Brad Kahlefeldt ( ๐Ÿ‡ฆ๐Ÿ‡บ ) William Clarke ( ๐Ÿ‡ฌ๐Ÿ‡ง )
2012 Mooloolaba ( ๐Ÿ‡ฆ๐Ÿ‡บ ) olympic world-cup Laurent Vidal ( ๐Ÿ‡ซ๐Ÿ‡ท ) Brad Kahlefeldt ( ๐Ÿ‡ฆ๐Ÿ‡บ )
2013 London ( ๐Ÿ‡ฌ๐Ÿ‡ง ) olympic WTCS Javier Gomez Noya ( ๐Ÿ‡ช๐Ÿ‡ธ ) Jonathan Brownlee ( ๐Ÿ‡ฌ๐Ÿ‡ง )
2013 Hamburg ( ๐Ÿ‡ฉ๐Ÿ‡ช ) sprint WTCS Jonathan Brownlee ( ๐Ÿ‡ฌ๐Ÿ‡ง ) Alistair Brownlee ( ๐Ÿ‡ฌ๐Ÿ‡ง )
2014 Chengdu ( ๐Ÿ‡จ๐Ÿ‡ณ ) olympic world-cup Wian Sullwald ( ๐Ÿ‡ฟ๐Ÿ‡ฆ ) Kevin McDowell ( ๐Ÿ‡บ๐Ÿ‡ธ )
2014 London ( ๐Ÿ‡ฌ๐Ÿ‡ง ) sprint WTCS Mario Mola ( ๐Ÿ‡ช๐Ÿ‡ธ ) Richard Murray ( ๐Ÿ‡ฟ๐Ÿ‡ฆ )
2014 Yokohama ( ๐Ÿ‡ฏ๐Ÿ‡ต ) olympic WTCS Javier Gomez Noya ( ๐Ÿ‡ช๐Ÿ‡ธ ) Mario Mola ( ๐Ÿ‡ช๐Ÿ‡ธ )
2014 Tiszaujvaros ( ๐Ÿ‡ญ๐Ÿ‡บ ) sprint world-cup รkos Vanek ( ๐Ÿ‡ญ๐Ÿ‡บ ) Rostislav Pevtsov ( ๐Ÿ‡ฆ๐Ÿ‡ฟ )
2015 Chengdu ( ๐Ÿ‡จ๐Ÿ‡ณ ) olympic world-cup Ryan Fisher ( ๐Ÿ‡ฆ๐Ÿ‡บ ) Rostislav Pevtsov ( ๐Ÿ‡ฆ๐Ÿ‡ฟ )
2016 Miyazaki ( ๐Ÿ‡ฏ๐Ÿ‡ต ) olympic world-cup Uxio Abuin Ares ( ๐Ÿ‡ช๐Ÿ‡ธ ) Joao Silva ( ๐Ÿ‡ต๐Ÿ‡น )
2016 Salinas ( ๐Ÿ‡ช๐Ÿ‡จ ) sprint world-cup David Castro Fajardo ( ๐Ÿ‡ช๐Ÿ‡ธ ) Matthew McElroy ( ๐Ÿ‡บ๐Ÿ‡ธ )
2016 Tongyeong ( ๐Ÿ‡ฐ๐Ÿ‡ท ) sprint world-cup Uxio Abuin Ares ( ๐Ÿ‡ช๐Ÿ‡ธ ) Matthew McElroy ( ๐Ÿ‡บ๐Ÿ‡ธ )
2018 Tongyeong ( ๐Ÿ‡ฐ๐Ÿ‡ท ) sprint world-cup Max Studer ( ๐Ÿ‡จ๐Ÿ‡ญ ) Felix Duchampt ( ๐Ÿ‡ท๐Ÿ‡ด )
2019 Cagliari ( ๐Ÿ‡ฎ๐Ÿ‡น ) sprint world-cup Alistair Brownlee ( ๐Ÿ‡ฌ๐Ÿ‡ง ) Justus Nieschlag ( ๐Ÿ‡ฉ๐Ÿ‡ช )
2019 Hamburg ( ๐Ÿ‡ฉ๐Ÿ‡ช ) sprint WTCS Jacob Birtwhistle ( ๐Ÿ‡ฆ๐Ÿ‡บ ) Vincent Luis ( ๐Ÿ‡ซ๐Ÿ‡ท )
2019 Montreal ( ๐Ÿ‡จ๐Ÿ‡ฆ ) sprint WTCS Jelle Geens ( ๐Ÿ‡ง๐Ÿ‡ช ) Mario Mola ( ๐Ÿ‡ช๐Ÿ‡ธ )
2019 Madrid ( ๐Ÿ‡ช๐Ÿ‡ธ ) sprint world-cup Justus Nieschlag ( ๐Ÿ‡ฉ๐Ÿ‡ช ) Lasse Lรผhrs ( ๐Ÿ‡ฉ๐Ÿ‡ช )
2019 Tiszaujvaros ( ๐Ÿ‡ญ๐Ÿ‡บ ) sprint world-cup Eli Hemming ( ๐Ÿ‡บ๐Ÿ‡ธ ) Ryan Fisher ( ๐Ÿ‡ฆ๐Ÿ‡บ )
2020 Mooloolaba ( ๐Ÿ‡ฆ๐Ÿ‡บ ) sprint world-cup Ryan Sissons ( ๐Ÿ‡ณ๐Ÿ‡ฟ ) Hayden Wilde ( ๐Ÿ‡ณ๐Ÿ‡ฟ )
2021 Edmonton ( ๐Ÿ‡จ๐Ÿ‡ฆ ) olympic WTCS Kristian Blummenfelt ( ๐Ÿ‡ณ๐Ÿ‡ด ) Marten Van Riel ( ๐Ÿ‡ง๐Ÿ‡ช )
2021 Hamburg ( ๐Ÿ‡ฉ๐Ÿ‡ช ) sprint WTCS Tim Hellwig ( ๐Ÿ‡ฉ๐Ÿ‡ช ) Paul Georgenthum ( ๐Ÿ‡ซ๐Ÿ‡ท )
2022 Bergen ( ๐Ÿ‡ณ๐Ÿ‡ด ) sprint world-cup Dorian Coninx ( ๐Ÿ‡ซ๐Ÿ‡ท ) Kristian Blummenfelt ( ๐Ÿ‡ณ๐Ÿ‡ด )
2022 Huatulco ( ๐Ÿ‡ฒ๐Ÿ‡ฝ ) sprint world-cup Genis Grau ( ๐Ÿ‡ช๐Ÿ‡ธ ) Tyler Mislawchuk ( ๐Ÿ‡จ๐Ÿ‡ฆ )
2023 Pontevedra ( ๐Ÿ‡ช๐Ÿ‡ธ ) olympic WTCS Dorian Coninx ( ๐Ÿ‡ซ๐Ÿ‡ท ) Tim Hellwig ( ๐Ÿ‡ฉ๐Ÿ‡ช )
2023 Sunderland ( ๐Ÿ‡ฌ๐Ÿ‡ง ) sprint WTCS Pierre Le Corre ( ๐Ÿ‡ซ๐Ÿ‡ท ) Lรฉo Bergere ( ๐Ÿ‡ซ๐Ÿ‡ท )
2023 Huatulco ( ๐Ÿ‡ฒ๐Ÿ‡ฝ ) sprint world-cup David Castro Fajardo ( ๐Ÿ‡ช๐Ÿ‡ธ ) Tyler Mislawchuk ( ๐Ÿ‡จ๐Ÿ‡ฆ )
2023 Valencia ( ๐Ÿ‡ช๐Ÿ‡ธ ) olympic world-cup David Cantero Del Campo ( ๐Ÿ‡ช๐Ÿ‡ธ ) Lasse Nygaard Priester ( ๐Ÿ‡ฉ๐Ÿ‡ช )

Not many highly contested finishes on women's world-series for a while!

By the way, World Triathlon rules that the win must be contested:

  • Triathletes should not "finish in a contrived tie situation where no effort to separate the finish times has been made".
  • At Tokyo ( ๐Ÿ‡ฏ๐Ÿ‡ต ) 2019, Jessica Learmonth ( ๐Ÿ‡ฌ๐Ÿ‡ง ) and Georgia Taylor-Brown ( ๐Ÿ‡ฌ๐Ÿ‡ง ) were disqualified after crossing line hand-in-hand.

๐Ÿ—“๏ธ LEVEL OVER YEARS

The same outlier removal is applied as described in the โฑ๏ธ PACES section.

repeated_events_standard_w.png
Olympic format. Women.
repeated_events_standard_m.png
Olympic format. Men.
Click to expand - ๐Ÿš€ Same plots for the Sprint format.
repeated_events_sprint_w.png
Sprint format. Women.
repeated_events_sprint_m.png
Sprint format. Men.
Click to expand - ๐Ÿ… Same plots for the Top-3.
repeated_events_standard_w_top3.png
Olympic format. Top-3 women in each leg.
repeated_events_standard_w_top3.png
Olympic format. Top-3 men in each leg.
repeated_events_sprint_w_top3.png
Sprint format. Top-3 women in each leg.
repeated_events_sprint_w_top3.png
Sprint format. Top-3 men in each leg.
Click to expand - ๐Ÿ Same plots for the Top-10.
repeated_events_standard_w_top10.png
Olympic format. Top-10 women in each leg.
repeated_events_standard_w_top10.png
Olympic format. Top-10 men in each leg.
repeated_events_sprint_w_top10.png
Sprint format. Top-10 women in each leg.
repeated_events_sprint_w_top10.png
Sprint format. Top-10 men in each leg.

In this early-2024 video, Vincent Luis ( ๐Ÿ‡ซ๐Ÿ‡ท ) explains:

"My running times are very similar to what I was doing in 2019-2020, when I was world champion. [...] It's just that the level has gone up [...]. I ran at 20km/h in Yokohama (2024). This is a pace that 5-6 years ago was enough to win some World Series."

The above plots seem consistent with this statement. ๐Ÿ“ˆ

  • The figures for women's and men's olympic-format show recent improvements of the 10k run pace (green bars) on WTCS races. For 2019 -> 2021 -> 2023, paces were:
    • 3:33 -> 3:28 -> 3:23 for women.
    • 3:06 -> 3:04 -> 2:59 for men.
  • The 3:00 /km pace mention by Luis was, in 2023, hardly enough to finish between 5-th and 9-th on WTCS men's races.
  • The same 3:00 /km was, in 2019, probably enough to win a WTCS, since the 5-th to 9-th places were about 3:06 /km at that time.

In this other 2024 video:

  • Alex Yee ( ๐Ÿ‡ฌ๐Ÿ‡ง ) mentions that the run of 2012 London ( ๐Ÿ‡ฌ๐Ÿ‡ง ) Olympics is a reference: "The run was held as the best run that has ever been done in triathlon".
  • He explains: "In the last season (2023), we had a few races which came very close to that."
  • This statement seems also to be correct: on the run subplot of the men's olympic format figure, London 2012 was the fastest until 2023.
  • This article from triathlon.org, released just before Paris ( ๐Ÿ‡ซ๐Ÿ‡ท ) 2024, confirms: "Brownleeโ€™s times (London ( ๐Ÿ‡ฌ๐Ÿ‡ง ) 2012) will likely come under threat. Indeed, it seems highly likely that we could see the first ever sub-29 and sub-33 minute 10km times in an Olympic triathlon this summer."
    • The 5-9th men ran at the 2024 Paris Olympics much slower that the year before for the test event.
    • Also, Alex Yee ( ๐Ÿ‡ฌ๐Ÿ‡ง ) won in 2024 with a 29:49 run, compared to 29:00 in 2023.
    • Because of the heat (the men's race started at 10:45 am instead of 8:00 am)?
    • Or the fatigue caused by the very long swim?
    • Or was the run course longer in 2024? In this case the women's run time improvement would be even more impressive: Beth Potter ( ๐Ÿ‡ฌ๐Ÿ‡ง ) did 32:57 in 2023, compared to 32:42 for Cassandre Beaugrand ( ๐Ÿ‡ซ๐Ÿ‡ท ) in 2024, who run faster than 30% of men finishers (15 / 50)!

Yokohama ๐Ÿ‡ฏ๐Ÿ‡ต

  • The event took place 13 times between 2009 and 2024 in an olympic format.
    • I do not know if the courses have changed.
    • But if not, the comparison should be very relevant.
  • Run times have decreased in the last five editions. ๐Ÿ‘Ÿ
    • Leading to the best ever times in 2024.
  • It is interesting to note that runs were particularly good on olympics years (2012, 2016, 2021, 2024), especially for women.
    • Can it be that this race (usually happening early in the season) was used as qualification criteria or as a rehearsal, and therefore attracting very fit and motivated athletes?

Hamburg ๐Ÿ‡ฉ๐Ÿ‡ช

  • The event took place 12 times between 2009 and 2024 in a sprint format.
  • Run times have been constant until 2024, where a clear improvement can be seen.
    • The 2024 race was used by many athletes as a final repetition before the Olympics. But still, the improvement is huge.
    • Can it be that the 2024 run course was shorter? Indeed, the 2024 result page indicates Run 4.905km (2 laps) for 2024, compared to Run 5km (2 laps) for 2022.

Impact of carbon plate running shoes? ๐Ÿ‘Ÿ

  • The Nike Vaporfly came out in 2017.
  • As for Yokohama ( ๐Ÿ‡ฏ๐Ÿ‡ต ), the running paces have been constantly improving on WTCS since 2017-2018.
    • Running times were already good before 2014.
    • But since 2021, paces have never been so low.
  • Carbon plate technology could be one of the main factors explaining this improvement.
    • But how to explain that running performances on world-cups have not followed the same trend?

The bike(s) ๐Ÿšฒ

  • On olympic format, bike times have been drastically improving (ignoring covid 2020 year) over the past 6 years, especially for the men.
  • Can it be due to tech innovations?
  • Maybe the level has gone up: it is no longer just about being a good swimmer and an excellent runner?
  • Maybe some athletes tend to take more risks on the bike, sometimes reckless as reports Vincent Luis ( ๐Ÿ‡ซ๐Ÿ‡ท ) in this interview.

World-cup vs WTCS ๐Ÿ†

  • Apart for 2018, the running level is consistently higher in WTCS than in world-cups.
  • Which could be expected since WTCS are so much selective.

The swim of 2024 Paris ( ๐Ÿ‡ซ๐Ÿ‡ท ) Olympics was very though because of the current on the way back:

  • 20:26 (01:22 /100m) for men.
  • 22:33 (01:30 /100m) for women.

Care is needed when comparing swim times:

  • In the WTCS (green bars), women's times appear to have reached historic lows, while 2024 men's times are the third slowest since 2009.
  • This is likely because, in the two 2024 olympic WTCS events considered (Yokohama ( ๐Ÿ‡ฏ๐Ÿ‡ต ) and Cagliari ( ๐Ÿ‡ฎ๐Ÿ‡น )), women swam with wetsuits while men did not.

Criticisms โš ๏ธ

  • Not all events have identical distances and conditions.
    • However, averaging many events (the number below the year on the x-axis), with multiple venues repeating every year, helps to mitigate this variability.
  • Swim ๐ŸŠ
    • Distances can vary based on buoy positions.
    • Differences in water conditions (e.g. rough sea, current) can also significantly affect swim times, making comparisons challenging.
    • Last but not least, the wetsuit may be allowed or not.
  • Bike ๐Ÿšด
    • Weather conditions, particularly wind and rain, can influence bike times.
    • Variations in course profiles (hilly vs. flat) can make direct comparisons of bike times unfair.
  • Run ๐Ÿƒ
    • Run times are generally more comparable as World Triathlon run courses tend to be predominantly flat, reducing variability.

๐ŸŒก๏ธ TEMPERATURES

This section examines the recorded water and air temperatures and, inspired by the recent work by Gibson (2024), investigates their impact on swimming and running performance, respectively.

temperatures.png
Recorded water and air temperatures.

The temperature ranges are broad:

  • ๐ŸŒŠ 80% of the recorded water temperatures are between 16.4 and 26.9 ยฐC.
    • (Mean: 21.5 ยฐC, SD: 3.9 ยฐC, Min: 13.8 ยฐC, Max: 31.0 ยฐC)
  • โ›ฑ๏ธ 80% of the recorded air temperatures are between 17.0 and 29.6 ยฐC.
    • (Mean: 23.3 ยฐC, SD: 4.9 ยฐC, Min: 11.0 ยฐC, Max: 34.0 ยฐC)

Several limitations should be considered:

  • Incomplete data: Over half of the events lack temperature data, limiting the analysis's comprehensiveness.
  • Pre-race measurements: Measurements are typically taken before the event.
    • Depending on the race timetable, the actual temperature experienced during the race can be lower or higher than the reported temperature.
    • "Water temperature must be taken one hour prior to the start of the event on competition day. It must be taken at the middle of the course and in two other areas on the swim course, at a depth of 60 cm."
    • Air temperatures are likely recorded before the race as well, potentially leading to significant temperature increases by the time athletes begin the run.
    • For instance, the air temperature for the men's race (10:45 am) at Paris 2024 ( ๐Ÿ‡ซ๐Ÿ‡ท ) is reported at 23.9 ยฐC, while it was closer to 28 ยฐC during the run.
  • Humidity Impact: While humidity could significantly affect running performance, its values are not reported.
  • Last but not least, swim and run courses vary in distance, and water conditions (e.g., waves, salinity) also differ between events, making comparisons challenging.

โš ๏ธ Given these limitations, especially the variation in course distances, conclusions should be drawn with caution.

๐ŸŒŠ Water temperatures and swim times

temperatures_water.png
Water temperature and swim times.

The women's race at Haeundae ( ๐Ÿ‡ฐ๐Ÿ‡ท ) (2021) is excluded due to an inconsistency with the 20ยฐC wetsuit rule:

  • The race report notes: "Water temperature 21.3ยบC. Air temperature 15.4ยบ C. Wetsuits allowed."
  • The 2024 rule book states in section 4.4.b. that "when the water temperature is at or below 22ยบC and the air temperature is at or below 15ยบC, then the value of the water temperature will be adjusted."
    • For instance: Air at 15ยฐC and water at 22ยฐC -> The water temperature is adjusted at 18.5ยฐC -> Wetsuit allowed.
    • It is possible that the 15ยฐC threshold was higher in 2021, leading to the discrepancy.

Swim appears slightly faster in water temperatures below 20ยฐC, which is particularly visible in the two "olympic" sub-plots.

  • Likely because wetsuits are permitted at these lower temperatures.
  • Further research could analyse the impact of temperature on swim performance (see already Gay et al, 2021), and try to determine the optimal water temperature range for races, with and without wetsuits.

๐ŸŠ Some cold and hot swims:

YEAR EVENT WATER TEMPERATURE DISTANCE EVENT CATEGORY
2017 Cape Town ( ๐Ÿ‡ฟ๐Ÿ‡ฆ ) 13.8 ๐Ÿฅถ sprint WORLD-CUP
2022 Vina del Mar ( ๐Ÿ‡จ๐Ÿ‡ฑ ) 14.3 ๐Ÿฅถ sprint WORLD-CUP
2018 Cape Town ( ๐Ÿ‡ฟ๐Ÿ‡ฆ ) 15.0 ๐Ÿฅถ sprint WORLD-CUP
2022 Bergen ( ๐Ÿ‡ณ๐Ÿ‡ด ) 15.0 ๐Ÿฅถ sprint WORLD-CUP
2014 Stockholm ( ๐Ÿ‡ธ๐Ÿ‡ช ) 15.0 ๐Ÿฅถ sprint WTCS
2015 London ( ๐Ÿ‡ฌ๐Ÿ‡ง ) 15.2 ๐Ÿฅถ sprint WTCS
2019 Cagliari ( ๐Ÿ‡ฎ๐Ÿ‡น ) 15.5 ๐Ÿฅถ sprint WORLD-CUP
2016 Cape Town ( ๐Ÿ‡ฟ๐Ÿ‡ฆ ) 15.6 ๐Ÿฅถ sprint WTCS
2019 Cape Town ( ๐Ÿ‡ฟ๐Ÿ‡ฆ ) 16.0 ๐Ÿฅถ sprint WORLD-CUP
2017 Rotterdam ( ๐Ÿ‡ณ๐Ÿ‡ฑ ) 16.1 ๐Ÿฅถ olympic WTCS
... ... ... ... ...
2022 Valencia ( ๐Ÿ‡ช๐Ÿ‡ธ ) 27.8 ๐Ÿฅต sprint WORLD-CUP
2021 Tokyo ( ๐Ÿ‡ฏ๐Ÿ‡ต ) 28.9 ๐Ÿฅต olympic GAMES
2016 Cozumel ( ๐Ÿ‡ฒ๐Ÿ‡ฝ ) 29.0 ๐Ÿฅต olympic WTCS
2017 Mรฉrida ( ๐Ÿ‡ฒ๐Ÿ‡ฝ ) 29.0 ๐Ÿฅต sprint WORLD-CUP
2018 Mooloolaba ( ๐Ÿ‡ฆ๐Ÿ‡บ ) 29.1 ๐Ÿฅต sprint WORLD-CUP
2019 Huatulco ( ๐Ÿ‡ฒ๐Ÿ‡ฝ ) 29.5 ๐Ÿฅต sprint WORLD-CUP
2018 Huatulco ( ๐Ÿ‡ฒ๐Ÿ‡ฝ ) 30.0 ๐Ÿฅต sprint WORLD-CUP
2021 Huatulco ( ๐Ÿ‡ฒ๐Ÿ‡ฝ ) 30.0 ๐Ÿฅต sprint WORLD-CUP
2022 Huatulco ( ๐Ÿ‡ฒ๐Ÿ‡ฝ ) 31.0 ๐Ÿฅต sprint WORLD-CUP
2021 Abu Dhabi ( ๐Ÿ‡ฆ๐Ÿ‡ช ) 31.0 ๐Ÿฅต sprint WTCS

โ›ฑ๏ธ Air temperatures and run times

temperatures_air.png
Air temperature and run times.

A 2nd degree fit is applied on the scatter plot using seaborn.regplot.

  • It suggests a trend where higher temperatures correlate with slower running paces.
  • While no definitive "optimal" temperature can be reliably determined, this trend aligns with both research findings and personal experiences ( ๐Ÿฅต ).
  • It also underscores the importance of hydration and cooling strategies, such as the use of cooling headbands.

๐Ÿƒ Some cold and hot runs:

YEAR EVENT AIR TEMPERATURE DISTANCE EVENT CATEGORY
2022 Vina del Mar ( ๐Ÿ‡จ๐Ÿ‡ฑ ) 11.0 ๐Ÿฅถ sprint WORLD-CUP
2016 Edmonton ( ๐Ÿ‡จ๐Ÿ‡ฆ ) 12.7 ๐Ÿฅถ sprint WTCS
2021 Tongyeong ( ๐Ÿ‡ฐ๐Ÿ‡ท ) 13.2 ๐Ÿฅถ sprint WORLD-CUP
2018 New Plymouth ( ๐Ÿ‡ณ๐Ÿ‡ฟ ) 13.7 ๐Ÿฅถ sprint WORLD-CUP
2017 Rotterdam ( ๐Ÿ‡ณ๐Ÿ‡ฑ ) 14.4 ๐Ÿฅถ olympic WTCS
2020 Hamburg ( ๐Ÿ‡ฉ๐Ÿ‡ช ) 14.9 ๐Ÿฅถ sprint WTCS
2021 Lisbon ( ๐Ÿ‡ต๐Ÿ‡น ) 15.0 ๐Ÿฅถ olympic WORLD-CUP
2018 Miyazaki ( ๐Ÿ‡ฏ๐Ÿ‡ต ) 15.0 ๐Ÿฅถ olympic WORLD-CUP
2015 London ( ๐Ÿ‡ฌ๐Ÿ‡ง ) 15.2 ๐Ÿฅถ sprint WTCS
2021 Haeundae ( ๐Ÿ‡ฐ๐Ÿ‡ท ) 15.4 ๐Ÿฅถ sprint WORLD-CUP
... ... ... ... ...
2019 Montreal ( ๐Ÿ‡จ๐Ÿ‡ฆ ) 30.9 ๐Ÿฅต sprint WTCS
2019 Antwerp MAS ( ๐Ÿ‡ง๐Ÿ‡ช ) 31.5 ๐Ÿฅต sprint WORLD-CUP
2016 Mooloolaba ( ๐Ÿ‡ฆ๐Ÿ‡บ ) 31.8 ๐Ÿฅต sprint WORLD-CUP
2017 Cagliari ( ๐Ÿ‡ฎ๐Ÿ‡น ) 32.0 ๐Ÿฅต sprint WORLD-CUP
2022 Arzachena ( ๐Ÿ‡ฎ๐Ÿ‡น ) 32.0 ๐Ÿฅต sprint WORLD-CUP
2018 Huatulco ( ๐Ÿ‡ฒ๐Ÿ‡ฝ ) 32.1 ๐Ÿฅต sprint WORLD-CUP
2015 Rio de Janeiro ( ๐Ÿ‡ง๐Ÿ‡ท ) 32.1 ๐Ÿฅต olympic GAMES
2014 Tiszaujvaros ( ๐Ÿ‡ญ๐Ÿ‡บ ) 32.4 ๐Ÿฅต sprint WORLD-CUP
2021 Abu Dhabi ( ๐Ÿ‡ฆ๐Ÿ‡ช ) 34.0 ๐Ÿฅต sprint WTCS
2016 Cozumel ( ๐Ÿ‡ฒ๐Ÿ‡ฝ ) 34.0 ๐Ÿฅต olympic WTCS

๐ŸŒ HOST COUNTRIES

Countries having hosted more than one world-series, world-cup or games-related event since 2009:

COUNTRY COUNT VENUES
JPN ( ๐Ÿ‡ฏ๐Ÿ‡ต ) 21 Yokohama (13), Miyazaki (6), Ishigaki (1), Tokyo (1)
AUS ( ๐Ÿ‡ฆ๐Ÿ‡บ ) 18 Mooloolaba (9), Gold Coast (5), Sydney (3), Wollongong (1)
ESP ( ๐Ÿ‡ช๐Ÿ‡ธ ) 17 Madrid (7), Valencia (3), Banyoles (2), Pontevedra (2), Palamos (1), Alicante (1), Huelva (1)
MEX ( ๐Ÿ‡ฒ๐Ÿ‡ฝ ) 16 Huatulco (8), Cozumel (4), Monterrey (2), Cancun (1), Mรฉrida (1)
KOR ( ๐Ÿ‡ฐ๐Ÿ‡ท ) 14 Tongyeong (9), Tongyeong (2), Seoul (1), Haeundae (1), Yeongdo (1)
GER ( ๐Ÿ‡ฉ๐Ÿ‡ช ) 14 Hamburg (14)
GBR ( ๐Ÿ‡ฌ๐Ÿ‡ง ) 14 London (7), Leeds (5), Birmingham (1), Sunderland (1)
CAN ( ๐Ÿ‡จ๐Ÿ‡ฆ ) 13 Edmonton (9), Montreal (4)
NZL ( ๐Ÿ‡ณ๐Ÿ‡ฟ ) 11 New Plymouth (6), Auckland (4), Napier (1)
ITA ( ๐Ÿ‡ฎ๐Ÿ‡น ) 11 Cagliari (7), Arzachena (3), Rome (1)
HUN ( ๐Ÿ‡ญ๐Ÿ‡บ ) 10 Tiszaujvaros (9)
CHN ( ๐Ÿ‡จ๐Ÿ‡ณ ) 10 Chengdu (5), Weihai (2), Beijing (1), Jiayuguan (1), Weihai (1)
UAE ( ๐Ÿ‡ฆ๐Ÿ‡ช ) 8 Abu Dhabi (8)
CZE ( ๐Ÿ‡จ๐Ÿ‡ฟ ) 6 Karlovy Vary (6)
USA ( ๐Ÿ‡บ๐Ÿ‡ธ ) 6 Des Moines (2), San Diego (2), Chicago (2)
SWE ( ๐Ÿ‡ธ๐Ÿ‡ช ) 5 Stockholm (5)
AUT ( ๐Ÿ‡ฆ๐Ÿ‡น ) 4 Kitzbuhel (4)
RSA ( ๐Ÿ‡ฟ๐Ÿ‡ฆ ) 4 Cape Town (4)
BER ( ๐Ÿ‡ง๐Ÿ‡ฒ ) 3 Bermuda (3)
BRA ( ๐Ÿ‡ง๐Ÿ‡ท ) 3 Rio de Janeiro (2)
ECU ( ๐Ÿ‡ช๐Ÿ‡จ ) 3 Salinas (3)
SUI ( ๐Ÿ‡จ๐Ÿ‡ญ ) 3 Lausanne (3)
CHI ( ๐Ÿ‡จ๐Ÿ‡ฑ ) 2 Vina del Mar (2)
TUR ( ๐Ÿ‡น๐Ÿ‡ท ) 2 Alanya (2)

It can be that some events are missing: these entries come from the data used for this report, after filtering and cleaning.

Click to expand - Top host countries for world-series.
COUNTRY COUNT VENUES
GER ( ๐Ÿ‡ฉ๐Ÿ‡ช ) 14 Hamburg (14)
JPN ( ๐Ÿ‡ฏ๐Ÿ‡ต ) 13 Yokohama (13)
GBR ( ๐Ÿ‡ฌ๐Ÿ‡ง ) 12 London (6), Leeds (5), Sunderland (1)
CAN ( ๐Ÿ‡จ๐Ÿ‡ฆ ) 10 Edmonton (7), Montreal (3)
AUS ( ๐Ÿ‡ฆ๐Ÿ‡บ ) 8 Gold Coast (5), Sydney (3)
UAE ( ๐Ÿ‡ฆ๐Ÿ‡ช ) 8 Abu Dhabi (8)
ESP ( ๐Ÿ‡ช๐Ÿ‡ธ ) 6 Madrid (5), Pontevedra (1)
SWE ( ๐Ÿ‡ธ๐Ÿ‡ช ) 5 Stockholm (5)
USA ( ๐Ÿ‡บ๐Ÿ‡ธ ) 4 San Diego (2), Chicago (2)
AUT ( ๐Ÿ‡ฆ๐Ÿ‡น ) 4 Kitzbuhel (4)
Click to expand - Top host countries for world-cups.
COUNTRY COUNT VENUES
MEX ( ๐Ÿ‡ฒ๐Ÿ‡ฝ ) 15 Huatulco (8), Cozumel (3), Monterrey (2), Cancun (1), Mรฉrida (1)
KOR ( ๐Ÿ‡ฐ๐Ÿ‡ท ) 12 Tongyeong (8), Tongyeong (2), Haeundae (1), Yeongdo (1)
ESP ( ๐Ÿ‡ช๐Ÿ‡ธ ) 11 Valencia (3), Banyoles (2), Madrid (2), Palamos (1), Alicante (1), Huelva (1), Pontevedra (1)
AUS ( ๐Ÿ‡ฆ๐Ÿ‡บ ) 10 Mooloolaba (9), Wollongong (1)
CHN ( ๐Ÿ‡จ๐Ÿ‡ณ ) 9 Jintang (3), Weihai (2), Chengdu (2), Gansu (1), Weihai, Shandong (1)
HUN ( ๐Ÿ‡ญ๐Ÿ‡บ ) 9 Tiszaujvaros (7), Tiszaujvaros (2)
ITA ( ๐Ÿ‡ฎ๐Ÿ‡น ) 8 Cagliari (4), Arzachena (3), Rome (1)
NZL ( ๐Ÿ‡ณ๐Ÿ‡ฟ ) 8 New Plymouth (6), Auckland (1), Napier (1)
JPN ( ๐Ÿ‡ฏ๐Ÿ‡ต ) 7 Miyazaki (6), Ishigaki (1)
CZE ( ๐Ÿ‡จ๐Ÿ‡ฟ ) 6 Karlovy Vary (6)

๐Ÿ“† SEASON DURATION

This section examines the duration of the competition season and the number of races athletes participate in.

season_duration.png
Duration of World Cup and World Series seasons, as well as the duration of the seasons of the top 50 athletes in the WTCS ranking.

World-Series

2020, 2021 and to some extent 2022 have been affected by the covid pandemic. Their rows are written in italic.

Year Num. events Season duration Start End First event Last event
2009 7 127 days (~ 4.2 m) 05-02 09-09 Tongyeong ( ๐Ÿ‡ฐ๐Ÿ‡ท ) Gold Coast ( ๐Ÿ‡ฆ๐Ÿ‡บ )
2010 7 147 days (~ 4.9 m) 04-11 09-08 Sydney ( ๐Ÿ‡ฆ๐Ÿ‡บ ) Budapest ( ๐Ÿ‡ญ๐Ÿ‡บ )
2011 8 160 days (~ 5.3 m) 04-09 09-19 Sydney ( ๐Ÿ‡ฆ๐Ÿ‡บ ) Yokohama ( ๐Ÿ‡ฏ๐Ÿ‡ต )
2012 8 186 days (~ 6.2 m) 04-14 10-20 Sydney ( ๐Ÿ‡ฆ๐Ÿ‡บ ) Auckland ( ๐Ÿ‡ณ๐Ÿ‡ฟ )
2013 7 155 days (~ 5.2 m) 04-06 09-11 Auckland ( ๐Ÿ‡ณ๐Ÿ‡ฟ ) London ( ๐Ÿ‡ฌ๐Ÿ‡ง )
2014 7 143 days (~ 4.8 m) 04-06 08-29 Auckland ( ๐Ÿ‡ณ๐Ÿ‡ฟ ) Edmonton ( ๐Ÿ‡จ๐Ÿ‡ฆ )
2015 9 189 days (~ 6.3 m) 03-06 09-15 Abu Dhabi ( ๐Ÿ‡ฆ๐Ÿ‡ช ) Chicago ( ๐Ÿ‡บ๐Ÿ‡ธ )
2016 9 186 days (~ 6.2 m) 03-05 09-11 Abu Dhabi ( ๐Ÿ‡ฆ๐Ÿ‡ช ) Cozumel ( ๐Ÿ‡ฒ๐Ÿ‡ฝ )
2017 9 191 days (~ 6.4 m) 03-03 09-14 Abu Dhabi ( ๐Ÿ‡ฆ๐Ÿ‡ช ) Rotterdam ( ๐Ÿ‡ณ๐Ÿ‡ฑ )
2018 8 190 days (~ 6.3 m) 03-02 09-12 Abu Dhabi ( ๐Ÿ‡ฆ๐Ÿ‡ช ) Gold Coast ( ๐Ÿ‡ฆ๐Ÿ‡บ )
2019 8 171 days (~ 5.7 m) 03-08 08-29 Abu Dhabi ( ๐Ÿ‡ฆ๐Ÿ‡ช ) Lausanne ( ๐Ÿ‡จ๐Ÿ‡ญ )
2020 1 0 days ( ๐Ÿ˜ท ๐Ÿค’ ) 09-05 09-05 Hamburg ( ๐Ÿ‡ฉ๐Ÿ‡ช ) Hamburg ( ๐Ÿ‡ฉ๐Ÿ‡ช )
2021 5 170 days (~ 5.7 m) 05-15 11-05 Yokohama ( ๐Ÿ‡ฏ๐Ÿ‡ต ) Abu Dhabi ( ๐Ÿ‡ฆ๐Ÿ‡ช )
2022 6 190 days (~ 6.3 m) 05-14 11-24 Yokohama ( ๐Ÿ‡ฏ๐Ÿ‡ต ) Abu Dhabi ( ๐Ÿ‡ฆ๐Ÿ‡ช )
2023 6 199 days (~ 6.6 m) 03-03 09-22 Abu Dhabi ( ๐Ÿ‡ฆ๐Ÿ‡ช ) Pontevedra ( ๐Ÿ‡ช๐Ÿ‡ธ )
2024 2 14 days (~ 0.5 m) 05-11 05-25 Yokohama ( ๐Ÿ‡ฏ๐Ÿ‡ต ) Cagliari ( ๐Ÿ‡ฎ๐Ÿ‡น )

World-Cups

Year Num. events Season duration Start End First event Last event
2009 3 131 days (~ 4.4 m) 06-27 11-08 Des Moines ( ๐Ÿ‡บ๐Ÿ‡ธ ) Huatulco ( ๐Ÿ‡ฒ๐Ÿ‡ฝ )
2010 5 193 days (~ 6.4 m) 03-27 10-10 Mooloolaba ( ๐Ÿ‡ฆ๐Ÿ‡บ ) Huatulco ( ๐Ÿ‡ฒ๐Ÿ‡ฝ )
2011 8 234 days (~ 7.8 m) 03-26 11-20 Mooloolaba ( ๐Ÿ‡ฆ๐Ÿ‡บ ) Auckland ( ๐Ÿ‡ณ๐Ÿ‡ฟ )
2012 7 193 days (~ 6.4 m) 03-24 10-07 Mooloolaba ( ๐Ÿ‡ฆ๐Ÿ‡บ ) Cancun ( ๐Ÿ‡ฒ๐Ÿ‡ฝ )
2013 10 221 days (~ 7.4 m) 03-16 10-27 Mooloolaba ( ๐Ÿ‡ฆ๐Ÿ‡บ ) Guatape ( ๐Ÿ‡จ๐Ÿ‡ด )
2014 9 213 days (~ 7.1 m) 03-15 10-18 Mooloolaba ( ๐Ÿ‡ฆ๐Ÿ‡บ ) Tongyeong ( ๐Ÿ‡ฐ๐Ÿ‡ท )
2015 7 220 days (~ 7.3 m) 03-14 10-24 Mooloolaba ( ๐Ÿ‡ฆ๐Ÿ‡บ ) Tongyeong ( ๐Ÿ‡ฐ๐Ÿ‡ท )
2016 10 227 days (~ 7.6 m) 03-12 10-29 Mooloolaba ( ๐Ÿ‡ฆ๐Ÿ‡บ ) Miyazaki ( ๐Ÿ‡ฏ๐Ÿ‡ต )
2017 12 263 days (~ 8.8 m) 02-11 11-04 Cape Town ( ๐Ÿ‡ฟ๐Ÿ‡ฆ ) Miyazaki ( ๐Ÿ‡ฏ๐Ÿ‡ต )
2018 11 269 days (~ 9.0 m) 02-11 11-10 Cape Town ( ๐Ÿ‡ฟ๐Ÿ‡ฆ ) Miyazaki ( ๐Ÿ‡ฏ๐Ÿ‡ต )
2019 13 257 days (~ 8.6 m) 02-09 10-26 Cape Town ( ๐Ÿ‡ฟ๐Ÿ‡ฆ ) Miyazaki ( ๐Ÿ‡ฏ๐Ÿ‡ต )
2020 4 233 days (~ 7.8 m) 03-14 11-07 Mooloolaba ( ๐Ÿ‡ฆ๐Ÿ‡บ ) Valencia ( ๐Ÿ‡ช๐Ÿ‡ธ )
2021 6 158 days (~ 5.3 m) 05-22 10-30 Lisbon ( ๐Ÿ‡ต๐Ÿ‡น ) Tongyeong ( ๐Ÿ‡ฐ๐Ÿ‡ท )
2022 9 165 days (~ 5.5 m) 05-28 11-13 Arzachena ( ๐Ÿ‡ฎ๐Ÿ‡น ) Vina del Mar ( ๐Ÿ‡จ๐Ÿ‡ฑ )
2023 14 226 days (~ 7.5 m) 03-25 11-11 New Plymouth ( ๐Ÿ‡ณ๐Ÿ‡ฟ ) Vina del Mar ( ๐Ÿ‡จ๐Ÿ‡ฑ )
2024 6 84 days (~ 2.8 m) 02-24 05-18 Napier ( ๐Ÿ‡ณ๐Ÿ‡ฟ ) Samarkand ( ๐Ÿ‡บ๐Ÿ‡ฟ )

Athlete seasons

The following plots represent the seasons (columns) of 50 athletes (rows).

athlete_season_duration_m.png
Seasons of the top-50 athletes. Men.
athlete_season_duration_w.png
Seasons of the top-50 athletes. Women.

Top athletes seem to prefer world series over world cups (more red than blue in the top rows).

  • Probably because WTCS events offer more points and prize money.
  • Some athletes do not race in any world cup events (light blue background).

Regarding the competition season: from 2009 to 2019, it became longer.

  • The 2009 WTCS season started in May, but from 2015-2019, it began in early March.
  • In 2023, the WTCS season was 6.5 months long, compared to 4.2 months in 2009.
  • The number of world cup events increased significantly during this period: from 3 in 2009 to 13 in 2019.
    • The 2018 world cup season lasted 9 months.
  • The COVID-19 pandemic halted this trend, although 2023 seems similar to 2019.
    • The ranking process became complicated: e.g., some 2021 races were considered for the 2022 ranking.

Regarding the athletes' season: it follows the competition season.

  • Their World Triathlon season extended from 130 days in 2009 to 200 days in 2023.
  • On average, athletes raced 10 times in 2019 and 2023, compared to 6 times in 2009.

The results found (number of races and season duration) are lower bounds:

  • Athletes participate in other formats besides World Triathlon olympic and sprint races.
    • For instance, the French Grand Prix was very popular in the 2010s.
    • Some athletes are racing the supertri and Ironman 70.3 formats as well.
  • Their competition season is probably longer, including indoor races in the winter.

๐ŸŒ ATHLETE NATIONS

This table examines the nationalities of athletes in the top-50s of the WTCS Ranking from 2009 to 2023:

RANGE (%) NATIONS (W) NATIONS (M)
11-12 ๐Ÿ‡บ๐Ÿ‡ธ (11.0%)
10-11 ๐Ÿ‡ฌ๐Ÿ‡ง (10.9%) ๐Ÿ‡ฆ๐Ÿ‡บ (10.1%)
9-10 ๐Ÿ‡ซ๐Ÿ‡ท (9.6%)
8-9 ๐Ÿ‡ฏ๐Ÿ‡ต (8.0%) ๐Ÿ‡ฌ๐Ÿ‡ง (8.4%) ๐Ÿ‡ฉ๐Ÿ‡ช (8.1%)
7-8 ๐Ÿ‡ฉ๐Ÿ‡ช (7.6%) ๐Ÿ‡ฆ๐Ÿ‡บ (7.7%) ๐Ÿ‡ช๐Ÿ‡ธ (7.0%)
6-7 ๐Ÿ‡บ๐Ÿ‡ธ (6.1%)
5-6 ๐Ÿ‡ณ๐Ÿ‡ฟ (5.7%) ๐Ÿ‡ซ๐Ÿ‡ท (5.0%)
4-5 ๐Ÿ‡ณ๐Ÿ‡ฟ (4.6%) ๐Ÿ‡จ๐Ÿ‡ญ (4.6%) ๐Ÿ‡จ๐Ÿ‡ฆ (4.4%) ๐Ÿ‡ต๐Ÿ‡น (4.3%)
3-4 ๐Ÿ‡จ๐Ÿ‡ฆ (3.7%) ๐Ÿ‡ช๐Ÿ‡ธ (3.7%) ๐Ÿ‡ฎ๐Ÿ‡น (3.7%) ๐Ÿ‡ณ๐Ÿ‡ฑ (3.6%) ๐Ÿ‡จ๐Ÿ‡ญ (3.3%) ๐Ÿ‡ท๐Ÿ‡บ (3.6%) ๐Ÿ‡ฟ๐Ÿ‡ฆ (3.4%)
2-3 ๐Ÿ‡ฆ๐Ÿ‡น (2.6%) ๐Ÿ‡ฒ๐Ÿ‡ฝ (2.1%) ๐Ÿ‡ง๐Ÿ‡ท (2.0%) ๐Ÿ‡ฏ๐Ÿ‡ต (2.9%) ๐Ÿ‡ง๐Ÿ‡ช (2.9%) ๐Ÿ‡ญ๐Ÿ‡บ (2.6%) ๐Ÿ‡ฎ๐Ÿ‡น (2.4%) ๐Ÿ‡ณ๐Ÿ‡ด (2.4%) ๐Ÿ‡ฒ๐Ÿ‡ฝ (2.3%)
1-2 ๐Ÿ‡ง๐Ÿ‡ช (1.9%) ๐Ÿ‡จ๐Ÿ‡ฟ (1.7%) ๐Ÿ‡ฟ๐Ÿ‡ฆ (1.7%) ๐Ÿ‡ท๐Ÿ‡บ (1.6%) ๐Ÿ‡ต๐Ÿ‡ฑ (1.3%) ๐Ÿ‡ญ๐Ÿ‡บ (1.3%) ๐Ÿ‡ง๐Ÿ‡ฒ (1.3%) ๐Ÿ‡จ๐Ÿ‡ฑ (1.1%) ๐Ÿ‡ฎ๐Ÿ‡ช (1.0%) ๐Ÿ‡ง๐Ÿ‡ท (1.7%) ๐Ÿณ๏ธ (1.4%) ๐Ÿ‡ฆ๐Ÿ‡น (1.1%) ๐Ÿ‡ธ๐Ÿ‡ฐ (1.1%) ๐Ÿ‡ฎ๐Ÿ‡ช (1.0%) ๐Ÿ‡ฎ๐Ÿ‡ฑ (1.0%)
0-1 ๐Ÿ‡ต๐Ÿ‡น (0.9%) ๐Ÿ‡ธ๐Ÿ‡ช (0.7%) ๐Ÿ‡ฉ๐Ÿ‡ฐ (0.7%) ๐Ÿ‡ฑ๐Ÿ‡บ (0.4%) ๐Ÿ‡บ๐Ÿ‡ฆ (0.4%) ๐Ÿ‡ธ๐Ÿ‡ฎ (0.4%) ๐Ÿ‡ณ๐Ÿ‡ด (0.4%) ๐Ÿ‡ช๐Ÿ‡ช (0.1%) ๐Ÿ‡ฐ๐Ÿ‡ฟ (0.7%) ๐Ÿ‡ฆ๐Ÿ‡ฟ (0.7%) ๐Ÿ‡ฉ๐Ÿ‡ฐ (0.7%) ๐Ÿ‡จ๐Ÿ‡ฟ (0.6%) ๐Ÿ‡บ๐Ÿ‡ฆ (0.3%) ๐Ÿ‡จ๐Ÿ‡ท (0.3%) ๐Ÿ‡ต๐Ÿ‡ท (0.3%) ๐Ÿ‡ง๐Ÿ‡ง (0.3%) ๐Ÿ‡ณ๐Ÿ‡ฑ (0.3%) ๐Ÿ‡ฆ๐Ÿ‡ท (0.3%) ๐Ÿ‡ฑ๐Ÿ‡บ (0.3%) ๐Ÿ‡ต๐Ÿ‡ฑ (0.1%) ๐Ÿ‡จ๐Ÿ‡ด (0.1%) ๐Ÿ‡ฒ๐Ÿ‡ฆ (0.1%) ๐Ÿ‡จ๐Ÿ‡ฑ (0.1%)

๐ŸŽซ Olympics qualification

The Olympics competition has 55 spots, with a limit of 3 athletes per nations.

Considering nations with a high percentage of athletes in top-50, let's estimate how many athletes miss the Olympics because of the 'max-3-rule':

  • Women:
    • ๐Ÿ‡บ๐Ÿ‡ธ : ~11.0% => ~3.0 top-50 athletes rejected. ๐Ÿ˜ž
    • ๐Ÿ‡ฌ๐Ÿ‡ง : ~10.9% => ~3.0 top-50 athletes rejected. ๐Ÿ˜ž
    • ๐Ÿ‡ฆ๐Ÿ‡บ : ~10.1% => ~2.6 top-50 athletes rejected. ๐Ÿ˜ž
    • ๐Ÿ‡ฏ๐Ÿ‡ต : ~8.0% => ~1.4 top-50 athletes rejected. ๐Ÿ˜ž
    • ๐Ÿ‡ฉ๐Ÿ‡ช : ~7.6% => ~1.2 top-50 athletes rejected. ๐Ÿ˜ž
    • ๐Ÿ‡ณ๐Ÿ‡ฟ : ~5.7% => ~0.1 top-50 athlete rejected. ๐Ÿ˜ž
  • Men:
    • ๐Ÿ‡ซ๐Ÿ‡ท : ~9.6% => ~2.3 top-50 athletes rejected. ๐Ÿ˜ž
    • ๐Ÿ‡ฌ๐Ÿ‡ง : ~8.4% => ~1.6 top-50 athletes rejected. ๐Ÿ˜ž
    • ๐Ÿ‡ฉ๐Ÿ‡ช : ~8.1% => ~1.5 top-50 athletes rejected. ๐Ÿ˜ž
    • ๐Ÿ‡ฆ๐Ÿ‡บ : ~7.7% => ~1.2 top-50 athletes rejected. ๐Ÿ˜ž
    • ๐Ÿ‡ช๐Ÿ‡ธ : ~7.0% => ~0.9 top-50 athlete rejected. ๐Ÿ˜ž
    • ๐Ÿ‡บ๐Ÿ‡ธ : ~6.1% => ~0.4 top-50 athlete rejected. ๐Ÿ˜ž

The estimated numbers of rejections are probably lower bounds:

  • World Triathlon limits the number of athletes per nation for its races too.
    • As a result, some strong athletes, such as US women, cannot participate in important World Triathlon races, and thus do not gain points for the ranking.

The percentages are averages since 2009. Some years, they can be much higher:

  • For instance, 8 women in the top-50 (16.0%) for:
    • ๐Ÿ‡บ๐Ÿ‡ธ 2015 (a pre-olympics year), with 6 top-30 and the 4th athlete was ranked 16th.
    • ๐Ÿ‡บ๐Ÿ‡ธ 2016, with 6 top-20.
    • ๐Ÿ‡ฆ๐Ÿ‡บ 2017.
    • ๐Ÿ‡ฌ๐Ÿ‡ง 2021, with 8 top-30.
      • The same year, ๐Ÿ‡บ๐Ÿ‡ธ had 5 athletes in the top-12.

It is no wonder that some athletes change nationality to try to qualify for the Olympics.

๐Ÿ‘ถ AGE

ages.png
Age of athletes ranked 5th-9th over years.

Athletes finishing 5th-9th are, on average, 26.3 to 27.7 years old.

  • Ages are similar for women and men.
  • Athletes are slightly older in the olympic format compared to the sprint format: about 1 year difference.
    • I would have expected a larger difference.
  • There are some small variations, but no significant trends over the years.

๐Ÿ AGE OF LAST RACE

ages_of_last_race.png
Age of last world-cup, world-series or major games.

The average age of the last race is similar for women and men: around 31 years.

The distribution is broad (~4y std), because there are various reasons for ending a World Triathlon sprint- and olympic-distance top career, such as:

  • Age limit for elite sport.
  • Transitioning to longer-distance triathlons
  • Injury.
  • Personal reasons, such as pregnancy or changing careers.

Some extreme values:

  • <25 years:
    • Hollie Avil ( ๐Ÿ‡ฌ๐Ÿ‡ง ): 21y, 13 races.
    • Kirsten Nuyes ( ๐Ÿ‡ณ๐Ÿ‡ฑ ): 22y, 21 races.
    • Ellen Pennock ( ๐Ÿ‡จ๐Ÿ‡ฆ ): 23y, 17 races.
    • Hanna Philippin ( ๐Ÿ‡ฉ๐Ÿ‡ช ): 24y, 32 races.
    • Marc Austin ( ๐Ÿ‡ฌ๐Ÿ‡ง ): 24y, 25 races.
    • Sophia Saller ( ๐Ÿ‡ฉ๐Ÿ‡ช ): 24y, 26 races.
    • Oliver Freeman ( ๐Ÿ‡ฌ๐Ÿ‡ง ): 24y, 20 races.
    • Raphael Montoya ( ๐Ÿ‡ซ๐Ÿ‡ท ): 24y, 21 races.
  • >38 years:
    • Magali Di Marco Messmer ( ๐Ÿ‡จ๐Ÿ‡ญ ): 39y, 52 races.
    • Greg Bennett ( ๐Ÿ‡ฆ๐Ÿ‡บ ): 39y, 70 races.
    • Hunter Kemper ( ๐Ÿ‡บ๐Ÿ‡ธ ): 39y, 65 races.
    • Kate Allen ( ๐Ÿ‡ฆ๐Ÿ‡น ): 39y, 24 races.
    • Samantha Warriner ( ๐Ÿ‡ณ๐Ÿ‡ฟ ): 42y, 48 races.
    • Kiyomi Niwata ( ๐Ÿ‡ฏ๐Ÿ‡ต ): 43y, 96 races.
Click to expand - ๐Ÿ“œ Full list.
ATHLETE COUNTRY AGE OF LAST RACE NUMBER OF RACES
Hollie Avil ๐Ÿ‡ฌ๐Ÿ‡ง 21 13
Kirsten Nuyes ๐Ÿ‡ณ๐Ÿ‡ฑ 22 21
Ellen Pennock ๐Ÿ‡จ๐Ÿ‡ฆ 23 17
Hanna Philippin ๐Ÿ‡ฉ๐Ÿ‡ช 24 32
Marc Austin ๐Ÿ‡ฌ๐Ÿ‡ง 24 25
Sophia Saller ๐Ÿ‡ฉ๐Ÿ‡ช 24 26
Oliver Freeman ๐Ÿ‡ฌ๐Ÿ‡ง 24 20
Raphael Montoya ๐Ÿ‡ซ๐Ÿ‡ท 24 21
Akane Tsuchihashi ๐Ÿ‡ฏ๐Ÿ‡ต 25 27
Sarissa De Vries ๐Ÿ‡ณ๐Ÿ‡ฑ 25 26
Ron Darmon ๐Ÿ‡ฎ๐Ÿ‡ฑ 25 44
Daniela Ryf ๐Ÿ‡จ๐Ÿ‡ญ 25 42
Artem Parienko ๐Ÿ‡ท๐Ÿ‡บ 26 23
Jose Miguel Perez ๐Ÿ‡ช๐Ÿ‡ธ 26 30
James Seear ๐Ÿ‡ฆ๐Ÿ‡บ 26 29
Gareth Halverson ๐Ÿ‡ฆ๐Ÿ‡บ 26 17
Lucy Buckingham ๐Ÿ‡ฌ๐Ÿ‡ง 26 44
David McNamee ๐Ÿ‡ฌ๐Ÿ‡ง 26 34
Wian Sullwald ๐Ÿ‡ฟ๐Ÿ‡ฆ 26 70
Natalie Milne ๐Ÿ‡ฌ๐Ÿ‡ง 27 14
Aaron Harris ๐Ÿ‡ฌ๐Ÿ‡ง 27 27
Franz Lรถschke ๐Ÿ‡ฉ๐Ÿ‡ช 27 36
Andrey Bryukhankov ๐Ÿ‡ท๐Ÿ‡บ 27 36
Paul Tichelaar ๐Ÿ‡จ๐Ÿ‡ฆ 27 33
Felicity Abram ๐Ÿ‡ฆ๐Ÿ‡บ 27 46
Mariya Shorets ๐Ÿ‡ท๐Ÿ‡บ 27 47
Sebastian Rank ๐Ÿ‡ฉ๐Ÿ‡ช 27 28
Matthew Sharp ๐Ÿ‡ฌ๐Ÿ‡ง 27 22
Jenna Parker ๐Ÿ‡บ๐Ÿ‡ธ 28 21
Denis Vasiliev ๐Ÿ‡ท๐Ÿ‡บ 28 29
Benjamin Shaw ๐Ÿ‡ฎ๐Ÿ‡ช 28 50
Sarah-Anne Brault ๐Ÿ‡จ๐Ÿ‡ฆ 28 32
Agnieszka Jerzyk ๐Ÿ‡ต๐Ÿ‡ฑ 28 52
Andrew Yorke ๐Ÿ‡จ๐Ÿ‡ฆ 28 42
Maaike Caelers ๐Ÿ‡ณ๐Ÿ‡ฑ 28 56
Tamara Gomez Garrido ๐Ÿ‡ช๐Ÿ‡ธ 28 32
Ivan Tutukin ๐Ÿ‡ฐ๐Ÿ‡ฟ 28 34
Rebecca Robisch ๐Ÿ‡ฉ๐Ÿ‡ช 28 46
Kathrin Muller ๐Ÿ‡ฉ๐Ÿ‡ช 28 39
Kirsten Sweetland ๐Ÿ‡จ๐Ÿ‡ฆ 28 49
Paula Findlay ๐Ÿ‡จ๐Ÿ‡ฆ 28 41
Peter Croes ๐Ÿ‡ง๐Ÿ‡ช 28 58
Kaitlin Donner ๐Ÿ‡บ๐Ÿ‡ธ 28 37
Jason Wilson ๐Ÿ‡ง๐Ÿ‡ง 28 45
William Clarke ๐Ÿ‡ฌ๐Ÿ‡ง 28 44
Jodie Swallow ๐Ÿ‡ฌ๐Ÿ‡ง 29 32
Vanessa Raw ๐Ÿ‡ฌ๐Ÿ‡ง 29 25
Emmie Charayron ๐Ÿ‡ซ๐Ÿ‡ท 29 43
Katie Hewison ๐Ÿ‡ฌ๐Ÿ‡ง 29 16
Jillian Elliott ๐Ÿ‡บ๐Ÿ‡ธ 29 31
Andreas Giglmayr ๐Ÿ‡ฆ๐Ÿ‡น 29 51
Radka Kahlefeldt ๐Ÿ‡จ๐Ÿ‡ฟ 29 31
Danne Boterenbrood ๐Ÿ‡ณ๐Ÿ‡ฑ 29 20
Charlotte Bonin ๐Ÿ‡ฎ๐Ÿ‡น 29 62
Svenja Bazlen ๐Ÿ‡ฉ๐Ÿ‡ช 29 28
Elizabeth May ๐Ÿ‡ฑ๐Ÿ‡บ 29 55
Annabel Luxford ๐Ÿ‡ฆ๐Ÿ‡บ 29 52
Anastasia Abrosimova ๐Ÿ‡ท๐Ÿ‡บ 29 44
Cameron Good ๐Ÿ‡ฆ๐Ÿ‡บ 30 33
Mari Rabie ๐Ÿ‡ฟ๐Ÿ‡ฆ 30 50
Brendan Sexton ๐Ÿ‡ฆ๐Ÿ‡บ 30 57
Kathy Tremblay ๐Ÿ‡จ๐Ÿ‡ฆ 30 54
Jodie Stimpson ๐Ÿ‡ฌ๐Ÿ‡ง 30 71
Laurent Vidal ๐Ÿ‡ซ๐Ÿ‡ท 30 61
Yulian Malyshev ๐Ÿ‡ท๐Ÿ‡บ 30 33
Mark Buckingham ๐Ÿ‡ฌ๐Ÿ‡ง 30 22
Gregor Buchholz ๐Ÿ‡ฉ๐Ÿ‡ช 30 59
Debbie Tanner ๐Ÿ‡ณ๐Ÿ‡ฟ 30 55
Pamella Oliveira ๐Ÿ‡ง๐Ÿ‡ท 30 62
India Lee ๐Ÿ‡ฌ๐Ÿ‡ง 31 18
Valentin Mechsheryakov ๐Ÿ‡ฐ๐Ÿ‡ฟ 31 51
Lauren Groves ๐Ÿ‡จ๐Ÿ‡ฆ 31 57
Christian Prochnow ๐Ÿ‡ฉ๐Ÿ‡ช 31 41
Matt Chrabot ๐Ÿ‡บ๐Ÿ‡ธ 31 43
Melanie Hauss ๐Ÿ‡จ๐Ÿ‡ญ 31 45
Clark Ellice ๐Ÿ‡ณ๐Ÿ‡ฟ 31 53
Gavin Noble ๐Ÿ‡ฎ๐Ÿ‡ช 31 38
Tony Dodds ๐Ÿ‡ณ๐Ÿ‡ฟ 31 58
Emma Snowsill ๐Ÿ‡ฆ๐Ÿ‡บ 31 44
Miguel Arraiolos ๐Ÿ‡ต๐Ÿ‡น 31 72
Ricarda Lisk ๐Ÿ‡ฉ๐Ÿ‡ช 31 63
Line Jensen ๐Ÿ‡ฉ๐Ÿ‡ฐ 31 20
Ruedi Wild ๐Ÿ‡จ๐Ÿ‡ญ 31 55
Kate Roberts ๐Ÿ‡ฟ๐Ÿ‡ฆ 31 56
Helle Frederiksen ๐Ÿ‡ฉ๐Ÿ‡ฐ 31 28
Joe Maloy ๐Ÿ‡บ๐Ÿ‡ธ 31 37
Aurelien Raphael ๐Ÿ‡ซ๐Ÿ‡ท 31 59
Dan Wilson ๐Ÿ‡ฆ๐Ÿ‡บ 31 59
Annamaria Mazzetti ๐Ÿ‡ฎ๐Ÿ‡น 31 77
Rebecca Spence ๐Ÿ‡ณ๐Ÿ‡ฟ 31 39
Simon De Cuyper ๐Ÿ‡ง๐Ÿ‡ช 32 55
David Hauss ๐Ÿ‡ซ๐Ÿ‡ท 32 54
Erin Densham ๐Ÿ‡ฆ๐Ÿ‡บ 32 64
Premysl Svarc ๐Ÿ‡จ๐Ÿ‡ฟ 32 74
Yurie Kato ๐Ÿ‡ฏ๐Ÿ‡ต 32 52
Helen Jenkins ๐Ÿ‡ฌ๐Ÿ‡ง 32 55
Emma Moffatt ๐Ÿ‡ฆ๐Ÿ‡บ 32 76
Jan Frodeno ๐Ÿ‡ฉ๐Ÿ‡ช 32 50
Leonardo Chacon ๐Ÿ‡จ๐Ÿ‡ท 32 86
Kyle Jones ๐Ÿ‡จ๐Ÿ‡ฆ 32 90
Alexander Bryukhankov ๐Ÿ‡ท๐Ÿ‡บ 32 86
Brent McMahon ๐Ÿ‡จ๐Ÿ‡ฆ 32 71
Manuel Huerta ๐Ÿ‡ต๐Ÿ‡ท 32 60
Liz Blatchford ๐Ÿ‡ฌ๐Ÿ‡ง 32 62
Mark Fretta ๐Ÿ‡บ๐Ÿ‡ธ 32 70
Ivan Vasiliev ๐Ÿ‡ท๐Ÿ‡บ 32 81
Nicky Samuels ๐Ÿ‡ณ๐Ÿ‡ฟ 33 76
Brad Kahlefeldt ๐Ÿ‡ฆ๐Ÿ‡บ 33 78
Bruno Pais ๐Ÿ‡ต๐Ÿ‡น 33 64
Lisa Norden ๐Ÿ‡ธ๐Ÿ‡ช 33 67
Kate Mcilroy ๐Ÿ‡ณ๐Ÿ‡ฟ 33 39
Jonathan Zipf ๐Ÿ‡ฉ๐Ÿ‡ช 33 46
Mariko Adachi ๐Ÿ‡ฏ๐Ÿ‡ต 33 61
Tomoko Sonoda ๐Ÿ‡ฏ๐Ÿ‡ต 33 34
Irina Abysova ๐Ÿ‡ท๐Ÿ‡บ 33 48
Anne Haug ๐Ÿ‡ฉ๐Ÿ‡ช 33 45
Gonzalo Raul Tellechea ๐Ÿ‡ฆ๐Ÿ‡ท 33 51
Frederic Belaubre ๐Ÿ‡ซ๐Ÿ‡ท 33 44
Mary Beth Ellis ๐Ÿ‡บ๐Ÿ‡ธ 33 24
Lindsey Jerdonek ๐Ÿ‡บ๐Ÿ‡ธ 33 38
Margit Vanek ๐Ÿ‡ญ๐Ÿ‡บ 33 59
Misato Takagi ๐Ÿ‡ฏ๐Ÿ‡ต 33 31
Carole Peon ๐Ÿ‡ซ๐Ÿ‡ท 34 55
Claude Eksteen ๐Ÿ‡ฟ๐Ÿ‡ฆ 34 25
Marina Damlaimcourt ๐Ÿ‡ช๐Ÿ‡ธ 34 53
Thomas Springer ๐Ÿ‡ฆ๐Ÿ‡น 34 41
Aileen Reid ๐Ÿ‡ฎ๐Ÿ‡ช 34 60
Danylo Sapunov ๐Ÿ‡บ๐Ÿ‡ฆ 34 78
Jarrod Shoemaker ๐Ÿ‡บ๐Ÿ‡ธ 34 93
Tim Don ๐Ÿ‡ฌ๐Ÿ‡ง 34 73
Christiane Pilz ๐Ÿ‡ฉ๐Ÿ‡ช 34 40
Grรฉgory Rouault ๐Ÿ‡บ๐Ÿ‡ธ 34 19
Katrien Verstuyft ๐Ÿ‡ง๐Ÿ‡ช 34 53
Zuriรฑe Rodriguez Sanchez ๐Ÿ‡ช๐Ÿ‡ธ 34 63
Sarah Haskins ๐Ÿ‡บ๐Ÿ‡ธ 34 41
Maik Petzold ๐Ÿ‡ฉ๐Ÿ‡ช 34 61
Felicity Sheedy-Ryan ๐Ÿ‡ฆ๐Ÿ‡บ 34 50
Hirokatsu Tayama ๐Ÿ‡ฏ๐Ÿ‡ต 35 87
Lisa Mensink ๐Ÿ‡จ๐Ÿ‡ฆ 35 42
Kris Gemmell ๐Ÿ‡ณ๐Ÿ‡ฟ 35 84
Sven Riederer ๐Ÿ‡จ๐Ÿ‡ญ 35 87
Bevan Docherty ๐Ÿ‡ณ๐Ÿ‡ฟ 35 78
Andy Potts ๐Ÿ‡บ๐Ÿ‡ธ 35 32
Zita Szabรณ ๐Ÿ‡ญ๐Ÿ‡บ 35 44
Adam Bowden ๐Ÿ‡ฌ๐Ÿ‡ง 35 49
Jessica Harrison ๐Ÿ‡ซ๐Ÿ‡ท 36 81
Cedric Fleureton ๐Ÿ‡ซ๐Ÿ‡ท 36 36
Mateja ล imic ๐Ÿ‡ธ๐Ÿ‡ฎ 36 47
Bryan Keane ๐Ÿ‡ฎ๐Ÿ‡ช 36 45
Marek Jaskolka ๐Ÿ‡ต๐Ÿ‡ฑ 36 49
Sarah True ๐Ÿ‡บ๐Ÿ‡ธ 36 85
Kerry Lang ๐Ÿ‡ฌ๐Ÿ‡ง 36 43
Vladimir Turbayevskiy ๐Ÿ‡ท๐Ÿ‡บ 36 67
Tony Moulai ๐Ÿ‡ซ๐Ÿ‡ท 37 58
Diogo Sclebin ๐Ÿ‡ง๐Ÿ‡ท 37 81
Laura Bennett ๐Ÿ‡บ๐Ÿ‡ธ 37 68
Courtney Atkinson ๐Ÿ‡ฆ๐Ÿ‡บ 37 68
Reto Hug ๐Ÿ‡จ๐Ÿ‡ญ 37 67
Simon Whitfield ๐Ÿ‡จ๐Ÿ‡ฆ 37 78
Ryosuke Yamamoto ๐Ÿ‡ฏ๐Ÿ‡ต 37 82
Ainhoa Murua Zubizarreta ๐Ÿ‡ช๐Ÿ‡ธ 38 100
Magali Di Marco Messmer ๐Ÿ‡จ๐Ÿ‡ญ 39 52
Greg Bennett ๐Ÿ‡ฆ๐Ÿ‡บ 39 70
Hunter Kemper ๐Ÿ‡บ๐Ÿ‡ธ 39 65
Kate Allen ๐Ÿ‡ฆ๐Ÿ‡น 39 24
Samantha Warriner ๐Ÿ‡ณ๐Ÿ‡ฟ 42 48
Kiyomi Niwata ๐Ÿ‡ฏ๐Ÿ‡ต 43 96

๐Ÿ“… MONTH OF BIRTH

For this section and the next one about Body mass index, a larger dataset is used:

  • All athletes part of a non-para ranking, and aged between 15 and 45 years, are considered.
    • The ranking categories defined by World Triathlon can be found here.
  • Leading to a total of 3,439 unique athletes, including 1,382 women and 2,057 men.

The month-of-birth distribution of the athletes is compared to two reference distributions:

  • Reference #1: Uniform month-of-birth distribution.
    • It could be expected that each month accounts for 1/12 = 8.3% of the births.
  • Reference #2: Birth data collected by the United Nation: data.un.org.
    • Birth data of people aged between 20 and 30 years are considered, leading to over 230 million entries.
    • It could be expected that World Triathlon (formerly ITU) and UN month-of-birth distributions match.
Month-of-birth of World Triathlon (formerly ITU) athletes.

The results are even more striking when considering the year quarters.
Year-quarter-of-birth of World Triathlon (formerly ITU) athletes.
Click to expand - ๐Ÿ‘ซ Same plot with gender comparison.
Year-quarter-of-birth of World Triathlon (formerly ITU) athletes.
Click to expand - ๐Ÿ‘ถ Same plot for the Junior categories.

The average age is 19.5 years.
Year-quarter-of-birth of World Triathlon (formerly ITU) athletes in the Junior categories.
Click to expand - ๐Ÿ“Š Reference distribution from the United Nations.
Month-of-birth distribution of the population aged 20-30 recorded by UN. Data source: data.un.org

Can these discrepancies be due to differences between the two datasets (ITU and UN), such as the geographical origin of the births?

Click to expand - ๐ŸŒ Continent analysis.

The continent distributions differ between the World Triathlon and the UN datasets:

  • ITU dataset: Europe is predominant (~59%). Asia (~13%) and North America (~11%) follow.
  • UN dataset: Asia (~34%), Europe (29%) and North America (24%) form a more uniform top-3.
Continents distribution of the World Triathlon (formerly ITU) dataset
Continents distribution of the UN dataset (used as reference).

Note: The UN dataset has probably missed data from African and Asian countries such as China, India and Nigeria.
Another visualization of the reference month-of-birth distribution.

The month-of-birth and quarter-of-birth distributions for each continent can be visualized:
Month-of-birth distribution, by continent (normalized).
Quarter-of-birth distribution, by continent (normalized).

Conclusion:

  • The continents mainly represented in the two datasets (ITU and UN) share very similar month-of-birth and quarter-of-birth distributions.
  • Therefore, the difference in continent distributions does not explain the discrepancy in month-of-birth and quarter-of-birth distributions between the ITU and UN datasets.
Click to expand - ๐ŸŽ‚ Age distribution of World Triathlon athletes used for this analysis.

The UN and ITU datasets share the same average age (25), but the ITU age distribution is not uniform, unlike the UN one.

  • Importance sampling could be applied to the UN dataset to make the two distributions match, but I would be very surprised if that had an impact on the overall conclusion.
age.png
Age of athletes considered for the month-of-birth analysis.

Is the difference between ITU and UN month- and quarter-of-birth distributions statistically significant?

Click to expand - ๐Ÿงฎ Statistical test.

3,439 birth entries have been collected ("observed") from the World Triathlon (formerly ITU) data.

  • A priori there is no link between the quarter-of-birth and the fact of being a high-level triathlete.
  • Therefore, it can be assumed that the ITU observations follow the UN distribution.

From the UN distribution, the expected number of births for each quarter is computed:

OBSERVED (ITU) EXPECTED (UN)
Q1 987 854.443
Q2 859 845.248
Q3 820 893.717
Q4 773 845.592

The number of births in the two columns look very different: more observations than expected for Q1, fewer for Q3 and Q4.

  • The situation is similar to the case of a 6-sided dice, which is suspected to be biased. ๐ŸŽฒ
    • After rolling the die many times and counting each outcome, one can ask:
    • Is just due to randomness or is it because the dice is unfair?

Back to our problem:

  • How to quantify the deviation between the two columns?
OBSERVED (ITU) EXPECTED (UN) DIFF DIFF^2 DIFF^2 / EXPECTED
Q1 987 854.443 132.557 17571.3 20.5646
Q2 859 845.248 13.7516 189.106 0.223728
Q3 820 893.717 -73.7166 5434.14 6.08039
Q4 773 845.592 -72.5916 5269.54 6.23178

One can compute the differences between columns, square them to ensure they are positive, normalize them and sum the results:

  • SUM OF [DIFF^2 / EXPECTED] = 20.5646 + 0.2237 + 6.0804 + 6.2318 = 33.10
  • It 33.10 large? What does it mean? What can be concluded?

There is a mathematical formula that, given this computed number (33.10), answers the following question:

  • What is the probability of observing such a discrepancy (33.10) or an even larger one, assuming that the ITU data should follow the UN quarter-of-birth distribution?
  • In other words: How likely is it that the observed deviation is due to random chance?

For 33.10, the formula gives p = 0.0000003, i.e. 0.00003%.

Conclusion:

  • The extremely low probability (0.00003%) indicates that the observed differences in quarters-of-birth among World Triathlon athletes are highly unlikely to be due to random chance.
  • Therefore, the observed differences are statistically significant.
  • This suggests a systematic deviation from the expected UN distribution.
  • In other words, the quarters-of-birth of high-level triathletes do not align with the general population as represented by the UN distribution.

For more details about the derivation:

โš ๏ธ These findings do NOT indicate that "World Triathlon athletes born in January are more performant than others born in later months"!

  • This birth-of-month analysis rather shows that someone born earlier in one year is MORE LIKELY TO BECOME PRO TRIATHLETE than one born later in the same year.

Here is one possible explanation:

  • Kids born on January, 1st and December 31st of the same year compete in the same age category.
  • For example, at the age of 12, a 12-month difference represents 10% of their lifetime.
  • A 12-year kid born in January could, in theory, be ~10% physically more developed (in terms of strength and stamina) than one born in December.
  • This edge could give an advantage to kids and teenagers born in the first months of a year during their early races in youth categories.
  • They are more likely to perform well, stand out, gain experience, and get selected for international competitions, eventually becoming professional. ๐Ÿ†
    • Similar to a snowball effect. ๐ŸŒ€
  • This phenomenon, known as relative age effect, was also observed on young Spanish triathletes by Ferriz et al. in 2020.

๐Ÿ‹๏ธ BODY MASS INDEX

"The BMI is defined as the body mass divided by the square of the body height, and is expressed in units of kg/m^2"

To maximize the amount of data, all athletes registered with World Triathlon are considered (not just those participating in world-cups and world-series).

  • In total, 504 athletes aged between 15 and 53 (average is 27) are included.
bmi.png
Body Mass Index.
weight_height.png
Weights and Heights.

First, is the BMI a relevant metric here? No!

  • It is useful for public health, providing a quick and simple measure to identify trends in obesity and underweight conditions.
  • But it does not measure fitness levels, physical endurance, strength, flexibility, or other aspects of physical health.
  • Many athletes may have a high BMI due to increased muscle mass, yet they are fit and healthy.

Second, there is no enough data.

  • Only 14% of athletes registered with World Triathlon have valid weight and height information (504 out of 3560).
  • Most athletes did not input their dimensions.
  • Sometimes for privacy reasons, e.g. the entry for Kristian Blummenfelt's ( ๐Ÿ‡ณ๐Ÿ‡ด ) weight: "None of your business".

Third, some data may be outdated.

  • The body weight that can vary during a career (many athletes enter the database as juniors) or even during a season.

Alternative metrics such as Body Fat Percentage or Muscle Mass Percentage would be more appropriate for fitness assessment.

If not much can be concluded, comparing oneself to the distributions can still be interesting.

Click to expand - ๐Ÿ“ Non-standardized data format!

Here are examples of height data. ๐Ÿ“

 '6 Foot',
 "194, 6'4",
 '1.76 mts',
 '62 kg ',
 '1.80 meters',
 '174m',
 '5\'2"',
 '165 cms',
 '1,81 CM',
 '6ft. 2in.',
 '6ft 2in',
 '5โ€™7โ€',
 '6 ft 0 in',
 '170cm/5.58ft',
 '57kg',
 '5.5ft',
 '1,77mts',
 'MT. 175',
 '164,4cm',
 '1.66 MTS',
 '5\'8"',
 '183 cm',
 "6'4",
 '1.9m',
 '6 Ft',
 "1'72",
 '186cm',
 '182',
 '180 ',
 '182CM'

And some weight data. ๐Ÿ‹๏ธ

 'None of your business ',
 '160',
 '70 kg โ€œin season โ€œ ;-)',
 '1:68 m ',
 '140 lb',
 '51kgs ',
 '138 LBs',
 '1,78',
 '75 KG',
 '135ibs',
 '54kg/119lb',
 '42 kilo',
 '135IIbs',
 '187 lbs',
 'KG. 62',
 '49,4 kg',
 '67KG',
 '62 Kg',
 '175',
 '145lb',
 '52kg',
 '67',
 '140 Ibs',
 '65',
 '75 ',
 '66KG'

Interesting to see the different units and formats.

Some processing and filtering are required! ๐Ÿงน

๐Ÿ“œ CONCLUSION

๐Ÿ’ก IDEAS

Here are some ideas for data to explore:

๐Ÿ” 1) MORE DETAILED RACES ANALYSES

As mentioned by Alex Yee in this video:

"You are doing a full gas effort at the start of a race which is 2 hours long. If you told somebody to do that on a marathon, they would laugh in your face."

World Triathlon data provides a single time for each leg: swim, t1, bike, t2, run. There is no detail about the pace evolution during each segment. For instance:

  • Fast swim start to reach the first buoy.
  • Fast bike start to break away or catch a pack.
  • Fast run start - I have never really understood the benefit compared to a steady effort.
  • Fast run finish.

Some events, such as Paris Olympics, offer detailed results with lap times.

  • It is for instance very instructive to compare the runs of Alex Yee ( ๐Ÿ‡ฌ๐Ÿ‡ง ), Hayden Wilde ( ๐Ÿ‡ณ๐Ÿ‡ฟ ) and Lรฉo Bergรจre ( ๐Ÿ‡ซ๐Ÿ‡ท ).
  • The first two completed the first 1050m in 2:42.
  • Alex Yee then "slowed down" to ~3:00 for the following three laps, while Lรฉo Bergรจre was a bit more regular: 2:49, 2:56, 3:01, 2:54.
  • Such data should be very interesting to analyse.

In addition to paces, it would be interesting to access data such as:

  • Swim stroke rate.
  • Bike power and cadence.
  • Run cadence.
  • Run maximum speed during the sprint.
  • Heart rate.
  • ...
  • For comparison, heart rate information from some athletes is shown in UCI mountain-bike world-cups videos.

Activity trackers would be needed for these recordings, but athletes rarely wear them while racing, making swim and run data recording difficult.

Strava could be used to retrieve activities:

  • Many athletes publish their activities there, either their races (often limited to the bike section) or training sessions.
  • However, as mentioned, almost no swim or run information is recorded during races.

๐Ÿ›๏ธ 2) EQUIPMENT CHOICE

What is the best wetsuit? What are the fastest running shoes?

  • There are already some tests and reports conducted by researchers.
  • However, I believe that examining athletes' preferences would provide more reliable and valuable results.

Not all athletes are sponsored by swimming or running brands.

  • Therefore, many have the freedom to experiment, compare, and choose the equipment they believe will enhance their performance.
  • As an example, one could track how many athletes wore Asics, Adidas, Nike, New Balance, etc., at major events, noting the models used, and exclude those provided by sponsors.

I recall reading about a similar project but can no longer find the reference.

๐Ÿ’ฐ 3) MONEY

It would be interesting to investigate the financial aspects of the competitions, such as:

  • The prize money for the different World Triathlon race categories.
  • The event registration costs.
  • Eventually, to estimate from which rank an athlete can make a descent living. (Of course, sponsoring and federation support also play a role).

๐Ÿ 4) MISCELLANEOUS

  • Analyse the correlation between transition ranking and finish ranking.
    • Especially for T2.
  • Compare WTCS and world-cups more thoroughly.
  • The arbitrary decision to focus on the top 5-9 was made to capture a stable and consistent representation of the general competitive field.
    • However, examining the top performance (e.g. top-1 or top-3) or using a broader range could also yield valuable insights.
  • Conduct advanced analyses of cycling performances would be interesting.
    • Additional data may be required: drawing conclusions based solely on bike split times is challenging, due to the influence of drafting and pack dynamics.
  • Investigate the impact of swim conditions - including water temperature, presence of waves, and salinity - on swim performance and race dynamics.
  • Take a closer look at the critical start of the bike segment.
    • For example, given a lag at T1, how likely is it to catch the first group?
  • Understand the system of penalties and their impact on race dynamics.
    • The 2024 Paris Olympics ( ๐Ÿ‡ซ๐Ÿ‡ท ) saw a surprising number of penalties: 6 women out of 51 and 10 men out of 50 received a 15s penalty.
    • Before focusing on gaining seconds, some athletes may need to prioritize avoiding penalties, such as by correctly timing their dive and making sure to place their helmet inside the transition box.
  • Analyze the trajectory of successful elite athletes.
    • How did they perform as juniors, and how did they progress from junior to U23 to elite?
  • Try to estimate the level of a race.
    • For example, using the results table and the ranking of the participating athletes.
  • ...

๐Ÿ’ป CODE

The python code to fetch the data, set the parameters and generate plots is available in this GitHub repository. To use it,

  • Create a key for the World Triathlon API: https://apps.api.triathlon.org/ and add it to a api_key.txt file directly placed in tri_stats/.
  • Install the required python packages.
  • Run the different scripts of this repository.

๐Ÿฅก TAKEAWAYS

โš ๏ธ Most summary numbers given in this section are AVERAGES.

  • For instance race-averages are computed from the 5th to 9th best times of each leg, and sometimes averaged over multiples years.

Here are some simplified key takeaways:

  • โฑ๏ธ The three sports account for 16.4% ( ๐ŸŠ ), 53.1% ( ๐Ÿšด ), 28.9% ( ๐Ÿƒ ) of the overall time. Transitions for 1.1% and 0.5%.
  • ๐Ÿ‘Ÿ While swim and bike paces are similar between sprint and olympic formats, the 10k run requires 7 s/km more than the 5k.
  • ๐ŸŠ Women swim at 1:18 / 100m, men at 1:12 / 100m.
  • ๐Ÿšด Women ride 4 km/h slower, at 37.4 km/h, compared to men at 41.4 km/h.
  • ๐Ÿƒ Women run the 10k at 3:33 min/km (3:26 for 5k), men at 3:07 min/km (3:00 for 5k).
  • ๐Ÿ‘ซ Women swim 8.8% slower than men with the same equipment. They also ride 10.6% and run 14.2% slower.
  • ๐Ÿ“‰ The women/men difference has not significantly reduced on the years, except for the run leg of the sprint-format races (-0.13 % / year) and for the swim of WTCS (-11 % / year).
  • ๐Ÿง There is no evidence that wetsuits reduce swim gaps between top and less competitive swimmers.
  • ๐Ÿฉฑ Swim is ~5% faster with wetsuit (to be refined).
  • ๐Ÿ‡ซ๐Ÿ‡ท The swim of 2024 Paris Olympics was unusually long (more than 2:30 longer), probably because of the current in La Seine. In particular, the 5-9th women swam more than 1:30 / 100m.
  • โšก Winning by a run comeback, i.e. after not ending the bike in the front group, is entertaining but rare in the olympic format (28% for men and 7% for women) and is getting even rarer.
  • ๐Ÿšด The size of the front group after bike averages around 15. It decreases to 4 or fewer (small breakaway) in about 1/4 of women's and 1/3 of men's olympic-races.
  • ๐Ÿ‘Ÿ Over 2/3 of races are won by the best runner.
  • ๐Ÿ“ธ In men's races, 17% (sprint format) and 10% (olympic) are won by a sprint finish, occurring 50% more often than in women's races.
  • ๐Ÿคธโ€โ™€๏ธ Women's races occasionally feature wins by very large margins.
  • ๐Ÿ“ The gaps between the winner and the second are, on average, twice as large in olympic formats compared to sprint formats, and twice as large for women compared to men.
  • ๐Ÿš€ Bike and run times in WTCS olympic races have reached all-time lows.
  • ๐ŸŒก๏ธ Athletes must compete in a variety of conditions: Notably, 80% of the recorded air temperatures (lower estimates) fall between 17ยฐC and 30ยฐC and almost uniformly so.
  • ๐Ÿฅต Heat tends to slow down running pace.
  • ๐Ÿ“† On average, athletes raced 10 times (world cups and WTCS) in 2019 and 2023, compared to 6 times in 2009.
  • ๐Ÿ“† Their World Triathlon sprint- and olympic-distance season has extended from 130 days in 2009 to 200 days in 2023.
  • ๐ŸŽซ The limit of 3 athletes per nation for the Olympics creates challenges for the highly represented nations such as ๐Ÿ‡บ๐Ÿ‡ธ, ๐Ÿ‡ฌ๐Ÿ‡ง, ๐Ÿ‡ฆ๐Ÿ‡บ, ๐Ÿ‡ฉ๐Ÿ‡ช and ๐Ÿ‡ซ๐Ÿ‡ท.
  • ๐ŸŽ‚ Athletes finishing 5th-9th are, on average, between 26 and 28 years old.
  • ๐Ÿ Women and men typically race their last world-cup or WTCS at an average age of 31 years, thought there are significant variations.
  • ๐Ÿ“Š Someone born earlier in the year is more likely to become professional triathletes compared to those born later in the same year.

Thank you for reading until the end!

  • If you have any questions, suggestions, corrections, or comments, please feel free to contact me at simon.chauvin.contact[at]gmail.com.
  • Cheers,
  • Simon ๐Ÿ˜ƒ

About

Analyse data of the World Triathlon (formerly ITU) to try to answer key questions about elite triathlon. ๐ŸŠ๐Ÿšด๐Ÿƒ

Topics

itu sports-stats triathlon sports-data olympics sports-analytics

Resources

Readme
Activity

Stars

0 stars

Watchers

2 watching

Forks

1 fork
Report repository

Releases

No releases published

Packages

No packages published

Languages