An athlete's pacing pattern is widely recognized to have a substantial influence on the performance in endurance sports (Abbiss and Laursen, 2008; Tucker, 2009; Roelands et al., 2015; Casado et al., 2021). As the pacing pattern composed by the athlete directly influences the energy turnover rate and thereby performance, a variety of patterns (e.g., negative, all-out, positive, parabolic, and variable) exists. In endurance sports of short durations (≲2 min), all-out and positive pacing profiles are advantageous and common, e.g., ≤800 m running (Tucker et al., 2006) and ≤ 200 m swimming (Menting et al., 2019). It has been suggested that for events lasting longer than 2 min, an even pacing strategy may be optimal to achieve the best time or highest mean power output (Abbiss and Laursen, 2008). In practice, this seems evident at least for relatively long durations (≳ 10 min), e.g., ≥5,000 m running (Tucker et al., 2006; Diaz et al., 2018). For events lasting approximately 2–10 min, the literature and applied practice seem less certain. Typically, various parabolic (U-, J-, or reverse J-shaped) pacing profiles are observed (Menting et al., 2019; Casado et al., 2021). Interestingly, several best times and world records are set with small variations in pace and relatively even profiles, e.g., 5,000 m speed skating (≈6 min) (ISU, 2020) and 4,000 m track cycling (≈4 min) (Tokyo Olympics, 2021). Moreover, in various sports, the more “calculated” even pacing pattern has occurred over the recent years, making the even pacing more even, cf. e.g., Foster et al. (2014) and Diaz et al. (2018). However, as stated by Casado et al. (2021), pacing profiles within sports differ as well as the type of competition (championship; finals vs. qualifications, goal to set best times) implying the need for study of pacing profiles in the sport specific content.

The finishing time in rowing is always more than at least 5 min in 2000 m events, that is, World Rowing's standard regatta distance, which is used in all major championships. The importance of viscous fluid (hydro- and aerodynamic) drag yields a highly non-linear relation between mechanical power output and velocity. Consequently, the energetic cost of an uneven pacing strategy may be higher than in sports where the total resistive force is less dominated by viscous drag, e.g., running (gravity), cross-country skiing (gravity and snow-ski-friction) or uphill cycling (gravity) in sufficiently steep slopes (van Druenen and Blocken, 2021). Therefore, an even pacing strategy seems the most obvious choice to optimize performance in rowing. In contrast, previous studies, that have used 500 m split times to assess various aspects of pacing strategies in rowing, agree that, typically, a reverse J-shaped pacing profile (fastest-slow-slow-fast) is applied in 2,000 m rowing (Garland, 2005; Brown et al., 2010; Muehlbauer et al., 2010; Muehlbauer and Melges, 2011). Interestingly, a fast-start approach is also found in long-distance rowing (Edwards et al., 2016).

However, it is unclear if rowers at the highest performance level will seek toward more even pacing strategies (flatten their reverse J), to further increase performance. The study by Brown et al. (2010) supports the hypothesis that there is a relation between performance level and pacing strategy. In contrast, Garland (2005) did not find any difference in pacing profiles between winners and the rest of the heat. Furthermore, the study by Muehlbauer and Melges (2011) confirmed that there are significant differences in pacing patterns in heats and finals. Thompson (2014) summarized that previous studies of pacing in rowing have found similar pacing profiles independent of finishing position. Notably, previous studies have proved differences in pacing profiles between performance levels in some sports, e.g., running (Hettinga et al., 2019), and insignificant differences in pacing between performance levels in other sports, e.g., 1,500 m swimming (Lara and Del Coso, 2021). Further, Losnegard et al. (2016) found that the pacing pattern depended on performance in cross-country skiing among males, but not among females.

Inspired by the previous studies of split times, we analyze pacing profiles, from recent World (2017–2019) and European (2017–2021) championships, in the present study. Official 2,000 m race times, as well as intermediate split times at each 500 m mark, are obtained from the website of the World Rowing Federation. In our study, we limit the analysis to A-finals only, in order to investigate the split time characteristics and pacing profiles at the very highest performance level and to limit tactical behaviors often seen in the qualifications. This obviously limits the number of events and boat crews, which we compensate for by assessing eight recent championships.

Our main objective was to establish the relation between rowing performance, in terms of the final placement, and the pacing strategy. Furthermore, we tried to establish whether or not the pacing strategy in the considered A-finals is a function of sex and/or the number of crew members in the boat class.

This study provides novel information on how rowers in A-finals of World and European championships apply their pacing strategy. 106 finals were considered, in which all had six competing boat crews. We used the coefficient of variation (CV) of the 500 m intermediate times to quantify the evenness of the pacing profiles. Analyses of an additional pacing variation parameter, the sum of the relative differences (SRD) of the 500 m split times, were overall consistent with analyses of the CV. We did not find significant differences in the CV of female vs. male crews—although females paced somewhat more evenly than men and had smaller SRD (p = 0.03, d = 0.18)—or in the CV of various crew members (singles vs. doubles/pairs vs. quads/fours), cf. Table 3. However, more even pacing profiles were evident for the medallists compared to that of the 4th–6th places, implying that the pacing pattern discriminates athletes at the highest performance level in World class rowing, cf. Figures 1, 2 and Table 2.

The mean pacing profile of all crews, in the subgroups (sex, crew members), and for all placements except the last, followed a reverse J-shape, cf. Figures 3, 4 and Tables 2, 3. In 78% of the crews, the first 500 m segment was the fastest of all four 500 m segments. These findings are in line with studies of previous rowing events (Garland, 2005; Brown et al., 2010; Muehlbauer et al., 2010; Muehlbauer and Melges, 2011). Moreover, segment analyses revealed the winner's dominance in the relatively slower mid-race sections, cf. Figure 6. Notably, for the winners, being the fastest crew in the last segment was considerably less common compared with the three other segments, and in 79% of the finals, the winner was also leading at the 1,500 m mark, cf. Figure 7. This is consistent with A-finals of previous World championships (2003–2007) and Olympic games (2004, 2008), studied by Brown et al. (2010), who found that 78% of winners were placed first at the 1,500 m mark. We note the lack of drafting benefits in rowing, which is an important difference between rowing and several other mass start sports, where the outcome often comes down to a so-called endspurt phenomenon, e.g., 1,500 m running (Casado et al., 2021). Consequently, there seems to be limited good reasons for saving energy for an endspurt (Tucker, 2009), a likely reason why the regular J-shaped pacing profile, often seen in other endurance sports of similar duration, does not seem to be that common in rowing, cf. Figure 5 and Table 4. Interestingly, in a recent rundown of the regattas at the 2021 Tokyo Olympics, Kleshnev (2021) showed that the mean pacing profile was J-shaped. Kleshnev (2021) also noted that the Tokyo Olympics was the second fastest of 28 World regattas since 1993, and the pacing profiles were characterized as “much more even” than previous events.

From a pure mechanical point of view, our main finding that performance depends on the evenness of the pacing profile, seems very likely. Considering the streamlined vessels and the mean velocities—which ranged from 4.1 m/ s to 6.0 m/ s in the considered events—the resistance force on the boat and crew, F, should be dominated by viscous forces that can be approximated as (Faltinsen, 2005),

Here ρ is the fluid density, u is the velocity, C is a dimensionless coefficient and S is a characteristic surface area, e.g., the friction forces along the wet hull are expressed with a friction coefficient, C_f, and wet hull surface, S_w. Alternatively, if pressure forces dominate—for instance the aerodynamic drag forces on the rower's upper body—the expression may be written in terms of a form drag coefficient, C_d, and the projected frontal area, A. In either case, Equation (6) predicts that these viscous drag forces are proportional to velocity squared, F∝u². Consequently, the power (P; the rate at which a force does work), needed to overcome the total resisting force in rowing, is approximately proportional to velocity cubed, P∝u³. More refined models and experimental investigations support these assumptions, e.g., P∝u^3.2 (di Prampero et al., 1971), P∝u^2.95 (di Prampero, 1986), P∝u^2.8 (Affeld et al., 1988), P∝u^2.7 (Hofmijster et al., 2007), and P∝u^2.92 (Hill and Fahrig, 2009). Arguments have been made that the exponent is somewhat smaller if a model of actual rowing is considered (Shephard, 1998; Hill and Fahrig, 2009). Nevertheless, a highly non-linear relation between power and velocity is obtained. Consequently, for a given mean power output, the mean velocity is maximized when rowing at a constant velocity, i.e., constant power (even pacing strategy), assuming similar conditions over the course of the race. This is impossible for two obvious reasons; 1) initially, the boat must be accelerated from rest; 2) intracyclic velocity variations are inevitable consequences of the propulsive mechanisms of rowing. However, these effects may be treated separately from changes to the mean stroke velocity during the greatest part of the race. Typically, the initial acceleration phase of a 2,000 m regatta is limited to the very start of the race, increasing the time spent on the first 50 m of the race by 1–4 s compared to the mean 50 m split time, but not affecting later 50 m split times (Thompson, 2014, Figure 11.1). Moreover, the intracyclic velocity variations may be treated as representing an additional resistance component—which requires an additional power output—compared to traveling the boat constantly at the mean stroke velocity (Hofmijster et al., 2018), with an associated increase in the net mechanical power of 2–10% (Nigg, 1984; Sanderson and Martindale, 1986; Hofmijster et al., 2007; Hill and Fahrig, 2009; de Brouwer et al., 2013).

A reason why relatively fast starts (despite the initial acceleration phase) are common in rowing, may be due to the fact that rowing regattas are lane-based mass start events with both mental and hydrodynamic arguments for leading, not following. Since the crews are faced backwards, being upstream of competitors is a visual advantage. Therefore, a lead may give a sense of control with positive mental feedback which could influence performance (Schiphof-Godart et al., 2018). Clear favorites that are likely winners in many race scenarios may enjoy the confidence in taking an early lead to have visual control of the race. This may explain why the winners had slightly larger CV, relatively faster starts and more often applied the fast-start and fade pacing strategy, compared to the other medallists, cf. Figures 1–3, and Tables 2, 5, 6. In addition to mental aspects, there are also hydrodynamic arguments for uneven pacing. Incident waves from the competitors' boats may yield both a larger resistance force and make it more challenging to row effectively for a crew downstream. The risk of incident waves from competing boats may be estimated by considering the Kelvin angle, that is, the angle between the boundary of the wave system and course of a vessel (Faltinsen, 2005). Assuming deep water conditions, the Kelvin angle is arcsin13≈ 19°. If the lane width is 12.5 m and the boats are placed in the center of their lanes, the course direction distance from the wave propagation of the upstream boat to the stern of the downstream boat, must be less than 35 m to avoid incident waves from the upstream boat on the downstream boat. This distance is reduced if the upstream crew position the boat closer to the line of buoys. Consequently, if competitors apply a very fast start, incident waves is a risk worth noting if applying a more conservative pacing strategy. However, in general, the smallest risk of incident waves from competing boats throughout the race is achieved by going as fast as possible for the whole race distance, not just the start.

Medal winners in major rowing championships pace with smaller variation than their competitors. This may imply that a more even pacing profile, than what is typically applied by non-podium finishers, is advantageous in rowing.

The dataset supporting the conclusions of this article will be made available by the corresponding author, without undue reservation.

Ethical review and approval or written informed consent were not required for the study on human participants in accordance with the local legislation and institutional requirements.

FM collected and analyzed the data. Both authors contributed to the design of the study, wrote the manuscript, and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.