Abstract
Background
The IRONMAN® (IM) triathlon is a popular multi-sport, where age group athletes often strive to qualify for the IM World Championship in Hawaii. The aim of the present study was to investigate the location of the fastest IM racecourses for age group IM triathletes. This knowledge will help IM age group triathletes find the best racecourse, considering their strengths and weaknesses, to qualify.
Objective
To determine the fastest IM racecourse for age group IM triathletes using descriptive and predictive statistical methods.
Methods
We collected and analyzed 677,702 age group IM finishers’ records from 228 countries participating in 444 IM competitions held between 2002 and 2022 across 66 event locations. Locations were ranked by average race speed (performance), and countries were sorted by number of records in the sample (participation). A predictive model was built with race finish time as the predicted variable and the triathlete’s gender, age group, country of origin, event location, average air, and water temperatures in each location as predictors. The model was trained with 75% of the available data and was validated against the remaining 25%. Several model interpretability tools were used to explore how each predictor contributed to the model’s predictive power, from which we intended to infer whether one or more predictors were more important than the others.
Results
The average race speed ranking showed IM Vitoria-Gasteiz (1 race only), IM Copenhagen (8 races), IM Hawaii (18 races), IM Tallinn (4 races) and IM Regensburg (2 races) in the first five positions. The XG Boost Regressor model analysis indicated that the IM Hawaii course was the fastest race course and that male athletes aged 35 years and younger were the fastest. Most of the finishers were competing in IM triathlons held in the US, such as IM Wisconsin, IM Florida, IM Lake Placid, IM Arizona, and IM Hawaii, where the IM World Championship took place. However, the fastest average times were achieved in IM Vitoria-Gasteiz, IM Copenhagen, IM Hawaii, IM Tallin, IM Regensburg, IM Brazil Florianopolis, IM Barcelona, or IM Austria with the absolutely fastest race time in IM Hawaii. Most of the successful IM finishers originated from the US, followed by athletes from the UK, Canada, Australia, Germany, and France. The best mean IM race times were achieved by athletes from Austria, Germany, Belgium, Switzerland, Finland, and Denmark. Regarding environmental conditions, the best IM race times were achieved at an air temperature of ∼27°C and a water temperature of ∼24°C.
Conclusions
IM age group athletes who intend to qualify for IM World Championship in IM Hawaii are encouraged to participate in IM Austria, IM Copenhagen, IM Brazil Florianopolis, and/or IM Barcelona in order to achieve a fast race time to qualify for the IM World Championship in IM Hawaii where the top race times were achieved. Most likely these races offer the best ambient temperatures for a fast race time.
Introduction
The IRONMAN® (IM) triathlon covering 3.8 km swimming, 180 km cycling and 42.195 km running, is one of the most popular triathlon formats [1]. Since the first edition of IM Hawaii, different topics such as training [2], nutrition [3], pacing [4], predictive variables [5], age of peak performance [6], and the determination of the most predictive split discipline [7] led to scientific insights for athletes and coaches in regards to IM race preparation. Most professional and recreational age group IM triathletes want to compete in IM Hawaii as the IM World Championship [8]. Therefore, it would be of great interest for IM age group triathletes to know where the fastest IM racecourse is.
IM age group triathletes need to qualify in an IM race by ranking within their age group (in 5-year intervals). Each IM race offers several slots for each age group category. The goal of many IM age group triathletes is to achieve a new personal best time. For these athletes, the knowledge of the fastest IM racecourse is highly valuable. There are many possibilities to get a slot for IM Hawaii for age group triathletes. Of all the 19 potential possibilities, the first is to win the age group category in a full-distance IM triathlon race [https://blog.irace.ai/how-to-qualify-for-kona-ironman-world-championships]. Considering this information and the knowledge of each athlete’s strengths and weaknesses, such as technical skill in cycling, ability to adapt to different weather conditions, or the number of slots available to compete in IM Hawaii, they could decide in which race to participate to achieve their goal.
To date, there is no knowledge about the predictive factors of racecourses in the IM triathlon races. However, originating from the United States of America (USA) [9, 10] seemed to be a predictive factor for a fast IM race time. An analysis of the IM World Championship IM Hawaii and its qualifying races showed that in 2010, US-American triathletes dominated participation and performance in both IM Hawaii and its qualifiers [8]. A study analyzing 39,706 finishers from 124 countries and competing between 1985 and 2012 in IM Hawaii showed that most finishers originated from the USA [9]. However, another study investigating the pacing of 302,535 IM triathletes competing between 2002 and 2015 in 253 different IM races showed that Germans were the fastest, followed by Australians, Austrians, and Brazilians, where IM triathletes from the USA were not among the fastest [10]. The mentioned studies showed where the fastest IM triathletes originated from, but IM age group triathletes cannot extract from these studies where the fastest IM racecourses are located for their intention. Furthermore, the disparate findings regarding the dominance of US-American or German triathletes need more investigation.
Overall, age group triathletes form the largest group of athletes competing in an IM triathlon [11]. Over the years, IM age group triathletes have improved their performance in IM qualifiers [12] and in IM Hawaii itself [13]. Since IM age group triathletes form the largest group of athletes in any IM triathlon race and these athletes need to qualify for IM Hawaii, it might be of high interest for them to know where the fastest IM racecourses are, in order to choose, considering their strengths and weakness, the best IM race to participate.
Environmental conditions such as water [14] and air temperature [15] have a considerable influence on endurance performance. When different weather parameters (i.e. air temperature, humidity, wind speed, and solar load) were evaluated in running events of different distances, air temperature showed the highest influence where lower air temperatures of 10°C-17.5°C led to better performances [16]. In marathon running, performance progressively slowed down with increasing air temperature, especially for slower [17] and older [18] runners. Regarding triathlon, an actual study investigating the Olympic distance triathlon showed differences between the split disciplines and the sexes regarding the ambient temperatures. While cycling showed the highest influence on overall race time at low temperatures, running had the highest influence at high temperatures. Regarding the sex, female sex showed a higher influence on overall race time in cycling but less in running when swimming was performed in a river [19]. A study regarding the influence of temperatures on race performance is, however, missing for IM triathletes.
Therefore, the present study aimed to find the fastest racecourses for IM age group triathletes from all IM races held between 2002 and 2022 worldwide. With this knowledge, IM age group triathletes might be able to select a specific IM racecourse to qualify for IM Hawaii. Nonetheless, the optimal selection of a racecourse should be carried out carefully, considering the increased effort and associated financial cost of participation. Since originating from the USA seemed to be a predictive factor for a fast IM race times [9, 10], we hypothesized that most IM age group triathletes would (i) originate from the USA and (ii) would compete in IM races held in the USA. A second aim was to investigate a potential influence of water and air temperatures as well as course profiles on overall race time. We hoped to find (i) the ideal ambient water and air temperatures and (ii) the ideal course profiles in cycling and running to achieve a fast overall race time.
Methods
Declarations ethics approval
This study was approved by the Institutional Review Board of Kanton St. Gallen, Switzerland, with a waiver of the requirement for informed consent of the participants as the study involved the analysis of publicly available data (EKSG 01/06/2010). The study was conducted in accordance with recognized ethical standards according to the Declaration of Helsinki adopted in 1964 and revised in 2013.
Data set and data preparation
The race data was downloaded from the official IRONMAN® website (www.ironman.com) using a Python script (www.python.org). The athletes’ sex, age and country of origin, the event location and year, the partial times (i.e., swimming, running, cycling, and transition times) and the overall race times were thus obtained. We also included the average air and water temperature during these races as provided by the race websites. After extensive clean-up and data pre-processing, a total of 677,702 IM finishers’ records were available for analysis from IM triathletes originating from 228 different countries and competing at 444 races held between 2002 and 2022 in 66 different event IM race locations worldwide. The raw dataset included records of athletes that did not complete the race (DNS–did not start, DNF–did not finish, DQ–disqualified), which were discarded. Further processing of the dataset included harmonization of age group and gender information, checking for data coherence (meaning that the sum of the split times would match the overall race time, no negative or impossible times, etc.), the harmonization of countries of origin of the athletes (discarding records without such information, ensuring all country acronyms corresponded to actual countries, etc.), the harmonization of event locations (we found that some generic locations such as the “World Championship” or the “European Championship” had taken place in different locations over the years, so we did a mapping of logical to physical locations). Each race record corresponds to one triathlete´s participation in one race and includes the gender, age group, country of origin, race location, year of the race, etc. The number of records is the number of qualifying race results after discarding any records with bad or incomplete data. Race records are classified in 5-years age groups: 18–24, 25–29, 30–34, 35–39, 40–44, 45–49, 50–54, 55–59, 60–64, 65–69, 70–74, 75–79, 80–84 and 85–89 years following the general classification of age group triathletes in all triathlon races (www.triathlon.org/agegroup).
Data analysis
Descriptive information is presented using frequencies, mean, standard deviation, and minimum and maximum values. A XG Boost Regression model was designed to use the race finish time (FinishTime) as the predicted variable, whereas Gender, AgeGroup, Country EventLocation, and the Air and Water temperatures were used as predictors. XG Boost stands for Extreme Gradient Boosting and is an optimized decision tree-based ensemble algorithm that uses gradient boosting and parallelization to outperform other algorithms in execution speed and prediction accuracy (www.sciencedirect.com/topics/computer-science/extreme-gradient-boosting). Whilst the air and water temperature variables were originally numerical, the other four predictive variables were of a categorical type and had to be numerically encoded before they could be used to train the model. The variable Gender was encoded as 0 (females) or 1 (males). The AgeGroup variable was encoded using the first two digits of the age group (so 18–24 became 18, 25–29 became 25, etc.). The Country variable was ranked by participation, that is, by number of race records, and the resulting order was used as the encoding index (so US is index 0 with the largest number of records.) Last, the EventLocation variable was ranked by average race finish time, with the fastest locations first. The dataset was then split randomly into two subsets, and the model was trained with 75% of the available data and validated against the remaining 25%. Two evaluation metrics were calculated: the Mean Absolute Error (MAE) and R-squared (coefficient of determination or R2). Several models were built to explore the best hyper-parameters combination. The optimal model was finally built with 1000 estimators (decision trees) with max depth of 5, and was trained with a learning rate of 0.2. All other model hyper-parameters were left as default. Model generalization (i.e. the ability of a predictive model to make accurate prediction with previously unseen data examples) was not among our priorities, as we rather sought to use the model with statistical descriptive purposes. After training the model, we asked what it learnt from the training data in order to understand how each feature contributes to the model predictions. Several techniques were used to this end: the model SHAP values and features relative importances, and the PDP-based model interpretation charts. SHAP values (SHapley Additive exPlanations) are a model-agnostic technique with origins in the cooperative game theory, which is widely used to increase the transparency and interpretability of machine learning models. SHAP values indicate both a measure of the importance of each feature and whether each feature has a positive or negative impact on the predictions. These impacts can be calculated for each individual prediction, for the whole validation data sample (in their aggregated form), or even for the combined effects of two features. The Partial Dependence Plots (PDP) model interpretability charts add an additional level of detail, showing the model distribution of race time predictions against reference averages and group sizes. The predictions distributions are represented by means of boxplots. All analyses, modeling and visualizations were done using Python (www.python.org/) and the specific statistical and machine learning libraries using a Google Colab notebook (https://colab.research.google.com/). The main libraries used in the study are: numpy, pandas, scipy, statsmodels, matplotlib, seaborn, country_converter, sklearn, xgboost, pdpbox, shap.
Results
Most of the finishers were competing in IM triathlons held in the USA such as IM Wisconsin, IM Florida, IM Lake Placid, IM Arizona, or IM Hawaii, where the IM World Championship took place (Table 1). The fastest average overall race times were achieved in IM Hawaii, IM Austria, IM Copenhagen, IM Brazil Florianopolis, and IM Barcelona (Table 2).
Table 1. Full list of the 66 IM race locations ranked by participation (number of finishers records).
| Event Location | Number of Races | Number of Records | Number of Unique Athletes | |
|---|---|---|---|---|
| 0 | IRONMAN® Wisconsin | 19 | 38,545 | 26,986 |
| 1 | IRONMAN® Florida | 19 | 38,157 | 29,081 |
| 2 | IRONMAN® Lake Placid | 17 | 34,341 | 24,382 |
| 3 | IRONMAN® Arizona | 17 | 34,246 | 25,689 |
| 4 | IRONMAN® Hawaii | 18 | 32,156 | 21,080 |
| 5 | IRONMAN® Austria | 16 | 30,970 | 23,560 |
| 6 | IRONMAN® France | 16 | 29,302 | 24,200 |
| 7 | IRONMAN® Canada | 15 | 27,135 | 19,771 |
| 8 | IRONMAN® Coeur d’Alene | 15 | 24,526 | 19,575 |
| 9 | IRONMAN® Louisville | 10 | 21,661 | 17,907 |
| 10 | IRONMAN® Texas | 11 | 21,137 | 16,001 |
| 11 | IRONMAN® Frankfurt | 13 | 20,057 | 16,579 |
| 12 | IRONMAN® Cozumel | 13 | 17,902 | 14,470 |
| 13 | IRONMAN® Copenhagen | 8 | 16,739 | 13,344 |
| 14 | IRONMAN® New Zealand | 18 | 16,623 | 11,218 |
| 15 | IRONMAN® Mont-Tremblant | 8 | 16,306 | 12,951 |
| 16 | IRONMAN® UK | 11 | 15,221 | 12,066 |
| 17 | IRONMAN® Kalmar | 8 | 14,640 | 9,826 |
| 18 | IRONMAN® Lanzarote | 11 | 14,530 | 11,066 |
| 19 | IRONMAN® Brazil Florianopolis | 10 | 14,010 | 9,656 |
| 20 | IRONMAN® Australia ‐ New South Wales | 12 | 13,668 | 8,454 |
| 21 | IRONMAN® Zurich Switzerland | 9 | 13,199 | 11,502 |
| 22 | IRONMAN® Wales | 8 | 12,496 | 9,157 |
| 23 | IRONMAN® Barcelona | 6 | 11,137 | 10,132 |
| 24 | IRONMAN® Chattanooga | 6 | 10,788 | 9,605 |
| 25 | IRONMAN® Western Australia | 10 | 10,651 | 7,835 |
| 26 | IRONMAN® South Africa | 7 | 10,407 | 7,132 |
| 27 | IRONMAN® Boulder | 6 | 8,492 | 7,244 |
| 28 | IRONMAN® Maryland | 6 | 8,391 | 7,448 |
| 29 | IRONMAN® Cairns | 10 | 8,253 | 6,348 |
| 30 | IRONMAN® Emilia Romagna | 4 | 8,065 | 7,261 |
| 31 | IRONMAN® Vichy | 6 | 7,419 | 6,798 |
| 32 | IRONMAN® Hamburg | 4 | 6,654 | 6,249 |
| 33 | IRONMAN® Mallorca | 5 | 6,262 | 5,790 |
| 34 | IRONMAN® Malaysia | 8 | 5,409 | 4,255 |
| 35 | IRONMAN® Melbourne | 4 | 4,667 | 3,985 |
| 36 | IRONMAN® Santa Rosa | 3 | 4,655 | 4,275 |
| 37 | IRONMAN® Maastricht | 4 | 4,066 | 3,737 |
| 38 | IRONMAN® St. George | 3 | 3,705 | 3,394 |
| 39 | IRONMAN® Japan | 3 | 3,515 | 2,746 |
| 40 | IRONMAN® Tallinn | 4 | 3,333 | 3,009 |
| 41 | IRONMAN® Regensburg | 2 | 2,825 | 2,703 |
| 42 | IRONMAN® Los Cabos | 5 | 2,645 | 2,404 |
| 43 | IRONMAN® Gurye Korea | 2 | 2,609 | 2,251 |
| 44 | IRONMAN® Lake Tahoe | 2 | 2,605 | 2,504 |
| 45 | IRONMAN® Tulsa | 2 | 2,311 | 2,244 |
| 46 | IRONMAN® New York | 1 | 1,955 | 1,950 |
| 47 | IRONMAN® Taiwan | 4 | 1,940 | 1,739 |
| 48 | IRONMAN® Brazil Fortaleza | 3 | 1,755 | 1,484 |
| 49 | IRONMAN® Vineman | 1 | 1,699 | 1,696 |
| 50 | IRONMAN® Vitoria-Gasteiz | 1 | 1,572 | 1,568 |
| 51 | IRONMAN® Mar del Plata | 3 | 1,492 | 1,372 |
| 52 | IRONMAN® Portugal-Cascais | 1 | 1,390 | 1,384 |
| 53 | IRONMAN® Subic Bay Philippines | 2 | 1,262 | 1,172 |
| 54 | IRONMAN® Indiana | 1 | 1,223 | 1,223 |
| 55 | IRONMAN® Muskoka | 1 | 1,046 | 1,044 |
| 56 | IRONMAN® Switzerland Thun | 1 | 884 | 884 |
| 57 | IRONMAN® Des Moines | 1 | 843 | 842 |
| 58 | IRONMAN® Haugesund Norway | 2 | 825 | 779 |
| 59 | IRONMAN® Weymouth | 1 | 700 | 697 |
| 60 | IRONMAN® Waco | 1 | 639 | 639 |
| 61 | IRONMAN® Finland Kuopio-Tahko | 1 | 624 | 624 |
| 62 | IRONMAN® Gdynia | 1 | 590 | 588 |
| 63 | IRONMAN® China | 2 | 415 | 385 |
| 64 | IRONMAN® Pays D’Aix | 1 | 383 | 383 |
| 65 | IRONMAN® North Carolina | 1 | 34 | 34 |
Table 2. Top 25 IM race locations by average finish time.
Times are presented as hours:minutes:seconds.
| Event Location | Race finish time | |||||
|---|---|---|---|---|---|---|
| Records | mean | std | min | max | ||
| 0 | IRONMAN® Vitoria-Gasteiz | 1,572 | 11:35:17 | 01:21:41 | 08:46:18 | 15:42:38 |
| 1 | IRONMAN® Copenhagen | 16,739 | 11:41:04 | 01:22:43 | 08:19:14 | 16:06:57 |
| 2 | IRONMAN® Hawaii | 32,156 | 11:43:02 | 01:51:33 | 08:24:36 | 17:50:33 |
| 3 | IRONMAN® Tallinn | 3,333 | 11:43:59 | 01:36:07 | 08:27:45 | 16:59:37 |
| 4 | IRONMAN® Regensburg | 2,825 | 11:45:31 | 01:28:43 | 08:42:34 | 15:55:22 |
| 5 | IRONMAN® Barcelona | 11,137 | 11:46:41 | 01:25:56 | 08:22:24 | 16:01:29 |
| 6 | IRONMAN® Brazil Florianopolis | 14,010 | 11:48:13 | 01:31:03 | 08:36:29 | 17:32:34 |
| 7 | IRONMAN® Hamburg | 6,654 | 11:50:55 | 01:30:02 | 08:12:43 | 15:48:47 |
| 8 | IRONMAN® Mallorca | 6,262 | 11:54:14 | 01:34:26 | 08:16:48 | 16:23:56 |
| 9 | IRONMAN® Austria | 30,970 | 11:55:11 | 01:35:13 | 08:33:27 | 17:10:50 |
| 10 | IRONMAN® Switzerland Thun | 884 | 11:55:40 | 01:33:31 | 08:28:57 | 16:27:00 |
| 11 | IRONMAN® Emilia Romagna | 8,065 | 11:57:27 | 01:33:22 | 08:34:42 | 15:58:53 |
| 12 | IRONMAN® Finland Kuopio-Tahko | 624 | 11:58:28 | 01:25:19 | 09:09:01 | 15:49:35 |
| 13 | IRONMAN® Melbourne | 4,667 | 12:00:33 | 01:41:59 | 08:47:42 | 17:03:49 |
| 14 | IRONMAN® Frankfurt | 20,057 | 12:01:36 | 01:23:00 | 08:43:51 | 15:53:45 |
| 15 | IRONMAN® Kalmar | 14,640 | 12:04:54 | 01:27:55 | 08:29:15 | 16:15:18 |
| 16 | IRONMAN® Maastricht | 4,066 | 12:09:22 | 01:32:00 | 08:55:44 | 16:40:13 |
| 17 | IRONMAN® Mar del Plata | 1,492 | 12:09:40 | 01:36:57 | 08:46:41 | 16:54:51 |
| 18 | IRONMAN® Gdynia | 590 | 12:17:43 | 01:33:46 | 08:52:37 | 16:19:51 |
| 19 | IRONMAN® Portugal-Cascais | 1,390 | 12:18:33 | 01:34:26 | 08:56:28 | 15:57:07 |
| 20 | IRONMAN® Vichy | 7,419 | 12:18:59 | 01:37:07 | 08:23:49 | 16:55:55 |
| 21 | IRONMAN® Pays D’Aix | 383 | 12:20:15 | 01:34:31 | 08:47:05 | 15:58:45 |
| 22 | IRONMAN® Zurich Switzerland | 13,199 | 12:21:28 | 01:31:56 | 08:52:38 | 16:01:34 |
| 23 | IRONMAN® Haugesund Norway | 825 | 12:22:24 | 01:32:27 | 08:48:41 | 16:15:48 |
| 24 | IRONMAN® France | 29,302 | 12:23:18 | 01:31:59 | 08:20:08 | 16:50:25 |
Regarding the triathlete´s countries of origin, the vast majority of the successful finishers originated from the US, followed by athletes from the UK, Canada, Australia, Germany, and France (Table 3), but the best mean times were achieved by athletes from Austria, Germany, Belgium, Switzerland, Finland, and Denmark.
Table 3. Ranking of top 25 countries by number of records of IM finishers, accounting for over 94% of the full sample.
Times are presented as hours:minutes:seconds.
| Overall race time | |||||||
|---|---|---|---|---|---|---|---|
| Country | Number of Records | Number of Unique Athletes | mean | std | min | max | |
| 0 | USA | 27,4553 | 12,4646 | 13:21:38 | 01:45:05 | 08:17:24 | 17:57:34 |
| 1 | GBR | 55,410 | 29,852 | 13:02:11 | 01:39:05 | 08:31:52 | 17:04:40 |
| 2 | CAN | 38,264 | 18,074 | 13:01:58 | 01:43:45 | 08:23:48 | 17:50:43 |
| 3 | AUS | 37,571 | 16,836 | 12:15:32 | 01:40:10 | 08:39:10 | 17:03:49 |
| 4 | DEU | 32,662 | 16,240 | 11:44:13 | 01:30:45 | 08:12:43 | 17:35:15 |
| 5 | FRA | 27,873 | 16,293 | 12:06:37 | 01:32:35 | 08:19:20 | 17:02:02 |
| 6 | ESP | 17,009 | 9,788 | 12:04:09 | 01:35:02 | 08:06:57 | 16:57:55 |
| 7 | SWE | 14,640 | 7,057 | 12:05:09 | 01:28:52 | 08:30:55 | 17:50:33 |
| 8 | BRA | 14,408 | 7,911 | 11:51:22 | 01:35:39 | 08:29:48 | 17:32:34 |
| 9 | AUT | 12,792 | 6,267 | 11:40:39 | 01:29:42 | 08:37:13 | 16:56:30 |
| 10 | ITA | 12,702 | 6,236 | 12:05:08 | 01:28:46 | 08:35:54 | 16:50:26 |
| 11 | MEX | 11,941 | 6,211 | 13:13:53 | 01:38:57 | 08:36:19 | 17:05:38 |
| 12 | DNK | 11,874 | 6,469 | 11:35:16 | 01:23:42 | 08:22:31 | 17:07:48 |
| 13 | NZL | 11,092 | 5,538 | 12:41:41 | 01:45:16 | 08:03:41 | 17:38:06 |
| 14 | JPN | 10,249 | 4,990 | 13:27:16 | 01:49:16 | 08:20:05 | 17:00:31 |
| 15 | ZAF | 9,675 | 5,732 | 13:17:24 | 01:38:51 | 08:41:51 | 16:58:05 |
| 16 | BEL | 8,444 | 3,896 | 11:28:32 | 01:28:10 | 08:16:08 | 16:59:56 |
| 17 | CHE | 7,570 | 3,381 | 11:37:15 | 01:29:16 | 08:23:49 | 16:53:40 |
| 18 | IRL | 6,337 | 3,625 | 12:32:17 | 01:39:02 | 08:26:02 | 17:02:23 |
| 19 | ARG | 5,470 | 2,679 | 12:06:03 | 01:35:32 | 08:33:44 | 17:19:28 |
| 20 | NLD | 4,750 | 2,841 | 11:49:39 | 01:30:01 | 08:16:48 | 16:59:30 |
| 21 | FIN | 3,497 | 1,785 | 11:40:55 | 01:23:43 | 08:32:40 | 16:35:12 |
| 22 | POL | 3,438 | 2,005 | 12:00:30 | 01:31:23 | 08:46:49 | 16:44:55 |
| 23 | RUS | 3,163 | 1,953 | 11:54:43 | 01:30:19 | 08:22:24 | 17:03:00 |
| 24 | ISR | 2,952 | 1,884 | 12:44:26 | 01:35:25 | 08:40:21 | 16:53:06 |
Table 4 presents the percentage of unsuccessful finishers (DNF) in all IM races sorted in alphabetical order of the races. The highest percentage of non-finishers was recorded for IM Korea, followed by IM North Carolina and IM Taiwan.
Table 4. The percentage of non-finishers (DNF) for each IM race sorted in alphabetical order.
| Event Status | ||||
|---|---|---|---|---|
| EventLocation | DNF | Finish | Both | DNF rate |
| IRONMAN® Arizona | 68 | 34325 | 34393 | 0,20% |
| IRONMAN® Australia ‐ New South Wales | 11 | 13673 | 13684 | 0,08% |
| IRONMAN® Austria | 1 | 33261 | 33262 | 0,00% |
| IRONMAN® Barcelona | 2 | 11141 | 11143 | 0,02% |
| IRONMAN® Boulder | 38 | 8494 | 8532 | 0,45% |
| IRONMAN® Brazil Florianópolis | 10 | 14021 | 14031 | 0,07% |
| IRONMAN® Brazil Fortaleza | 4 | 1756 | 1760 | 0,23% |
| IRONMAN® Cairns | 49 | 9458 | 9507 | 0,52% |
| IRONMAN® Canada | 26 | 27146 | 27172 | 0,10% |
| IRONMAN® Chattanooga | 78 | 10799 | 10877 | 0,72% |
| IRONMAN® China | 0 | 415 | 415 | 0,00% |
| IRONMAN® Coeur d’Alene | 31 | 24542 | 24573 | 0,13% |
| IRONMAN® Copenhagen | 24 | 16751 | 16775 | 0,14% |
| IRONMAN® Cozumel | 87 | 17933 | 18020 | 0,48% |
| IRONMAN® Des Moines | 14 | 843 | 857 | 1,63% |
| IRONMAN® Emilia Romagna | 11 | 8078 | 8089 | 0,14% |
| IRONMAN® Finland Kuopio-Tahko | 0 | 624 | 624 | 0,00% |
| IRONMAN® Florida | 134 | 38204 | 38338 | 0,35% |
| IRONMAN® France | 7 | 29337 | 29344 | 0,02% |
| IRONMAN® Frankfurt | 1 | 23342 | 23343 | 0,00% |
| IRONMAN® Gdynia | 0 | 601 | 601 | 0,00% |
| IRONMAN® Gurye Korea | 81 | 2609 | 2690 | 3,01% |
| IRONMAN® Hamburg | 23 | 6668 | 6691 | 0,34% |
| IRONMAN® Haugesund Norway | 1 | 825 | 826 | 0,12% |
| IRONMAN® Hawaii | 40 | 32186 | 32226 | 0,12% |
| IRONMAN® Indiana | 17 | 1223 | 1240 | 1,37% |
| IRONMAN® Japan | 4 | 3516 | 3520 | 0,11% |
| IRONMAN® Kalmar | 0 | 14651 | 14651 | 0,00% |
| IRONMAN® Lake Placid | 93 | 34369 | 34462 | 0,27% |
| IRONMAN® Lake Tahoe | 2 | 2606 | 2608 | 0,08% |
| IRONMAN® Lanzarote | 5 | 14800 | 14805 | 0,03% |
| IRONMAN® Los Cabos | 0 | 2645 | 2645 | 0,00% |
| IRONMAN® Louisville | 50 | 21678 | 21728 | 0,23% |
| IRONMAN® Maastricht | 0 | 4068 | 4068 | 0,00% |
| IRONMAN® Malaysia | 79 | 5427 | 5506 | 1,43% |
| IRONMAN® Mallorca | 0 | 6265 | 6265 | 0,00% |
| IRONMAN® Mar del Plata | 2 | 1496 | 1498 | 0,13% |
| IRONMAN® Maryland | 10 | 8393 | 8403 | 0,12% |
| IRONMAN® Melbourne | 3 | 4669 | 4672 | 0,06% |
| IRONMAN® Mont-Tremblant | 21 | 16309 | 16330 | 0,13% |
| IRONMAN® Muskoka | 0 | 1047 | 1047 | 0,00% |
| IRONMAN® New York | 0 | 1955 | 1955 | 0,00% |
| IRONMAN® New Zealand | 10 | 16637 | 16647 | 0,06% |
| IRONMAN® North Carolina | 1 | 34 | 35 | 2,86% |
| IRONMAN® Pays D’Aix | 0 | 383 | 383 | 0,00% |
| IRONMAN® Portugal-Cascais | 11 | 1390 | 1401 | 0,79% |
| IRONMAN® Regensburg | 0 | 2826 | 2826 | 0,00% |
| IRONMAN® Santa Rosa | 19 | 4670 | 4689 | 0,41% |
| IRONMAN® South Africa | 24 | 11997 | 12021 | 0,20% |
| IRONMAN® St. George | 0 | 3708 | 3708 | 0,00% |
| IRONMAN® Subic Bay Philippines | 15 | 1262 | 1277 | 1,17% |
| IRONMAN® Switzerland Thun | 1 | 885 | 886 | 0,11% |
| IRONMAN® Taiwan | 49 | 2718 | 2767 | 1,77% |
| IRONMAN® Tallinn | 0 | 3333 | 3333 | 0,00% |
| IRONMAN® Texas | 82 | 21148 | 21230 | 0,39% |
| IRONMAN® Tulsa | 11 | 2312 | 2323 | 0,47% |
| IRONMAN® UK | 7 | 15224 | 15231 | 0,05% |
| IRONMAN® Vichy | 1 | 7457 | 7458 | 0,01% |
| IRONMAN® Vineman | 4 | 1699 | 1703 | 0,23% |
| IRONMAN® Vitoria-Gasteiz | 1 | 1572 | 1573 | 0,06% |
| IRONMAN® Waco | 3 | 639 | 642 | 0,47% |
| IRONMAN® Wales | 0 | 12505 | 12505 | 0,00% |
| IRONMAN® Western Australia | 28 | 10660 | 10688 | 0,26% |
| IRONMAN® Weymouth | 0 | 700 | 700 | 0,00% |
| IRONMAN® Wisconsin | 33 | 38568 | 38601 | 0,09% |
| IRONMAN® Zurich Switzerland | 7 | 13220 | 13227 | 0,05% |
| 1334 | 687696 | 689030 | ||
Regarding the predictive model performance, the XG Boost model produced an MAE of 4503.6 seconds (approx. 1h15min) over the 25% test sample. This metric, however, does not tell us much about the “goodness of fit”, that is, whether the model has strong or weak predictive power. The R2 score is a better metric to get an idea of the algorithm’s accuracy. The model produced an R2 score of 0.27, meaning that 27% of the predicted race finish time variability can be explained through the joint effect of the six predicting variables. R2-values of up to 0.25 or 0.3 are usually considered weak effects. Naturally, this model is just too simplistic to accurately represent reality, and the R2 result just indicates that there must be other variables with a stronger influence on an IM race finish time. But at 0.27, it is not irrelevant, and we can still use several model interpretability tools to extract some insights as to how each feature contributed to the model predictions.
Features relative importance (SHAP based)
Our XG Boost Regressor determined that Country (origin of the athlete) was the most important variable. However, the event location (where the race was held) wasthe feature with the second biggest relative importance (Fig 1).
Fig 1. XG Boost Regressor features importances (SHAP-based).
Aggregated SHAP values
Fig 2 shows the aggregated SHAP values of the XG Boost model. SHAP values present a quantitative view of the amount of change in the output (predicted FinishTime) for each input (predictor). The x-axis (SHAP value) represents seconds added to (or subtracted from) the reference line (set at zero). For the Country, the blue points concentrated in several clusters around zero and left and right of the reference line. Blue points correspond to low values of Country_ID. These countries have the largest number of participants, which are the USA (by a large margin) followed by the UK, Canada, Australia, Germany, France, Spain, Sweden, Brazil, etc., indicating a broad distribution of performances at all levels of competence. Red dots represent countries with marginal participation and accumulate on the left and right extremes of the chart. The EventLocation feature produces a chart where blue dots (low values of event location) accumulate on the left side, whilst increasing values of the location ID (violets and reds) accumulate towards the right. Given the rank encoding based on the average race finish time, blue points represent the fastest IM event locations (IM Hawaii, IM Copenhagen, IM Tallinn, IM Vitoria-Gasteiz, IM Regensburg, IM Barcelona, and IM Brazil Florianopolis. The AgeGroup feature gives the broadest range of change in the output (from around -3,000 to well over +12,000 seconds, about 4 hours). The red dots represent the older age groups making positive contributions to the FinishTime predictions whilst blue dots (low age groups, i.e., 18, 25, 30…) cluster on the negative side of the chart, confirming low age as a decisive predicting factor in deducting seconds from the race time. And in the last place, the Gender feature contributes to the prediction in an almost binary way, either adding seconds for females (value 0) or subtracting for males (value 1).
Fig 2. Aggregated SHAP values for the XG Boost predictive model.
Advanced model interpretability charts
These charts are built using the pdpbox and matplotlib libraries, to produce a customized view of our model. For each predicting variable (gender, age group, air temperature, etc.), a set of two charts is produced, describing the relationship between that predicting variable and the target variable (i.e. the race time in seconds). The lower chart is just a bar chart specifying the groups sizes (for instance, for the gender chart group, it represents the number of race records from male athletes and from female athletes). The upper chart is the most interesting and represents the model prediction distribution, for each possible value of the predicting variable visualized as box plots, i.e through the min, max, 25%, 50% and, 75% percentile values, and against the average group value in red. More men than women competed and men were faster than women (Fig 3). Athletes in age group 40–45 years were the most numerous, but athletes in age group 30–35 years were the fastest (Fig 4). Athletes from the USA were the most numerous, but athletes from Belgium, Denmark, Austria, Switzerland, Germany, and Finland were the fastest (Fig 5). The fastest overall race times were achieved in IM Hawaii (Fig 6). Regarding environmental conditions, the best race times were obtained an air temperature of ∼27°C (Fig 7) and a water temperature of ∼24°C (Fig 8).
Fig 3. Advanced model interpretability charts for gender.
Fig 4. Advanced model interpretability charts for age group.
Fig 5. Advanced model interpretability charts for country of origin of the athlete.
Fig 6. Advanced model interpretability charts for country of event location.
Fig 7. Advanced model interpretability charts for air temperature.
Fig 8. Advanced model interpretability charts for water temperature.
Discussion
This study intended to find the fastest racecourse for age group triathletes competing in IM triathlons by interpreting an XG Boost predictive model. Based on existing literature, we hypothesized that most IM age group triathletes would originate from the US and they would preferably compete in IM races held in the US. We found that (i) IM Hawaii was the fastest racecourse overall, (ii) most of the successful and most numerous IM finishers originated from the US, but athletes from Belgium, Denmark, Austria, Switzerland, Germany, and Finland were the fastest, (iii) athletes in age group 40–45 were the most numerous, but athletes in age group 30–35 years were the fastest and (iv) the best IM race times were achieved at an air temperature of ∼27°C and a water temperature of ∼24°C.
IM Hawaii is the fastest racecourse for IM age group triathletes
The first and most important finding was that the racecourse of IM Hawaii was the fastest course for most IM age group triathletes. It is well known that IM age group triathletes intending to compete in the IM World Championship IM Hawaii need to qualify [8]. For example, for 2023, IM age group triathletes have eight possibilities to qualify for the IM World Championship (www.ironman.com/news_article/show/1241591). The standard qualification competes in one of the 47 full-distance IM races in the qualifying series. A specific number of qualification slots are offered for each age group regarding the number of finishers per age group (www.ironman.com/im-world-championship-2023-slot). These characteristics guarantee the best IM athletes competing in this event, given that sustained performance and a high level of performance throughout the season are required. This qualification system works since only the best qualifiers can compete in IM Hawaii. Future studies need to consider mixed methods to understand the atmosphere behind this event, including motivational aspects and preparation characteristics.
US-American IM triathletes were the most numerous but not the fastest
A second important finding was that most of the successful IM finishers originated from the USA, followed by athletes from the UK, Canada, Australia, Germany, and France, but athletes from Belgium, Denmark, Austria, Switzerland, Germany, and Finland achieved the best mean times. These findings confirmed our first hypothesis that most IM age group triathletes would originate from the USA, but IM age group triathletes from the USA were not the fastest.
To date, three studies investigated the aspect of nationality and performance in an IM triathlon. One study analyzed participation and performance in the IM World Championship IM Hawaii and its qualifying races in 2010 showed that IM triathletes from the USA were both the most numerous and the fastest [8]. A study investigating a larger sample of finishers of IM Hawaii over a longer time frame confirmed that successful finishers at IM Hawaii were mainly US-American triathletes [8]. A study analyzing more than 300,000 IM triathletes showed that US-Americans were not the fastest at IM Hawaii, but IM triathletes originating from Germany, Australia, Austria, and Brazil [10]. In the present analysis with more than 670,000 finishers, IM triathletes from Austria, Germany, Belgium, Switzerland, Finland, and Denmark were the fastest, confirming recent findings investigating half of the sample and during an earlier time frame.
The finding that local athletes are not the fastest has already been shown in open-water long-distance swimmers. For example, in the ‘English Channel Swim’, where open-water swimmers cross the ‘English Channel’ between England and France, most swimmers were from Great Britain, USA, Australia and Ireland. However, swimmers from the USA and Australia achieved the fastest swim times before swimmers from Great Britain [20]. However, in crossing the ‘Strait of Gibraltar’, local Spanish swimmers were faster than US-American swimmers [21].
It is also important to note that Hawaii is within the territory of the USA but quite far from the mainland. For example, the time difference between New York and Hawaii is 6 hours. Furthermore, Hawaii has a tropical climate. A different time zone and a different climate might counteract the advantages of locals (i.e., US-Americans). Similar to data about the fastest racecourse, future studies need to consider different research approaches in order to expand our understanding of how the travel distance, the knowledge about the racecourse, and geographical characteristics affect athletes’ participation and performance.
Most of the finishers were competing in IM triathlons held in the USA
We could confirm our second hypothesis that most IM age group triathletes would compete in IM races held in the USA. Most of the successful finishers were competing in IM triathlons held in the USA, such as IM Wisconsin, IM Florida, IM Lake Placid, IM Arizona, and IM Hawaii, where the IM World Championship took place.
The fact that local athletes mainly compete in their home country has been described in different races such as the ‘Swiss Alpine Marathon’ in Switzerland where Swiss athletes were the most numerous and the best [22]. Similarly, in the ‘Powerman Duathlon World Championship’ held in Switzerland, duathletes from Switzerland were the most numerous and the fastest [23]. Also, Swiss triathletes represented the largest group of participants and were the fastest finishers in IM Switzerland held in Zurich, Switzerland [24].
An explanation of the increased number of local athletes in these races might be the advantages of competing at home in terms of budget, traveling and time management, as well as adaptation to weather conditions and no need to adapt to time zones. Furthermore, it should be noted that the above-mentioned races occur in countries where multi-sports are popular, which facilitates increased rates of participation of local athletes. The host effect was presented previously in the context of the Olympic Games [25]. Positive effects were shown for political, economic, cultural, environmental, tourism, and psychological aspects, as well as the number of medals in hosting countries’ [25]. In another way, little information is available about the role of hosting events in triathlon performance. Future studies need to consider investigating the relationship between the number of events, countries’ performance, and racecourse characteristics.
The fastest IM race times at the age of 30–35 years
A further important finding was that the fastest IM race times were achieved at the age of 30–35 years. The age of 35 years corresponds to the age of peak IM race performance, which has been described and confirmed in several studies investigating different samples of IM triathletes [6, 26, 27]. One study specifically investigated the age of peak triathlon performance in elite IM triathletes competing in IM Hawaii [28]. In the annual top ten women and men competing in the IM World Championship IM Hawaii between 1983 and 2012, the age of the annual top ten women and men increased over time to 35±5 and 34±3 years, respectively [28]. Lepers et al. suggested that the age-related decline in performance in the IM distance was larger than in the Olympic distance and attributed this observation to the longer duration of the former race [29].
The age of peak performance in the IM distance observed in the present study was older than what was shown by Malcata et al. in the Olympic distance triathlon [30], highlighting the role of the race distance. Although the purpose of this study was not to find the age of peak performance, our findings reinforce the interplay of different variables (i.e. sex, country, and event location) to the performance of amateur (i.e. recreational or age group) triathletes. Despite the fastest race time, as shown at the age of 35 years and younger, the IM Hawaii presented the higher contribution to athletes’ performance. However, the magnitude of this contribution was different for athletes of different ages. These initial results reinforce the complexity associated with endurance performance and suggest that future studies need to consider a holistic approach that considers the interplay between different variables and characteristics.
The aspect of environmental conditions
A last important finding was that the best IM race times were obtained at an air temperature of ∼27°C and at a water temperature of ∼24°C. It is well known that environmental conditions can have an impact on ultra-endurance performance [31–33]. A very recent study investigating Olympic distance triathletes competing in China between 2015 and 2021 showed that the specific split disciplines cycling and running had a different impact on overall race time depending upon the environmental temperatures [19]. Regarding swimming in a full-distance triathlon such as the ‘Norseman Xtreme Triathlon’, a water temperature of 14.4–16.4°C may lead to hypothermia even when wearing a wetsuit [34]. On the other hand, running in IM Hawaii at 22.44–28.50°C, high-level age group triathletes reached a state of hyperthermia during the marathon [35]. Both conditions of hypothermia in swimming and hyperthermia in running might have a considerable impact on overall race time in an IM triathlon. Most probably the air temperature of ∼27°C and the water temperature of ∼24°C were optimal temperatures for a fast race time in IM Hawaii.
World Triathlon has established specific rules regarding environmental temperatures (www.triathlon.org/about/downloads/category/competition_rules). Specifically for high ambient temperatures, prevention guidelines have been established for triathlon races (https://www.triathlon.org/uploads/docs/World_Triathlon_Guidelines_for_Exertional_Heat_Illness_Prevention_201902261.pdf). Also for running races, specific rules have been set (IIRM_Brochure_AW_Lo.pdf (emos1storage.blob.core.windows.net). Overall, high temperatures should be avoided for endurance athletes. For example, it was suggested that road-based marathon running races should not be started when wet bulb globe temperature (WBGT) at the start is > 21°C [36]. Another study defined upper thresholds for marathon to WBGT 24–27°C and for triathlon to WBGT 23–26°C [37]. Interestingly, these IM triathletes were the best at higher ambient temperatures than suggested in literature.
The question remains now whether the environmental conditions have an influence on race performance or not. IM age group athletes who intend to qualify for the IM World Championship in IM Hawaii compete in different races [8, 9, 11] with different environmental conditions as described. Most probably the selection of the best IM age group athletes leads to the fact that only the most experienced and fastest IM age group athletes can compete in IM Hawaii also evidenced by the very low non-finisher race of 0.12%. We assume that the best IM age group athletes are not affected by environmental conditions.
Limitations and implications for future studies
Although this analysis of a very large data set provides valuable insights for athletes and coaches regarding the selection of the fastest IM racecourse, some limitations should be acknowledged. The races provided only a general temperature for air and water temperatures and an overall description of the cycling and running course. Aspects such as food [3] and fluid [38] intake, hydration status [39] and thermoregulation [40], pacing during the race [41], elevations in the race course [41] and previous experience [2] were not considered. All these aspects might have a considerable influence on IM race performance. Future studies could benefit from using a qualitative approach to deeply understand the athletes’ perception of the racecourse characteristics and the motivation and budget associated with participating in events (i.e., training preparation and travel distances). Detailed information about weather (e.g. wind speed, humidity, solar radiation, altitude, and changes in elevation would be helpful for a detailed analysis. Future studies might incorporate detailed weather data from the specific regions using a high temporal resolution. Another important point is to deeply understand why low and middle-income countries are unrepresentative in participation and performance in IM race events. On the other hand, the study’s strength was its novelty, as it was the first to examine this topic using this methodological approach (the popular machine learning XG Boost algorithm). Our findings have practical information for triathletes in the context of selecting suitable races to compete in, especially considering the increased budget of participating and associated training workload.
Conclusion
In summary, IM Hawaii was the fastest racecourse overall for IM age group triathletes, most of the successful IM finishers originated from the US, but athletes from Belgium, Denmark, Austria, Switzerland, Germany, and Finland were the fastest, athletes in age group 40–45 were the most numerous, but athletes in age group 30–35 years were the fastest and the best race times were obtained an air temperature of ∼27°C and a water temperature of ∼24°C. The results of the present study are helpful for coaches and athletes to attend the most important events and fastest racecourses.
Supporting information
(CSV)
Data Availability
All relevant data are within the paper and its Supporting Information files
Funding Statement
The author(s) received no specific funding for this work.
References
- 1.Cushman DM, Dowling N, Ehn M, Kotler DH. Triathlon Considerations. Phys Med Rehabil Clin N Am. 2022;33(1):81–90. Epub 20211020. doi: 10.1016/j.pmr.2021.08.006 . [DOI] [PubMed] [Google Scholar]
- 2.Sinisgalli R, de Lira CAB, Vancini RL, Puccinelli PJG, Hill L, Knechtle B, et al. Impact of training volume and experience on amateur Ironman triathlon performance. Physiol Behav. 2021;232:113344. Epub 20210129. doi: 10.1016/j.physbeh.2021.113344 . [DOI] [PubMed] [Google Scholar]
- 3.Pfeiffer B, Stellingwerff T, Hodgson AB, Randell R, Pöttgen K, Res P, et al. Nutritional intake and gastrointestinal problems during competitive endurance events. Med Sci Sports Exerc. 2012;44(2):344–51. doi: 10.1249/MSS.0b013e31822dc809 . [DOI] [PubMed] [Google Scholar]
- 4.Knechtle B, Käch I, Rosemann T, Nikolaidis PT. The effect of sex, age and performance level on pacing of Ironman triathletes. Res Sports Med. 2019;27(1):99–111. Epub 20181111. doi: 10.1080/15438627.2018.1546703 . [DOI] [PubMed] [Google Scholar]
- 5.Johnson EC, Pryor JL, Casa DJ, Belval LN, Vance JS, DeMartini JK, et al. Bike and run pacing on downhill segments predict Ironman triathlon relative success. J Sci Med Sport. 2015;18(1):82–7. Epub 20131221. doi: 10.1016/j.jsams.2013.12.001 . [DOI] [PubMed] [Google Scholar]
- 6.Rüst CA, Knechtle B, Rosemann T, Lepers R. Sex difference in race performance and age of peak performance in the Ironman Triathlon World Championship from 1983 to 2012. Extrem Physiol Med. 2012;1(1):15. Epub 20121214. doi: 10.1186/2046-7648-1-15 ; PubMed Central PMCID: PMC3710130. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Sousa CV, Aguiar S, Olher RR, Cunha R, Nikolaidis PT, Villiger E, et al. What Is the Best Discipline to Predict Overall Triathlon Performance? An Analysis of Sprint, Olympic, Ironman(®) 70.3, and Ironman(®) 140.6. Front Physiol. 2021;12:654552. Epub 20210505. doi: 10.3389/fphys.2021.654552 ; PubMed Central PMCID: PMC8131838. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Stiefel M, Knechtle B, Rüst C, Rosemann T. Analysis of performances at the ‘Ironman Hawaii triathlon’ and its qualifier events with respect to nationality. Journal of Science and Cycling. 2023;2(2):27–34. [Google Scholar]
- 9.Dähler P, Rüst CA, Rosemann T, Lepers R, Knechtle B. Nation related participation and performance trends in ‘Ironman Hawaii’ from 1985 to 2012. BMC Sports Science, Medicine and Rehabilitation. 2014;6(1):16. doi: 10.1186/2052-1847-6-16 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Nikolaidis PT, Kach I, Rosemann T, Knechtle B. The Role of Nationality on the Pacing of Ironman Triathletes. Asian J Sports Med. 2017;8(4):e57130. Epub 2017-10-02. doi: 10.5812/asjsm.57130 [DOI] [Google Scholar]
- 11.Stiefel M, Rüst CA, Rosemann T, Knechtle B. A comparison of participation and performance in age-group finishers competing in and qualifying for Ironman Hawaii. Int J Gen Med. 2013;6:67–77. Epub 20130222. doi: 10.2147/IJGM.S40202 ; PubMed Central PMCID: PMC3582317. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Stiefel M, Knechtle B, Lepers R. Master triathletes have not reached limits in their Ironman triathlon performance. Scand J Med Sci Sports. 2014;24(1):89–97. Epub 20120514. doi: 10.1111/j.1600-0838.2012.01473.x . [DOI] [PubMed] [Google Scholar]
- 13.Lepers R, Rüst CA, Stapley PJ, Knechtle B. Relative improvements in endurance performance with age: evidence from 25 years of Hawaii Ironman racing. Age (Dordr). 2013;35(3):953–62. Epub 20120226. doi: 10.1007/s11357-012-9392-z ; PubMed Central PMCID: PMC3636391. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Dressendorfer R. Triathlon swim deaths. Curr Sports Med Rep. 2015. May-Jun;14(3):151–2. doi: 10.1249/JSR.0000000000000142 [DOI] [PubMed] [Google Scholar]
- 15.Del Coso J, González C, Abian-Vicen J, Salinero Martín JJ, Soriano L, Areces F, Ruiz D, Gallo C, Lara B, Calleja-González J. Relationship between physiological parameters and performance during a half-ironman triathlon in the heat. J Sports Sci. 2014;32(18):1680–7. doi: 10.1080/02640414.2014.915425 [DOI] [PubMed] [Google Scholar]
- 16.Mantzios K, Ioannou LG, Panagiotaki Z, Ziaka S, Périard JD, Racinais S, Nybo L, Flouris AD. Effects of Weather Parameters on Endurance Running Performance: Discipline-specific Analysis of 1258 Races. Med Sci Sports Exerc. 2022. Jan 1;54(1):153–161. doi: 10.1249/MSS.0000000000002769 ; PMCID: PMC8677617. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Ely MR, Cheuvront SN, Roberts WO, Montain SJ. Impact of weather on marathon-running performance. Med Sci Sports Exerc. 2007. Mar;39(3):487–93. doi: 10.1249/mss.0b013e31802d3aba . [DOI] [PubMed] [Google Scholar]
- 18.Weiss K, Valero D, Villiger E, Scheer V, Thuany M, Cuk I, Rosemann T, Knechtle B. The Influence of Environmental Conditions on Pacing in Age Group Marathoners Competing in the "New York City Marathon". Front Physiol. 2022. Jun 14;13:842935. doi: 10.3389/fphys.2022.842935 ; PMCID: PMC9237513. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Zhao J, Wu Y, Zhang J. A study of triathletes’ race strategies in different competition environments. Heliyon. 2024. Apr 9;10(8):e29454. doi: 10.1016/j.heliyon.2024.e29454 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Knechtle B, Rosemann T, Rüst CA. Participation and performance trends by nationality in the ’English Channel Swim’ from 1875 to 2013. BMC Sports Sci Med Rehabil. 2014;6:34. Epub 20140827. doi: 10.1186/2052-1847-6-34 ; PubMed Central PMCID: PMC4160141. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Nikolaidis PT, Sousa CV, Knechtle B. The relationship of wearing a wetsuit in long-distance open-water swimming with sex, age, calendar year, performance, and nationality ‐ crossing the "Strait of Gibraltar". Open Access J Sports Med. 2018;9:27–36. Epub 20180221. doi: 10.2147/OAJSM.S158502 ; PubMed Central PMCID: PMC5825996. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Rosemann T, Lepers R, Rust CA, Knechtle B, Eichenberger E, Onywera V. The aspect of nationality and performance in a mountain ultra-marathon-the ‘Swiss Alpine Marathon’. J Hum Sport Exerc. 2012;7(4):748–62. [Google Scholar]
- 23.Rüst CA, Knechtle B, Knechtle P, Rosemann T, Lepers R. The aspect of nationality in participation and performance at the ‘Powerman Duathlon World Championship’–The ‘Powerman Zofingen’ from 2002 to 2011. J Sci Cycling. 2012;2:33–9. [Google Scholar]
- 24.Jürgens D, Knechtle B, Rüst CA, Knechtle P, Rosemann T, Lepers R. An analysis of participation and performance by nationality at ‘Ironman Switzerland’ from 1995 to 2011. J Sci Cycling. 2012;1(2):10–20. [Google Scholar]
- 25.Asgari B, Khorshidi R. The effect of hosting on performance of host countries in summer and winter Olympic Games. European Journal of Sports and Exercise Science. 2013;2:1–6. [Google Scholar]
- 26.Rüst CA, Knechtle B, Knechtle P, Rosemann T, Lepers R. Age of peak performance in elite male and female Ironman triathletes competing in Ironman Switzerland, a qualifier for the Ironman world championship, Ironman Hawaii, from 1995 to 2011. Open Access J Sports Med. 2012;3:175–82. Epub 20121102. doi: 10.2147/OAJSM.S37115 ; PubMed Central PMCID: PMC3781912. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Stiefel M, Knechtle B, Rüst CA, Rosemann T, Lepers R. The age of peak performance in Ironman triathlon: a cross-sectional and longitudinal data analysis. Extreme Physiology & Medicine. 2013;2(1):27. doi: 10.1186/2046-7648-2-27 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Gallmann D, Knechtle B, Rüst CA, Rosemann T, Lepers R. Elite triathletes in ’Ironman Hawaii’ get older but faster. Age (Dordr). 2014;36(1):407–16. Epub 20130417. doi: 10.1007/s11357-013-9534-y ; PubMed Central PMCID: PMC3889914. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Lepers R, Knechtle B, Stapley PJ. Trends in Triathlon Performance: Effects of Sex and Age. Sports Med. 2013;43(9):851–63. doi: 10.1007/s40279-013-0067-4 . [DOI] [PubMed] [Google Scholar]
- 30.Malcata RM, Hopkins WG, Pearson SN. Tracking career performance of successful triathletes. Med Sci Sports Exerc. 2014;46(6):1227–34. doi: 10.1249/MSS.0000000000000221 . [DOI] [PubMed] [Google Scholar]
- 31.Burdon CA, Johnson NA, Chapman PG, Munir Che Muhamed A, O’Connor HT. Case study: beverage temperature at aid stations in ironman triathlon. Int J Sport Nutr Exerc Metab. 2013;23(4):418–24. Epub 20130107. doi: 10.1123/ijsnem.23.4.418 . [DOI] [PubMed] [Google Scholar]
- 32.Stearns RL, Nolan JK, Huggins RA, Maresh CM, Munõz CX, Pagnotta KD, et al. Influence of cold-water immersion on recovery of elite triathletes following the ironman world championship. J Sci Med Sport. 2018;21(8):846–51. Epub 20180131. doi: 10.1016/j.jsams.2017.12.011 . [DOI] [PubMed] [Google Scholar]
- 33.Moehrle M. Ultraviolet exposure in the Ironman triathlon. Med Sci Sports Exerc. 2001;33(8):1385–6. doi: 10.1097/00005768-200108000-00021 . [DOI] [PubMed] [Google Scholar]
- 34.Høiseth LØ, Melau J, Bonnevie-Svendsen M, Nyborg C, Eijsvogels TMH, Hisdal J. Core Temperature during Cold-Water Triathlon Swimming. Sports (Basel). 2021. Jun 20;9(6):87. doi: 10.3390/sports9060087 ; PMCID: PMC8234067. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Olcina G, Crespo C, Timón R, Mjaanes JM, Calleja-González J. Core Temperature Response During the Marathon Portion of the Ironman World Championship (Kona-Hawaii). Front Physiol. 2019. Dec 3;10:1469. doi: 10.3389/fphys.2019.01469 ; PMCID: PMC6901617. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Roberts WO. Determining a "do not start" temperature for a marathon on the basis of adverse outcomes. Med Sci Sports Exerc. 2010. Feb;42(2):226–32. doi: 10.1249/MSS.0b013e3181b1cdcf [DOI] [PubMed] [Google Scholar]
- 37.Oyama T, Fujii M, Nakajima K, Takakura J, Hijioka Y. Validation of upper thermal thresholds for outdoor sports using thermal physiology modelling. Temperature (Austin). 2023. May 14;11(1):92–106. doi: 10.1080/23328940.2023.2210477 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Meyer M, Knechtle B, Bürge J, Knechtle P, Mrazek C, Wirth A, et al. Ad libitum fluid intake leads to no leg swelling in male Ironman triathletes ‐ an observational field study. J Int Soc Sports Nutr. 2012;9(1):40. Epub 20120901. doi: 10.1186/1550-2783-9-40 ; PubMed Central PMCID: PMC3524467. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Sousa CV, Aguiar SDS, Olher RDR, Sales MM, de Moraes MR, Nikolaidis PT, et al. Hydration Status After an Ironman Triathlon: A Meta-Analysis. J Hum Kinet. 2019;70:93–102. Epub 20191130. doi: 10.2478/hukin-2018-0096 ; PubMed Central PMCID: PMC6942476. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Baillot M, Hue O. Hydration and thermoregulation during a half-ironman performed in tropical climate. J Sports Sci Med. 2015;14(2):263–8. Epub 20150508. ; PubMed Central PMCID: PMC4424453. [PMC free article] [PubMed] [Google Scholar]
- 41.Pryor JL, Adams WM, Huggins RA, Belval LN, Pryor RR, Casa DJ. Pacing Strategy of a Full Ironman Overall Female Winner on a Course With Major Elevation Changes. J Strength Cond Res. 2018;32(11):3080–7. doi: 10.1519/JSC.0000000000002807 . [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
(CSV)
Data Availability Statement
All relevant data are within the paper and its Supporting Information files