Abstract
Each season, coaches must decide which competitions to enter, how many to schedule, and how to sequence them; however, competition programme planning remains among the least evidence-based areas of coaching practice. This study examined how the number of races and days to a season-fastest swim vary by age, competition level and competition format. An observational longitudinal design was used, analysing performance data from 116 elite and sub-elite Irish swimmers (72 male, 44 female) across 23 seasons (2011–2024), comprising 31,136 race-level performances and 11,288 season-events. Mixed effects two-part models (zero-inflated negative binomial for days-to-fastest; hurdle model for races-to-fastest) separated the probability of an immediate season-fastest from subsequent progression. Older swimmers required fewer races and fewer days to achieve their season-fastest swim, with the steepest developmental changes occurring from late childhood to mid adolescence. Long course seasons were characterised by later season-fastest performances in calendar days but required fewer race attempts than short course seasons. Seasons including National or International competitions were associated with 91% lower likelihood of achieving a season-fastest swim at the first opportunity (OR = 0.09), while season-fastest swims were over four times more likely to occur in multi-swim competitions (OR = 4.26). When heats and finals were available, approximately 77% of season-fastest swims occurred in the final. The timing and location of season-fastest performances are shaped by developmental stage, competition level, and competition structure, rather than occurring randomly across the season. Coaches working with younger swimmers should plan for a longer competition window and interpret early-season performances with caution.
Introduction
Coaches make competition programme decisions every season, such as how many competitions to enter, which events to target, and how to sequence races in the build-up to a major championship. However, these decisions remain among the least evidence-based in swimming practice. Despite the ever-growing body of research on training load, periodisation, and performance progression, the competition programme itself has received remarkably little direct research attention. 1 This is particularly consequential in swimming, as the planning of an effective competition programme can be complicated by swimmers often competing across multiple strokes, distances, and rounds, sometimes within a single swim or multi-swim/day competitions 2 and a competition calendar that is largely fixed by external requirements from national federations, swimming structures and club commitments. As such, these factors may constrain rather than enable deliberate programme design. Competition serves functions beyond just performance measurement. It can provide swimmers with exposure to race-day conditions, pacing demands, emotional management, support fundamental skill development and acquisition in early years of their career, 3 as a structured programme targeted towards qualification and performance delivery 4 and as a way to develop competition routines under pressure that are difficult to replicate in training.5,6
McGibbon et al. 6 is one of the few studies to examine coaches’ approaches to competition programme structure, albeit through the lens of pacing development. They found considerable variation in practice, ranging from racing every 3–6 weeks to only 2–3 competitions ahead of a major championship, with some coaches advocating intensive weekly race blocks. Hellard et al. 2 reported that elite swimmers competed in 7 ± 4 competitions per season on average, a variation that itself signals the absence of a consensus approach. Stone et al. 7 similarly found that coaches distinguished between competitions used as training stimuli and those warranting a full taper, drawing on deep practical experience to do so. Combining that experience with evidence-based quantitative analyses of competition performance has the potential to further inform the management of competition scheduling in high-performance swimming.
Swimmers and coaches attempt to plan seasons around two linked uncertainties: when form typically peaks and how many race opportunities are needed before a season-fastest swim is likely to emerge. Most evidence appears to be descriptive, often conflating calendar time with opportunity (a season-fastest may appear “early” simply because there were few chances to race), and rarely tracing how these patterns evolve across development within the same swimmer. Thus, it remains poorly understood how competition programmes could be structured across a swimmer's career, how many competitions are needed to reach their season's fastest, over what time frame this occurs and if (and how) these patterns change as a swimmer develops.1,8
Johns et al. 1 began to address this gap by quantifying the number of days and races from the first swim in a season to the season's fastest swim across strokes, distances, course format, and sex, using performance data from elite and sub-elite Irish swimmers over eight seasons (2012–2020). The analysis showed that season-fastest swims are frequently achieved early and that subgroup differences are most pronounced by stroke and distance. This provided an important baseline description of season's fastest swim timing, but it left three practical questions unresolved, questions that matter precisely because coaching decisions are developmental and context-dependent. Firstly, does the timing and attempt requirements for reaching a season-fastest change systematically with age, such that “early fast” means something different for younger vs. older swimmers. Secondly, does exposure to higher-level competition shift these trajectories beyond age and event structure, altering the likelihood of immediate season-fastest swims and the expected progression when the season's fastest performance is not achieved immediately. Thirdly, how does competition format (i.e., multi-swim competitions vs. single swim competitions) relate to where season-fastest swims occur. In attempting to answer these questions, this research sought to contribute to a more nuanced understanding of the structure of a swimmer's competition programme.
The present study addresses these questions using an extended version of the previously used dataset, 1 now spanning 2011–2024 within a modelling framework that separates calendar timing from opportunity. Two complementary outcomes were analysed to allow developmental profiles to be estimated directly, while testing how those profiles are modified by course format (i.e., short course (SC) vs. long course (LC)), competition exposure, and competition structure. The study provides a practical, developmentally grounded account of when a season-fastest swim is likely to emerge, how many chances are typically needed, and how competitive context shifts both expectations and interpretation of those patterns.
Methods
Participants
Participants were elite and sub-elite Irish swimmers, who were part of the National Squad programme, between 2011 and 2024 and whose performances were recorded in the national competition results database across 23 seasons. The raw dataset comprised 31,136 race-level performances from 116 unique swimmers (72 male, 44 female), collected across 531 competitions (578 competition-days). The participants had an overall average age of 21.4 (SD ± 4.2) years at data collection. For season-level analyses, races were aggregated to season-events defined as unique combinations of swimmer × season × stroke × distance (11,341 season-events). After excluding 25 m and 75 m events (i.e., non-regulation distances), the analytic season-event sample comprised 11,288 season-events, distributed across distances as follows: 50 m (n = 2672), 100 m (n = 3978), 200 m (n = 3210), 400 m (n = 1037), 800 m (n = 233), and 1500 m (n = 158).
Data extraction
All data included in the study were provided by Swim Ireland. Using an observational design, race results were extracted at the individual-performance level, including swimmer identifier (ID), age, sex, date of competition, stroke, distance, course format (SC vs. LC), competition name (and round where available), and performance time. The SC season was defined as races occurring between October and March (inclusive), with the LC season being competitions occurring between April and September (inclusive); races outside these windows were omitted. Within each season-event, the season start was defined by the swimmer's first recorded race date in that window (rather than a fixed calendar date). Where competitions involved multiple rounds (i.e., heats, semi-finals and finals), each swim was treated as a separate race opportunity and ordered as completed.
Design and unit of analysis
The unit of analysis was a swimmer-season-event (e.g., 50 m freestyle within a given season window), defined as all recorded races for that swimmer and event within said season. For each season-event, the first race date, the season-fastest race (minimum time), the number of race opportunities (n_races_season), and the number of distinct competition days on which the swimmer contested that event, were identified.
Outcomes
Days-to-fastest was defined as the number of days between the first race date in the season-event and the date of the season-fastest race (days_to_fastest ≥ 0). Races-to-fastest was defined as the within-season-event index of the race in which the season-fastest occurred (races_to_fastest ≥ 1). Given that “fastest at the first opportunity” is substantively meaningful, but can also be structural (e.g., when n_races_season = 1), each outcome was deconstructed into (i) whether the season-fastest occurred at the first opportunity (day 0 / race 1) and (ii) conditional on not occurring at the first opportunity, the additional days or races required to reach the season-fastest.
Predictors
The primary predictors were age (modelled nonlinearly), course format (SC vs. LC), and competition exposure. Competition names were cleaned and classified into competition levels (Club/Local, Regional, National and International) using rule-based string matching with manual override via a lookup table; exposure variables were then aggregated at the swimmer × season × stroke × distance level. The main text reports a binary exposure (any_high: any National/International racing in the season-event), with proportion- and count-based alternatives reported in the Supplement. Stroke, distance, and sex were included as covariates to adjust for event structure.
Statistical analysis
The analysis began by summarising the data using five-number summaries and mean (SD). Given repeated season-events within swimmers and zero-heavy outcomes (many season-events achieved the fastest at the first opportunity), mixed-effects two-part count models were used. Days-to-fastest was modelled using a zero-inflated negative binomial mixed model with a spline age term and swimmer random intercept. The conditional component estimated expected days-to-fastest among non-day-0 cases, while the zero-inflation component estimated the probability of a day-0 fastest. Races-to-fastest were modelled using a hurdle-at-1 strategy on season-events with at least two race opportunities (n_races_season ≥2). The first component was a mixed-effects logistic model of fastest-on-first-race; the second component modelled the number of additional races required given not-first (extra_races = races_to_fastest - 1, restricted to >0) using a zero-truncated count model. Both hurdle components adjusted for opportunity via log(n_races_season) and included swimmer random intercepts to account for repeated observations within individuals.
Estimation and presentation
Models were fitted in R using glmmTMB9,10 with spline terms using the base R splines package. 11 Continuous predictors for which nonlinear effects were expected were modelled using natural cubic splines, with degrees of freedom chosen a priori to allow smooth nonlinear trends while avoiding highly flexible, overfitted curves. Results are presented as marginal predicted curves with 95% confidence intervals, computed on the population-average scale (random effects set to zero). Coefficient summaries are reported as incidence-rate ratios/odds ratios with Wald 95% confidence intervals. For the hurdle models, combined predictions for the overall expected number of races to fastest swim were obtained by propagating fixed-effect uncertainty through both model components. For the figures presented in the results section, all predictions are shown for a female swimmer specialising in backstroke with race opportunity held at the median, unless otherwise stated. This profile was used to keep figures comparable across analyses; the substantive contrasts are based on model-predicted differences rather than on this profile alone.
Results
Descriptive statistics
The included sample comprised 11,288 season-events from 116 swimmers (72 male, 44 female) across 23 seasons. SC seasons contributed 6242 season-events and LC seasons contributed 5046. Table 1 presents the distribution of season-events by stroke and distance. Freestyle events were most common (n = 4310 season-events across 114 swimmers), followed by backstroke (n = 1986), butterfly (n = 1821), individual medley (n = 1623), and breaststroke (n = 1539). The 100 m and 200 m distances accounted for the largest share (35.2% and 28.4%, respectively), with 800 m (2.1%) and 1500 m (1.4%) representing the smallest proportion of the dataset. Swimmers had a median of two race opportunities per season-event (M = 2.7); 32.6% of season-events contained only a single race.
Season-Event counts by stroke, distance, and course format.
Note. SC = short course; LC = long course; IM = individual medley. Swimmers = number of unique swimmers contributing season-events in that event. Counts represent season-events (unique swimmer × season × stroke × distance combinations). Only internationally recognised SC and LC events are included.
Across all season-events, the season-fastest performance occurred at the first opportunity in 61.7% of cases (days-to-fastest median = 0, M = 23.1; races-to-fastest median = 1, M = 1.8). The distributions of both outcomes were heavily right-skewed, with most season-events concentrated at the earliest values and a long tail extending to 175 days and 13 races. This concentration at the first opportunity reflects both structural cases where only one race was available, as well as genuinely early season-fastest performances. This is the structural feature that motivated the two-part modelling approach described below.
Table 2 presents outcome descriptives by age band and course format, alongside the proportion of season-events with high-level competition exposure. Several patterns are notable prior to modelling. First, the raw proportions of fastest-first season-events were relatively flat across age bands (range: 55.9%–64.6%). This suggests that age effects on timing are more nuanced than a simple shift in proportions; an observation the models address by separating the probability of an immediate fastest from the expected additional attempts when it does not occur at first attempt. Second, LC seasons showed wider interquartile ranges for days-to-fastest at every age band (e.g., IQR = 0–75 at ages 14–15 in LC vs. 0–34 in SC), reflecting the longer calendar structure. Third, the proportion of season-events with any National or International exposure increased sharply with age, from 36.2% of SC season-events for swimmers under 14 to over 80% from age 16 onward. For LC it was from 59.6% to over 95%. Therefore, competition exposure and age are closely linked in this sample, a confound the models address by including both as predictors.
Days-to-fastest and races-to-fastest by age band, course format, and competition exposure.
Mdn=Median.
Age and season fastest swim
As swimmers develop, they reach their season-fastest performance earlier in the season and with fewer race attempts. This pattern follows a nonlinear trajectory. The most pronounced improvements occur from late childhood through mid-adolescence, after which the curves plateau toward adulthood.
Probability of swimming fastest at the first race
The probability that a swimmer's season-fastest occurs at the very first race increased with age (p < .001). Older swimmers were progressively more likely to “arrive ready” at their first competition. For younger swimmers, particularly those in late childhood and early adolescence, the first race of a season was less likely to be their fastest. Coaches should expect that it may take several race opportunities before these swimmers reach their season-fastest. Figure 1 presents the probabilities of swimming the season-fastest time on the first try. A swimmer's first LC race is roughly three times more likely to be their fastest of the season than their first SC race. This is consistent with LC seasons typically beginning later in the preparation cycle, closer to competition readiness.
Model-based age curves for the probability of achieving the season-fastest on the first race, by course format and distance. Panels show the population-average predicted probability that the season-fastest performance occurs on the first race of a season-event, plotted against age and faceted by distance (50–1500 m). Predictions are from the first component of the two-part hurdle model (mixed-effects logistic regression), adjusting for sex, stroke, and race opportunity (held at the sample median) and accounting for repeated observations within swimmers via a random intercept. Predictions are shown for a female backstroke swimmer. Red lines denote short-course (SC) and blue lines long-course (LC) seasons; shaded ribbons indicate 95% confidence intervals.
Males were less likely than females to swim their fastest at the first race (21% lower odds, OR = 0.79, p = .044). Across strokes, freestyle swimmers were the least likely to peak at the first attempt relative to backstroke (28% lower odds, OR = 0.72, p < .001). Race distance had a strong and progressive effect: compared with 50 m events, the probability of producing the season-fastest at the first race declined steadily – 23% lower at 200 m (OR = 0.77, p = .002), 45% lower at 400 m (OR = 0.55, p < .001), 84% lower at 800 m (OR = 0.16, p < .001), and 86% lower at 1500 m (OR = 0.14, p < .001).
Additional races needed when the fastest is not first
When the season-fastest was not achieved at the first race, a second question becomes relevant: how many more race attempts are typically required? With increasing age, fewer additional races were needed (p ≤ .002), though this effect was more modest than the fastest-first probability. Males required more additional races than females (10% more, IRR=1.10, p < .001), and butterfly swimmers required more additional races than backstroke swimmers (11% more, IRR=1.11, p = .018). Notably, once the season-fastest was not first, distance had no reliable effect on how many extra races were needed (all p > .23). This means that the extra-race burden was similar whether a swimmer was competing in a 100 m or a 1500 m event. The key difference across distances lies in the probability of achieving the fastest at the first attempt, not in the progression thereafter.
Figure 2 presents the combined expected races-to-fastest across ages, integrating both components: the probability of being fastest-first and the expected additional races when not. The curves decline with age and separate by course format, with LC seasons requiring fewer total races. This is driven primarily by the higher fastest-first probability in LC.
Model-based age curves for expected additional races to season-fastest when the fastest was not achieved on the first race, by course format and distance. Panels show the population-average expected number of additional races required to reach the season-fastest, conditional on the season-fastest not occurring at the first race, plotted against age and faceted by distance (50–1500 m). Predictions are from the second component of the two-part hurdle model (zero-truncated Poisson with mixed effects), adjusting for sex, stroke, and race opportunity and accounting for repeated observations within swimmers via a random intercept. Predictions are shown for a female backstroke swimmer. Red lines denote short-course (SC) and blue lines long-course (LC) seasons; shaded ribbons indicate 95% confidence intervals. Note the y-axis is restricted to 1.0–2.8 to show the modest separation between course formats; the near-overlap of SC and LC curves indicates that course format primarily affects whether the fastest occurs first (Figure 1A-(i)), not how many additional races are needed when it does not.
Days to fastest swim
LC seasons were associated with substantially longer time-to-fastest in calendar days than SC seasons, controlling for age, stroke, distance, and swimmer-level clustering (65% longer, IRR=1.65, p < .001). This is expected given the longer calendar window of the LC season (April-September vs. October-March) and does not imply that LC swimmers are slower to find form. As the races analysis above shows, they require fewer race attempts. Males showed longer days-to-fastest than females (11% longer, IRR=1.11, p = .002), and breaststroke swimmers took marginally longer than backstroke swimmers to reach their season-fastest (7% longer, IRR=1.07, p = .047).
Differences by distance in the time taken to reach the season-fastest were generally small: 100 m events took longer than 50 m events (9% longer, IRR=1.09, p < .001), while longer distances showed no reliable additional differences in expected days (all p > .19). The probability of achieving the season-fastest on the very first day of racing that event also varied by distance: compared with 50 m, swimmers in 100 m and 200 m events were less likely to peak immediately (25% lower odds, OR = 0.75, p < .001 and 18% lower odds, OR = 0.83, p = .002, respectively), while age had only a small effect on this probability (Figure 3).
Model-based age curves for days to fastest by format and distance. Panels show the population-average expected number of calendar days from the first race in a season-event to the season-fastest performance (days_to_fastest), plotted as a function of age and faceted by race distance (50–1500 m). Curves are predicted from a zero-inflated negative binomial mixed model with a natural-spline age term, adjusting for sex and stroke and including a swimmer random intercept; predictions are shown for short-course (SC; red) and long-course (LC; blue). Expected values include the contribution of the zero-inflation component (i.e., incorporate the probability that the season-fastest occurs on the first competition day, days_to_fastest = 0).
Competition level
Competition exposure was defined as whether a swimmer had any high-level racing (National/International) within a season-event (any_high).
Probability of swimming fastest at the first race
Figure 4 shows a parallel pattern for the number of attempts. Exposure reduced the chance that the first race of the season-event was the season-fastest (48% lower, OR = 0.52, p < .001). In practical terms, a swimmer whose season includes National or International competition is roughly half as likely to produce their fastest time at the first race, compared with a swimmer racing only at Club or Regional level. This is consistent with high-exposure seasons being structured around a target competition later in the season, with earlier races functioning as preparation (in addition to swimmers being fully tapered for a benchmark target competition).
Model-based age curves for race to fastest by competition exposure, format and distance. Panels show the overall expected number of races required to achieve the season-fastest within a season-event (races_to_fastest), plotted against age and faceted by distance (50–1500 m). Solid lines are population-average predictions from the two-part hurdle model that combines (i) the probability that the season-fastest occurs on the first race and (ii) the expected number of additional races when the fastest is not first. Line colour indicates season type (SC red; LC blue). Line type indicates competition exposure within the season-event (solid = no National/International exposure; dashed = any National/International exposure).
Additional races needed when the fastest is not first
When the fastest race was not first, exposure was associated with longer progression; 28% more races were required before reaching the season-fastest (IRR=1.28, p < .001). However, this effect was smaller than the shift in fastest-first probability, indicating that exposure primarily changes whether swimmers peak early, rather than substantially altering how long the progression takes once it is underway. Figure 4 presents the combined expected races-to-fastest by age, course format, and exposure, integrating both components. Across ages and distances, swimmers with high-level exposure require more total races on average, with the gap widest in younger age groups and in SC formats.
Days to fastest swim
Figure 5 shows that exposure primarily changes whether swimmers achieve their fastest swim immediately. Swimmers with any high-level exposure were less likely to swim their fastest performance immediately, approximately 91% lower odds (OR = 0.09, p < .001). This is the strongest single effect in the analysis. In high-exposure seasons, the probability of going fastest on day one drops dramatically, reflecting the fact that major competitions are typically scheduled later in the season window.
Model-based age curves for days-to-fastest by competition exposure, course format, and distance. Panels show the population-average expected number of calendar days from the first race in a season-event to the season-fastest performance (days_to_fastest), plotted against age and faceted by distance (50–1500 m). Curves are predicted from the zero-inflated negative binomial mixed model, adjusting for sex and stroke and accounting for repeated observations within swimmers via a random intercept. Line colour indicates season type (SC red; LC blue). Line type indicates competition exposure within the season-event (solid = no National/International exposure; dashed = any National/International exposure).
Among season-events where the fastest did not occur immediately, exposure was associated with approximately 15% fewer days to reach season-fastest (IRR=0.85, p < .001).
Single vs. multi swim competition format
For this question, the following were examined; (i) whether the season-fastest swim was more likely to occur in multi-swim competitions, and (ii) when both a heat and final exist, whether season-fastest swims were more likely to occur in a final, compared to the heats, as well as whether these tendencies differ with high-level competition exposure. Multi-swim formats provide multiple opportunities within the same competition and are likely to place the swimmer and the swim under higher stakes (i.e., a final).
The model was adjusted for age at fastest, overall competition opportunity within that season-event (swimmers with more competitions have more chances to encounter multi-swim formats), event characteristics (sex, course, distance, stroke), any high-level exposure (National/International; any_high), and clustering of repeated observations within swimmers (random intercept). Swimmers with high-level exposure were over four times more likely to achieve their season-fastest at a multi-swim competition (OR = 4.26, p < .001); a large effect that remained after accounting for overall competition opportunity and event structure. Independently, greater competition opportunity across the season-event also increased the likelihood that the fastest swim occurred in a multi-swim competition (84% higher, OR = 1.84, p < .001). LC format showed higher probabilities than in SC format (87% higher, OR = 1.87, p < .001). Figure 6 visualises these effects, where the probability that the season-fastest is achieved in a multi-swim competition rises steeply through adolescence, peaks in late teens/early twenties, and is consistently higher for exposed swimmers and in LC than SC.
Probability that the season fastest swim occurred in a multi-competition format. Panels show the model-predicted probability that a season-fastest was achieved in a multi-swim competition (i.e., a competition in which the swimmer raced the same event more than once, such as heats and a final), plotted against age at the season-fastest and faceted by distance (50–1500 m). Predictions come from a mixed-effects logistic model adjusting for event characteristics and accounting for repeated season-events within swimmers via a random intercept. Line colour indicates season type (SC red; LC blue). Line type indicates competition exposure within the season-event (solid = no National/International exposure; dashed = any National/International exposure).
Heats vs. finals in producing fastest swim
Round structure was examined directly by restricting to season-events where the fastest competition contained both a heat and a final. Within this subset, it was modelled whether the season-fastest occurred in the final rather than the heat. Descriptively, finals dominated (∼77% fastest-in-final overall), consistent with a “final advantage” in which swimmers are more likely to produce their top performance in the decisive round. The model showed systematic differences by sex, course, and distance. Males had lower odds of fastest-in-final (33% lower, OR = 0.67, p = .002), LC had higher odds (29% higher probability, OR = 1.29, p < .001), and longer events, (especially 200 m and 400 m distances), showed higher odds (200 m: 62%, OR = 1.62, p < .001; 400 m: over twice as likely, OR = 2.15, p < .001). Figure 7 shows how these probabilities vary over age and are shifted upward for exposed swimmers.
Probability that the season-fastest occurred in the final (heat + final competitions only). Panels show the model-predicted probability that the season-fastest was achieved in the final rather than the heat, conditional on the fastest competition having both a heat and a final, plotted against age at the season-fastest and faceted by distance (50–1500 m). Predictions come from a mixed-effects logistic model adjusting for event characteristics and accounting for repeated season-events within swimmers via a random intercept. Line colour indicates season type (SC red; LC blue). Line type indicates competition exposure within the season-event (solid = no National/International exposure; dashed = any National/International exposure).
Discussion and practical applications
This longitudinal follow-up study to previous work by Johns et al. 1 provides a developmental account of how quickly swimmers achieve their fastest swim within a season and how many attempts it typically takes, while separating calendar timing (days-to-fastest) from opportunity (races-to-fastest). The present analysis clarifies where the developmental and contextual effects are strongest, thus providing coaches and support staff with data-informed insights into how a season can be planned.
Three key findings emerged from this body of work. First, the likelihood of producing a season-fastest at the first opportunity increases with age. Second, course format and competition exposure shift the two components in different ways, where LC seasons and high-level exposure both reduce the probability of an immediate season-fastest. Third, season-fastest performances in high-exposure contexts are disproportionately realised in multi-swim competitions and in finals, indicating that the competitive conditions for fastest performance change systematically across the development pathway.
Age and the timing of season-fastest performances
For coaches working with younger swimmers, particularly those aged 10–15, the data indicate that the first race of a season is unlikely to be the fastest. Therefore, programme planning should accommodate a longer window and more race opportunities before a season-fastest is likely to emerge. Over-interpreting an early-season result (good or bad) risks misjudging where the swimmer is in their progression. For older swimmers, from mid-to-late adolescence onward, the season-fastest increasingly occurs at the first opportunity. This shifts the coaching emphasis from “race exposure” toward quality of preparation, repeatability and execution across a season. The progressive decline in fastest-first probability with increasing distance, from 50 m through to 1500 m, suggests that coaches should apply wider windows and more race opportunities specifically in longer events, not just for younger swimmers.
Relative to the earlier descriptive work, 1 the present analyses demonstrate that these patterns are systematically structured by development, and modified by competitive context. Course format provides a clear illustration of why separating timing from opportunity matters. LC seasons are associated with a later season-fastest in calendar time (approximately 65% more days than the SC season), yet LC swimmers require fewer race attempts to reach their fastest. This apparent paradox is consistent with LC seasons having a longer calendar structure, while concentrating meaningful performance attempts into fewer, more targeted races, whereas SC seasons may provide more frequent early-season racing without necessarily accelerating the calendar timing of the fastest. In addition, the SC season generally takes place after a swimmer has had an extended break from competition, during which swimmers may still be building form and fitness. In practical terms, an early-season fastest in a LC season may carry different meaning than one in a SC season. In LC, achieving the fastest at the first race is substantially more common (3.11 times as likely as in SC), meaning it is less likely to signal premature peaking and more likely to reflect genuine readiness. In SC, the same early fastest should be interpreted with more caution, as the format routinely requires several race attempts before the fastest emerges.
This is consistent within the coaching literature. For the SC season, swimmers often enter competitions with shorter deloads rather than a full taper and competitions are stacked more frequently, acting as race pace stimuli.6,7 The data support this approach, where more frequent SC racing does not appear to accelerate the arrival of the season-fastest in calendar time, but it does mean more race attempts before the fastest emerges. This highlights why evaluating performance should be considered in both days and attempts. A swimmer can be early in the race sequence, yet late in the calendar (or vice versa), and these have different implications for programme adjustment and the approach to competition selection across each format. This would appear to be supported by the evidence that LC places greater physiological demands on swimmers than the SC format due to fewer turns, reduced opportunities for lactate clearance and recovery.12,13
Stroke and sex differences, while smaller than the age and format effects, are also worth considering for programme design. Males took longer than females to reach their season-fastest in calendar days (11%) and were less likely to produce their fastest swim in the first race (21% lower odds). Freestyle swimmers were the least likely stroke group to be fastest-first (28% lower odds than backstroke), while butterfly swimmers required more additional races when the fastest was not first (11% more). These patterns are broadly consistent with the stroke specific developmental literature. 14 Consistent with this study, Born et al. 14 showed that breaststrokers plateaus later in development, suggesting that breaststroke swimmers may continue to benefit from accumulated competition exposure for longer. Similarly, Iglesias Garcia et al. 13 showed that backstroke and breaststroke had the largest performance differences between SC and LC, across all ages groups and distances (particularly 100/200 m), suggesting these swimmers face greater adjustments (physiological and technical) between formats. For coaches, this means that stroke-specific expectations for competition programme length are warranted. For example, a breaststroke swimmer who has not yet produced their fastest performance after three or four competitions is showing a ‘normal’ pattern, whereas the same observation for a freestyle swimmer may be more problematic.
Competition level and performance trajectories
Competition exposure further clarifies how season-fastest should be interpreted. Any exposure to National/International racing sharply reduced the likelihood of an immediate season-fastest swim. Swimmers with high-level exposure were roughly half as likely to go fastest at the first race (OR = 0.52) and showed dramatically lower odds of going fastest on the first competition day (OR = 0.09). By contrast, the effect on the extra-race component was much weaker. When the fastest was not first, exposure added only modestly to the number of additional races required. This asymmetry is important. It means exposure primarily changes whether swimmers peak early, not how long it takes once they don’t.
In practical terms, high-exposure seasons look less like “find form quickly” seasons and more like “build-to-peak” seasons, where early racing functions as preparation and the fastest swim is expected later. This is consistent with the interpretation offered by Stone et al., 7 who found that coaches distinguish between competitions used as training stimuli and those warranting a full taper. The present data provide quantitative support for that distinction, showing that the competition programme structure itself shapes when peak performances can emerge. For coaches, the practical implication is that in seasons targeting a major championship (e.g., Olympic Games and World Aquatics Championships), an early season-fastest should not automatically be interpreted as a positive sign. Instead, it may signal that the swimmer has peaked prematurely relative to the intended trajectory. Conversely, coaches should not be alarmed if a swimmer with a full championship programme has not yet produced their fastest time early in the season. The data suggest that this is the expected pattern, and not an indication of poor preparation. Once progression was underway, exposed swimmers reached their season-fastest marginally more quickly in calendar time than non-exposed swimmers (15% fewer days, IRR=0.85), consistent with more deliberately structured competition schedules concentrating the progression into a shorter window.
Competition format and the distribution of season-fastest swims
Exposed season-events were substantially more likely to achieve their season-fastest in multi-swim competitions (i.e., formats that offer multiple swims), and, when heats and finals were available, season-fastest occurred disproportionately in finals rather than heats (approx.77% of season-fastest performances occurred in finals when both rounds existed). Together with the exposure effects on days- and races-to-fastest, these findings converge on a coherent applied account. Season-fastest in high-exposure contexts tend to be realised under higher-stakes, multi-round conditions rather than at the first low-stakes opportunity.
This “final advantage” was more pronounced in longer events (200 m and 400 m), in LC format, and for female swimmers. This pattern may partly reflect a depth of field issue, whereby the substantially greater number of competitors in shorter events (50 m and 100 m) at all levels of competition creates a different competitive dynamic, with qualification pressure in heats potentially reducing the capacity to build into a final compared with longer distances. Males were less likely to produce their fastest in the final (33% lower odds), a finding that warrants further investigation. This may reflect that male swimmers at national level more frequently produce qualifying-standard performances in heats, reducing the incentive to build into a final, whereas female competition structures may more consistently reward final-round performance. The finding that finals are the predominant site of season-fastest performances has direct implications for preparation. Coaches should design competition routines that explicitly target multi-round readiness, including controlled heat swims (qualifying safely without excessive energy expenditure), structured between-round recovery protocols, and the ability to build into a final performance.
To support coaches in applying these findings in practice, a reference tool providing expected benchmarks for fastest-first probability, expected races, and expected days-to-fastest across age, distance, course format, and exposure level, is provided in the Supplementary Material (Tables S1–S3). Further, an interactive tool accompanying this paper is intended to support this reflective process by allowing coaches to explore the data for their own swimmer profiles. The link is available in the supplementary material.
Limitations
Several limitations should be noted. The data are observational and do not include direct measures of training load, tapering, illness/injury, 15 or explicit competition-targeting decisions; therefore, it is not possible to distinguish between a deliberately small/low race volume competition programme and one reduced by injury/illness, in any given season. Accordingly, participation in competitions likely reflects a varying combination of deliberate planning and some structural constraints (opportunity, financial, timing etc). Exposure effects may also reflect selection into higher-level competitions (i.e., qualification and/or selection) rather than competition effects, per se. Similarly, because the models adjust for the number of race opportunities, the results describe expected patterns conditional on a given level of opportunity. They do not prescribe an optimal number of competitions.
Additionally, the sample reflects swimmers represented in the national competition database and results may not generalise to all levels or alternative season definitions. For example, the Irish national competition structure may have structural features that differ from other national contexts. Therefore, the analysis can only provide a partial answer to the design of effective competition programmes, and indicative benchmarks for coaches working with similar populations. The reference predictions in the supplementary tables are derived from a specific population (elite and sub-elite Irish swimmers, 2011–2024) and a specific set of model assumptions (female backstroke reference profile, race opportunity held at the median). Coaches working in different national systems, with different competition calendars or squad structures, should treat these benchmarks as indicative rather than prescriptive. Despite these limitations, the central inference is clear: developmental profiles, and their modification by opportunity, exposure, and competition format, systematically shape when and where season-fastest performances occur.
Future work could seek to link competition results directly to training load and periodisation data. This would enable researchers to distinguish between competition programmes that are deliberately designed and those that emerge from circumstance. The inclusion of pacing data (split times within races) would also allow examination of whether the progression toward season-fastest reflects improving execution, greater willingness to commit to race strategy, or physiological readiness. Given the complexity of linking these variables in a large observational dataset such as this one, a pragmatic first step would be to investigate a smaller, more targeted sample — for example, a single cohort tracked across one season or one Olympic cycle. Beyond this, a valuable step would be to evaluate the utility of the practitioner reference tool directly with coaches to understand if access to age and event related benchmarks change how coaches interpret competition results and plan programmes. A validated, coach facing tool, informed by data from a wider population, would represent a meaningful applied contribution and is the direction we intend to pursue.
Conclusions
This study examined how quickly swimmers reach their season-fastest performance and how this varies with age, course format, competition exposure, and competition structure. The analysis provides benchmarks that coaches can apply directly to competition programme planning. For younger swimmers (approximately 10–15 years), the likelihood of producing a season-fastest at the first race is low, and coaches should plan for multiple race opportunities before the fastest is likely to emerge. From mid-to-late adolescence onward, swimmers increasingly arrive at their first race ready to perform at or near their best. This reflects a shift from accumulating race exposure toward refining execution and repeatability of performance, supported by the reference benchmarks provided in the Supplementary Material.
Competitive context matters. In seasons that include National or International competition, the season-fastest is substantially less likely to occur at the first opportunity and more likely to emerge later, after earlier races have served as preparation. An unusually early season-fastest in such a season may warrant attention, as it could indicate that the swimmer peaked ahead of the target competition rather than reflecting genuine early-season form.
Finally, as swimmers progress through the development pathway, their season-fastest performances become increasingly likely to occur in multi-swim competition formats and, within those, in finals rather than heats. Preparation for championship performance should therefore explicitly target multi-round readiness, including controlled heat swims, structured between-round recovery, and the ability to build into a final.
Supplemental Material
sj-docx-1-spo-10.1177_17479541261466580 - Supplemental material for How long to fast? Development and competition context in season-fastest swimming performances
Supplemental material, sj-docx-1-spo-10.1177_17479541261466580 for How long to fast? Development and competition context in season-fastest swimming performances by Karen L Johns, Dijana Cocic, Merim Bilalic and Cormac Powell in International Journal of Sports Science & Coaching
Footnotes
Acknowledgements
The authors would like to thank Swim Ireland for providing the performance data used in this study.
Ethical considerations
Ethical approval was not required for the study in accordance with the local legislation and institutional requirements.
Consent to participate
Not applicable
Consent for publication
Not applicable
Author contributions
KJ : Conceptualisation, Investigation, Writing – original draft, review & editing.
DC: Formal analysis, Visualization, Writing – original draft, review.
MB: Formal analysis, Visualization, Writing – original draft, review, methodology,
CP : Conceptualisation, Investigation, Writing – original draft, review & editing.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of conflicting interests
The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: CP was previously employed by Swim Ireland but is no longer in said position.
Data availability statement
Supplemental material
Supplemental material for this article is available online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
