Sports performance is strongly determined by body features (1, 2). The same occurs in competitive swimming. Studies reported that swimmers who have bigger somatic features (specifically, limbs’ lengths, girths, and hand surface area) are more likely to deliver better performances (3, 4).

Notwithstanding, most research about this topic is related to the front-crawl stroke. Being the fastest stroke (5), researchers put a lot of focus in this swim stroke. Overall, the better performing swimmers in front-crawl are taller, bigger, with a larger arm span, and greater hand and foot surface areas (6, 7). By contrast, scarce evidence can be found in the remaining swim strokes (backstroke, breaststroke, and butterfly) (8), and even less in medley (9). Nonetheless, for young male and female breaststrokers aged between 10 and 13 years, it was found that the forearm girth and the leg length were (among others) the strongest predictors of the 100 m breaststroke event with a positive association (i.e., larger dimensions led to fastest performances) (10). In backstroke, the same research group found out that the strongest predictors of the 100 m event were the sitting height (SH), leg length, and two girths (forearm and arm relaxed girth) (11). In butterfly, Dimitric et al. (12), noted that arm span in girls was significantly related with better performances in the 50 m event. As for medley, a study with young swimmers noted significant correlations between the 200 m medley event and height, hand length, arm span, and body fat percentage (this with a negative correlation) in boys. As for girls, non-significant correlations were found between anthropometrics and the 200 m medley event. Notwithstanding, all these findings are for sprinting events, i.e., 50 or 100 m (and middle-distance for the medley) and in youth swimming. Thus, the literature lacks information about the role that somatic features play on other race distances in all swim strokes and medley, and in older age groups.

Some of these outputs were based on allometric modeling (10, 11). This approach consists in fitting a straight line to logarithmic transformations of the original bivariate data and then back-transforming the resulting equation to form a two-parameter power function in the arithmetic scale. The method has the dual advantages of enabling investigators to fit statistical models that describe multiplicative growth while simultaneously addressing the multiplicative nature of residual variation in response variables (heteroscedasticity) (13). Moreover, applying these allometric models to adult swimmers by providing deeper insights about the somatic features than can have a meaningful effect on each swim stroke performance can also help youth swimming coaches on understanding their swimmers’ handicaps and strengths.

The current scientific knowledge indicating that not having “ideal” anthropometrics can compromise the swimmers’ energetics and ultimately their performance. Therefore, it was hypothesized that applying these allometric models to adult swimmers will provide deeper insights about the somatic features of swimmers allowing us to identify those of them that provide the determinate effect on performance in each swim stroke. With this tool, coaches and their swimmers will be able to understand the handicaps and strengths to be determined by their unique somatic potential. To support our hypothesis, the aim of this study was to explore which key somatic features are common to all four swim strokes and individual medley, and specifically to identify which characteristics benefit only specific strokes, i.e., that are “stroke specific.”

Table 2 presents the descriptive statistics (mean ± standard deviation) of the swimmers’ age and somatic measurements by stroke. As a non-significant sex effect was noted in the following allometric model, data are presented with both sexes plotted together. The TH (0.037) and WC (0.032) were the only variables presenting a significant stroke effect, i.e., differences between strokes. In Figure 1, it is possible to note that associations between swimming speed and arm length are similar in breaststroke and medley, and in freestyle, backstroke and butterfly.

The parsimonious solution to the backward elimination regression analysis of log-transformed swim speed [Ln(SS)] resulted in the multiple regressions model presented in Table 3. The multiplicative allometric model relating swim speeds to the predictor variables found six variables that are significantly associated/predictors of Ln(SS) (note that the whole body size variables of BM and H were not significant and dropped from the analysis). Five of these predictors were found to be “common” to, or associated with, all five strokes. Three of these were positively associated with Ln(SS), which are Ln(UBC), Ln(HB), and age². The other two “common” predictor variables were Ln(WC) and SST, both found to be negatively associated with Ln(SS) performance.

One other predictor variable, Ln(AL), was also found to be strongly associated with Ln(SS), but this association varied significantly between the four different swim strokes and medley (Table 3). This was identified by introducing stroke-by-predictor-variable interaction (see Statistical analysis section). The significant “stroke-by-arm length” interaction identified the arm length of freestyle (p < 0.001; η_p²= 0.332), backstroke (p < 0.001; η_p²= 0.163), and butterfly (p < 0.001; η_p²= 0.175) swimmers to have a much stronger predictive ability on swim speed Ln(SS), compared to the baseline breaststroke and medley swimmer groups (p > 0.05) (see Table 3). The coefficient of determination, R² for the fitted multiplicative allometric model was 60.9% with the log-transformed error ratio being 0.079 or 8.2%, having taken antilogs.

The aim of this study was to explore which key somatic features are common to all four swim strokes and medley, and specifically to identify which characteristics benefit only specific strokes, i.e., that are “stroke specific.” Main findings revealed that associations between SS and AL were similar in breaststroke and medley, and in freestyle, backstroke and butterfly. The multiplicative allometric model found six variables significantly associated/predictors of Ln(SS) [(UBC), Ln(HB), Ln(WC), SST, age², and Ln(AL)]. Ln(AL) was also found to be strongly associated with Ln(SS), but this association varied significantly between the four different swim strokes plus medley.

Since decades, the scientific papers strongly report the importance of anthropometric and somatic features in swimming performance (19). These are also pointed out as significant predictors in swimming performance talent identification programs allowing us to early identify which swimmers are more likely to perform better in when achieving adulthood (10, 20). As aforementioned, solid evidence can be found about the front-crawl stroke but less about the remaining swim strokes and medley. Moreover, and as far as our understanding goes, only one study included the four swim strokes aiming to identify which key somatic features are common to all of them, and which ones benefit only specific strokes, i.e., ones that are “stroke specific” (4). Notwithstanding, this was conducted with adolescent age-group swimmers. Our findings, including young adult national and international level swimmers (Table 3), revealed that Ln(UBC), Ln(HB), Ln(WC), SST, age², and Ln(AL) were the predictors of SS. From these, Ln(UBC), Ln(HB), Ln(WC), SST, and age² were common to all four swim strokes plus medley. Regarding the positive association of age² with Ln(SS), swimming speed is rising within our swimmers’ age range (a quadratic within an exponential term). It was noted that elite swimmers are characterized by a high-level performance from 12 years on and progressively outperform swimmers from similar age (21). Thus, understanding this moment can help coaches and swimmers to better design and set realistic short- and long-term goals.

Regarding the predictors related to the swimmers’ trunk (Table 3), Ln(UBC) had a positive association and Ln(WC) a negative association. That is, greater UBC and smaller WC led to faster SS. At least in front-crawl, Papic et al. (22) aimed to determine the influence of torso morphology on maximum instantaneous hydrodynamic resistance. The authors noted that a larger area at the waist level was significantly correlated with a greater drag coefficient (greater drag leads to slower SS). Others, based on numerical simulations of male swimmers, analyzed the effect of different body shapes on hydrodynamic drag in the streamlined position (23). Four body shapes were scanned and analyzed: inverted triangle (“v-shape”), inverted trapezoid, rectangle, and oval shapes. From these, the inverted triangle was the one that presented the smallest projected area and surface area. Consequently, for speeds between 1.2 and 2.2 m/s, the inverted triangle shape was the one that presented smaller drag and drag coefficient values (23). This highlights the “v-shaped” body that swimmers have and allow them to achieve better performances. The SST also presented a negative association with SS, i.e., smaller thickness in the subscapular area led to faster SS. Overall, it was reported that a lean body mass and skinfolds thickness decrease were related to an enhancement of SS (24).

Ln(HB) was also retained as a significant predictor of Ln(SS) with a positive association with Ln(SS) (Table 3). Both experimental (25) and numerical studies (26) indicate the key role that the hand's dimensions (area, length, or width) have on the swimmers’ propulsion in front-crawl and, consequently, in the SS enhancement. The upper limbs, mainly the hands, are propelling surfaces that act as “paddles.” Thus, swimmers with larger hand areas are more likely to generate greater propulsive forces and hence faster SS (25). This rationale can be applied to the remaining swim strokes. Based on numerical simulations, it was possible to determine the hydrodynamic characteristics of the hands in various combinations of angle of attack and angle of orientation, as well as hand shape and speed, simulating the different phases of the stroke cycle in front-crawl (26). Based on four positions of the fingers (thumb adducted, thumb abducted, all fingers spread, and spreading fingers and thumb adducted), it was noted that at different moments within the stroke cycle, different finger positions delivered diverse propulsion outputs. Thus, swimmers can and should alter finger position and hand angles during the stroke cycle to generate more propulsion (26).

From the six predictors, Ln(AL) varied significantly between the four different swim strokes and medley with a positive association. Specifically, the arm length of freestylers, backstrokers, and butterflyers have a much stronger predictive ability on swim speed Ln(SS), compared to baseline breaststrokers and medley swimmers. Curiously, the three former swim strokes are the fastest ones and the two latter ones are the slowest (5). Biomechanically, SS can be improved by increasing stroke frequency, stroke length, or both (27). If stroke frequency can be trained or manipulated, stroke length is strongly correlated to the upper limbs’ length, namely, the arm span (28). It was found out that the arm length itself was a significant predictor of SS with a positive association in the front-crawl stroke (29). Moreover, swimmers with larger arm span (which includes the arm's length) are more likely to achieve greater distances with the hand entry and pull more water backward. This will lead to more propulsion and hence more SS. The arm span was one of the variables that was positively related to performance in all four swimming techniques and in individual medley (9). Nonetheless, our data revealed that longer upper limbs lengths (specifically the arm) are more important in the fastest swim strokes. Indeed, these (freestyle, butterfly, and backstroke) are more dependent from the SS achieved by the upper limbs rather than breaststroke (30).

Our findings demonstrate that it was possible to model SS in the four swim strokes plus medley based on allometric modeling. It should be mentioned that sex, swimmers’ level, and race distance were not retained as significant predictors. This means that the predictive model is suitable for both sexes, both competitive levels (i.e., national and elite), and all race distances. The five common predictors indicate that for all four swim strokes plus medley, there is a given moment in the swimmers’ career (age²), where performance may enhance exponentially, and a “v-shaped” trunk associated with a large hand area are significant predictors of SS. Swimmers with longer arms are more likely to deliver better performances in freestyle, butterfly, and backstroke. Based on these outputs (from older age groups where the growth likelihood is smaller), the practical applications become apparent—coaches of younger age-group swimmers must be aware of the importance that anthropometrics have on their swimmers’ performance. Due to the smaller number of swimmers, particularly in butterfly and medley, in comparison to the other strokes, the sample size can be considered as the main limitation of this study.

As a conclusion, allometric modeling indicated that five predictors were common to all swimmers—age², UBC, HB, WC, and SST. These indicate that the large trunk “v-shape” morphology and the large hand areas are important factors in determining fast performances. One predictor (arm length) varied significantly between the four different swim strokes plus medley. This highlights the role that long upper limbs play in fastest swimming performance, particularly in freestyle, butterfly, and backstroke.