Measures of Rowing Performance T

Home , Rowing (sport)

Measures of Rowing Performance T. Brett Smith^ and Will G. Hopkins^ 1 Department of Sport & Leisure Studies, University of Waikato, Hanrtilton, New Zealand 2 Sport Performance Research Institute NZ, AUT University, Auckland, New Zealand

Contents Abstract 343 1. introduction 344 2. Measures of On-Water Rowing Performance 345 2.1 impeller iVieasurement of Boat Speed 346 2.2 Global Positioning System Measurement of Boat Speed 347 2.3 On-Water Ergometry 348 3. Measures of Off-Water Rowing Performance 349 3.1 Off-Water versus On-Water Time Trials 349 3.2 Reliability of the Off-Water Time Trial 350 3.3 Oti-ier Off-Water Test Measures 35I 4. Conciusian ' 356

Abstract Accurate measures of performance are important for assessing competitive athletes in practical and research settings. We present here a review of rowing performance measures, focusing on the errors in these measures and the implications for testing rowers. The yardstick for assessing error in a performance measure is the random variation (typical or standard error of measurement) in an elite athlete's competitive performance from race to race: ~ 1.0% for time in 2000 m rowing events. There has been little research interest in on-water time trials for assessing rowing performance, owing to logistic difficulties and environmental perturbations in performance time with such tests. Mobile ergometry via instrumented oars or rowlocks should reduce these problems, but the associated errors have not yet been reported. Mea- surement of boat speed to monitor on-water training performance is common; one device based on global positioning system (GPS) technology contributes negligible extra random error (0.2%) in speed measured over 2000 m, but extra error is substantial (1-10%) with other GPS devices or with an impeller, especially over shorter distances. The problems with on-water testing have led to widespread use of the Concept II rowing ergometer. The standard error of the estimate of on-water 2000 m time predicted by 2000 m ergometer performance was 2.6% and 7.2% in two studies, reflecting different effects of skill, body mass and environment in on-water versus ergometer performance. However, well trained rowers have a typical error in performance time of only -0.5% between repeated 2000 m time trials on this ergometer, so such trials are suitable for tracking changes in physiological performance and factors affecting it. Many researchers have used the 2000 m ergometer performance 344 Smith & Hopkins

time as a criterion to identify other predictors of rowing performance. Standard errors of the estimate vary widely between studies even for the same predictor, but the lowest errors (-1-2%) have been observed for peak power output in an incremental test, some measures of lactate threshold and measures of 30-second all-out power. Some of these measures also have typical error between repeated tests suitably low for tracking changes. Combining measures via multiple linear regression needs further investigation. In summary, measurement of boat speed, especially with a good GPS device, has adequate precision for monitoring training performance, but adjustment for environmental effects needs to be investigated. Time trials on the Concept II ergometer provide accurate estimates of a rower's physiological ability to output power, and some submaximal and brief maximal ergometer performance measures can be used frequently to monitor changes in this ability. On- water performance measured via instrumented skiffs that determine individual power output may eventually surpass measures derived from the Concept II.

1. introduction to search for investigations on measures of rowing performance published since 1970. To examine Rowing competitions usually involve races last- whether we had missed any relevant material, we ing 6-8 minutes over a 2000 m regatta course. undertook follow-up searches in Google Scholar Rowing demands a high level of endurance,!'^ esti- and SPORTDiscus^"^ where we combined the term mates of the aerobic contribution being 70-87%.^"^^ 'row' with the specific measure/s of interest 'GPS', Successful rowers tend to be tall and lean.'•*"*! 'Concept IF, 'time-trial', 'lactate tests', etc. There are up to 22 boat classes in international The yardstick for evaluating a measure of ath- rowing regattas, consisting of hea'vyweight or light- letic performance is the variability that top athletes weight rowers, male or female rowers, singles or show from one race or competition to the next. crewed boats, coxed or coxless boats and sweep or This variability, which is expressed as a within- sculling oars. In sculling, each rower has two oars, subject standard deviation, is analogous to the and boat types involve one rower (single), two standard error of measurement (or typical error of rowers (double) or four rowers (quad). In sweeping, measurement) in a reliability study of a perfor- each rower has one oar and boat types include mance test, with the repeated trials replaced by two rowers without a coxswain (coxless pair), two races. The race-to-race variability in finish times rowers with a coxswain (coxed pair), four rowers for elite rowers in world cups, world champion- without coxswain (four) and eight rowers with a ships and Olympic competitions is ~\%.^^^ These coxswain (eight). races are maximal efforts for highly motivated and Measuring changes in performance is Important well conditioned rowers, so the ~l% race-to-race for monitoring the progress of rowers during train- variability is, at first glance, an irreducible error for ing and for research assessing the effect of training any measure of rowing performance. However, and other interventions. In the only review^^ of the variability arising from environmental and other tools and tests available for monitoring rowing factors probably adds to the rower's inherent performance, the authors focused on monitoring physiological variability in performance from race for overtraining, with only a cursory examination of to race,'^l so measures of performance derived from the accuracy of the various measures of rowing a rowing ergometer can and do have standard performance. In this review we describe these mea- error of measurement less than 1% in reliability sures, their errors and the practical implications. studies. We used Google Scholar, SPORTDiscus™ A useful measure of rowing performance must and reference lists in reviews and research articles have acceptable validity, as well as reliability.

Validity of a measure is determined by comparing other score that results, on average, in one extra its values with those of a criterion measure, which medal in every ten competitions.''^ Simulations in rowing is competitive performance time over showed that this enhancement is a factor of 0.3 of 2000 m. The single best validity statistic is the the standard deviation of within-athlete race-to- standard error of the estimate (or typical error of race variability in performance, which for row- the estimate), which is a standard deviation rep- ing is therefore a 0.3% change in race time resenting the error in an individual's criterion (0.3% X 1.0%). Thresholds for quantifying magni- value predicted by the test or other measure. The tudes based on winning three, five, seven and nine standard error of the estimate and the other extra medals per ten races are 0.9% (moderate), regression validity statistics (the correlation co- 1.6% (large), 2.5% (very large) and 4.0% (ex- efficient and the regression or calibration equation) tremely large).t'"' To interpret the magnitude of an are specific to the population represented by the error (standard error of measurement or standard sample from which they are derived, and the error of the estimate) on this scale of magnitudes, standard error of the estimate is misleadingly the error should be doubled.'**! A good measure of smaller in samples with a narrower between- rowing performance would therefore need a subject standard deviation. Indeed, with a homo- standard error of measurement of less than 0.3% if geneous sample, the standard error of the estimate one wanted to be confident about trivial changes is simply the noise (standard error of measure- in performance. We will see that no rowing tests ment) in the criterion. For this reason, we have reach this level of precision. used a method^^l to adjust the standard error of the estimate from each study to a population with the 2. Measures of On-Water Rowing widest possible range of values (infinite standard Performance deviation). See Appendix 1 for the formulae. The standard error of the estimate is then an unbiased Although the 2000 m on-water time is the cri- estimate of the error in the criterion value arising terion measure of rowing performance, this mea- from error in the test measure and, as such, can be sure has a number of limitations. Environmental compared between studies for the purpose of conditions are the most important limitation, choosing measures with the smallest errors. The which contributes substantial variability to the adjusted standard error of the estimate of a test competitive performance of elite rowers.'*! Even measure is inevitably larger than the standard error for a group of rowers competing together in a of measurement of the criterion, because the single race, and therefore under seemingly iden- standard error of the estimate includes contribu- tical environmental conditions, rowers could be tions from the standard error of measurement of affected differently; for example, a headwind the criterion, the standard error of measurement of could hinder some rowers more than others, and the test measure and differences between subjects a side wind could unfairly benefit rowers in lanes that are not accounted for by the test measure. in which there is a wind shadow. Furthermore, The smallest worthwhile change in perfor- without special instrumentation in the boat, it is mance is another important consideration in the not possible to quantify an individual rower's assessment of measures of performance. When the performance from the speed of a boat with two, standard error of measurement of a performance four or eight rowers. Rowers from a crewed boat measure is similar in magnitude to the smallest could be tested individually in single sculls, but worthwhile change, the measure is sufficiently sen- the faster movement speeds in a crew boat could sitive to quantify small but meaningful changes, easily create a difference in technique. There are either when monitoring individual athletes or also biomechanical differences between single when performing controlled trials with realistic sculling and sweeping.t"-'^^ Instrumenting the sample sizes.t^l The smallest worthwhile enhance- boat (see below) may solve the problems of as- ment of an elite athlete's performance has been sessing the on-water performance of individual defined as the change in performance time or rowers in single or crewed boats. Many coaches

will not permit regular maximal 2000 m on-water Kellerman, Boothwyn, PA, USA) and the Cox- tests during the earlier stages of the season, owing mate (Coxmate, St Peters, SA, Australia). The to concerns about impairing aerobic develop- impeller has two advantages over the GPS. One ment, but performance would be available from advantage is that it can give accurate readings the competitions over 2000 m that occur regularly over any distance, whereas GPS has unaccept- during the ~4-month competitive season (Smith able error over short distances (see below). The TB, personal observations). other is that the impeller measures the speed of Despite problems with the 2000 m on-water the boat relative to any water current, so impeller performance tests, maximal boat speed over dis- speed more accurately reflects the performance tances other than 2000 m is a commonly used of the rowers. Windy conditions can create measure in rowing.''^^ The distances range from water currents,''^-'^' and while the boat speed 250 m to 15 km, depending on the energy sys- calculated by the impeller accounts for these tem(s) being trained in a given phase of the season currents, the impeller does not adjust for the di- (Smith TB, personal observations). A common rect effect of the wind on the boat, rowers and practice in this context is the use of 'prognostic oars;''^"' for example, when there is a headwind, speeds', which are either world record or some an impeller will indicate, appropriately, a slower other target speed for each boat class.['•'•'•*' The boat speed. performance of each boat is evaluated as a per- The impellers are calibrated upon installation centage of its prognostic speed to provide a rank- and checked regularly thereafter, as there is an- ing within a boat class and a comparison for boats ecdotal evidence that weed or other debris in the from different boat classes. Environmental condi- water can upset the calibration. In an effort to tions can still affect the accuracy of the ranking, increase the accuracy of the impellers in flowing because the environment has a greater effect on water, they are calibrated by travelling a known boats with fewer rowers and boats with female land distance upstream and downstream. The crews.'**' The boats are often handicapped for these combined land distance for the two runs is com- tests in an effort to improve the sense of competi- pared with the combined impeller distance, which tion. While handicapping does reduce environ- allows the appropriate caUbration to be calcu- mental effects, provided that conditions do not lated for that stretch of water. Accurate mea- change substantially between boats, the turbu- surement of speed with an impeller requires a lence (boat wash) can disadvantage the trailing constant speed and direction of the water current boats. over the period of testing and through all parts of The development of various rowing speed- the waterway that the boats travel. ometers has increased the popularity of the mea- There are no published studies examining the surement of boat speed during training and reliability and validity of the impeller in rowing. competition.''^J These devices are more con- To gain some understanding of their accuracy, venient than stop-watches, which require timing one of the authors examined the Nielsen Keller- at each end of the course and accurate measure- man impeller (NK) during regattas over 2000 m ment of the distance. The two common speed- on various international rowing courses. In 61 ometers are based either on an impeller or the observations of NK versus true boat speed over global positioning system (GPS). The impeller 2000 m, the NK units showed a negligible fixed measures speed relative to water (true speed), error (0.1%), but there was a moderate random while GPS measures speed relative to land. error of 1.2% even when wind direction and speed were taken into account (Smith TB, personal observations). While this amount of error is only 2.1 Impeller Measurement of Boat Speed slightly larger than the 1% yardstick considered The two popular devices that measure rowing appropriate to accurately monitor training, the NK boat speed via impellers attached to the hull are is not sufficiently accurate to quantify small but the Nielsen Kellerman Speed Coach (Nielsen meaningful changes in competitive performance.

2.2 Global Positioning System iVIeasurement unit reliability has established that there is little of Boat Speed difference between units of the same model of GPS, at least over long durations and distances,'-^'' GPS requires an unobstructed view of its sat- so it is probably safe to assume that findings with elhtes, but this requirement is seldom a problem one unit apply to all such units of a given model. on the waterways where rowers train and com- GPS technology has undergone a series of pete. The devices are easily swapped between rapid advances since May 2000, when the US boats and do not require calibration. Government made full precision available. We The early GPS devices sampled and calculated have therefore limited this review to GPS studies position once per second (1-Hz), but with recent published since then. We have included data from technical developments, units that sample as high studies with movement patterns similar to that of as 20-Hz are now available. Higher sampling fre- rowing, that is, straight-line movements. We have quencies are needed for accurate speed measure- also included data from movements around ment over shorter intervals or distances, but the 400 m running tracks but have excluded zigzag or 1 -Hz units are still in principle adequate (< 1 % error) T-shaped shuttle runs. In all, three studies pro- for quantifying boat speed over durations in excess vided useful information for rowing,P-"-'*^ and we of 100 seconds. have included some unpublished observations Proprietary algorithms employed by the var- (see figure 1). ious manufacturers of GPS are also considered to The accuracy of the 5-Hz MinimaxX (Catapult, influence accuracy, so the findings from a parti- Melbourne, VIC, Australia), the 1-Hz SPI-10 and cular GPS model should be considered to apply the 5-Hz SPI-Pro (GPSports, Fyshwick, ACT, only to that model.t-^°l Previous research on inter- Australia) was examined over a range of speeds

A MinimaxX 1 Hz O MinimaxX 5-1 OHz À SPI 1 Hz • SPI 5 Hz * NKXL4 35 , SEE (%) 5 -, SEE (%)

30-

25- O A A 3 - O 20 O O

15- 2 - © 10-

1 - 5-

0- 0-1 10 20 30 40 100 500 1000 2000 5000 10000 Distance (m)

Flg. 1. Plot of the standard error of the estimate for each measurement device over various distances. NK=Nieisen Kelierman impeller; SEE = standard error of the estimate.

and distances by Petersen et al.P^' The distances these two GPSs are shown in figtxre 1 as the points recorded by the GPS were compared with the with the smallest error. These errors are well within actual distances travelled on a 400 m running the 1 % yardstick and are sufficiently low to quan- track for walking 8800 m through to sprinting tify small but meaningful changes in boat speed in 20 m. It is apparent in figure 1 that the accuracy 2000 m time trials. Further research is required to of the GPS is more dependent on the manu- determine the GPS accuracy during rowing over facturer than the signal frequency. Figure 1 also distances shorter than 2000 m, when aliasing might shows that the 5-Hz version of the MinimaxX begin to make a substantial contribution. improved the error relative to the 1-Hz version From the data presented in figure 1 we make regardless of distance and speed.'^^l Pyne et al.t^"*' the following conclusions. The SPI elite and compared the 10-Hz MinimaxX with previously MinimaxX GPS are more accurate than the NK published results for the 5-Hz version for straight over 2000 m and the various SPI GPS units are line sprinting over 10, 20 and 40 m. Not apparent more accurate than the MinimaxX except for in figure 1 is the fact that the accuracy of the 10-Hz 2000 m rowing. The MinimaxX has less error at version was better than that of the 5-Hz version. higher sampling frequencies but the effect of fre- Rowing has a unique problem for speed mea- quency is not clear for the various SPI units. sured over a short distance. The average speed During rowing, the SPI-Elite has a smaller error for the various boat types during competition is than that determined over similar distances on a 4-6 m/sec, but speed varies by 2-3 m/sec during 400 m running track for other SPI models, even each stroke.P^'^*' An aliasing error may arise for those with a higher sampling frequency. It is from a combination of a 1-Hz sampling fre- therefore likely that cornering on athletic tracks quency, the large oscillations in velocity and a increases GPS error, f-^"' short sample period, and this error may worsen Environment changes will cause changes in the already poor accuracy of the GPS over boat speed, so regardless of how speed is mea- shorter durations and distances. Over the short- sured, it can be an inaccurate method for tracking est distances of interest to rowers (-250 m), GPS rowing performance. Even if ideal environmental sampling at greater frequencies (>1-Hz) would conditions could be guaranteed between trials (no overcome this aliasing error. wind, no water currents, no changes in the com- The accuracy of GPS in rowing has been position, depth and temperature of the water), examined in one published study, which is avail- boat speed is still not an accurate measure of an able only as a conference abstract.'^^^ In this study individual rower's performance within a crew boat. the race time recorded by 5-Hz MinimaxX was compared with the official race time for 244 row- 2.3 On-Water Ergometry ing boats during major competitions over 2000 m. The standard error of the estimate was 0.45 sec- The measurement of a rower's power output in onds, but not enough data were presented to a boat is now possible with on-water ergometers, convert this error to a percentage. However, if we which have been constructed for both sculling and assume the race time was 6-8 minutes, the sweeping.[-^^ These ergometers calculate power standard error of estimate would be -0.1%. output from kinetic data measured by sensors in One of the authors also examined the accuracy the rowlock and/or oar(s).P^' Although in theory of GPS in rowing by comparing the distance re- the power output from these devices should cor- corded by 10 SPI-Elite 1-Hz units with the 2000-m relate strongly with boat velocity, findings have distance travelled by 22 rowing boats during been mixed.'"'^''^"' On-water ergometers are also various regattas. The standard error of the esti- expensive, time consuming to install and calibrate mate was negligible (0.2%; Smith TB, personal and often fragile (Smith TB, personal observa- observations) and similar to the -0.1% estimated tions). Despite the potential benefits of these above from the 5-Hz MinimaxX for 2000 m ergometers, it remains to be determined how well rowing. The standard error of the estimate for the power measured by their sensors represents

power propelling the boat forward. We therefore versions of the Concept II lack the data to cal- advise caution in the use of on-water ergometers culate the standard error of the estimate.P-^^'^^' until the associated errors have been reported. The only other rowing ergometer in contention is the Rowperfect (Devon, UK); while it may more 3. Measures of Off-Water Rowing closely simulate on-water rowing movement,^^'^ Performance performance on this ergometer is less reliable (see below). The difficulties associated with assessment of on-water performance have led to widespread use 3.1 Off-Water versus On-Water Time Trials of stationary ergometers that simulate the action of on-water rowing. Studies of various rowing er- To examine how accurately rowing-ergometer gometers have found some differences in arm mo- performance predicts on-water performance, we tion,'^'' handle force and acceleration profiles,'^^] reviewed two studies where comparisons were and consistency in stroke timing'-'^' between off- made between 2000 m time trials conducted on water and on-water rowing performance. Despite water versus on a Concept II (see table I). We had these differences, the rowing ergometer is widely to exclude two further studies comparing Con- used by rowers, and the 2000 m ergometer time cept II performance with rankings from world trial is the most common measure of rowing championships,''*^-'*^l because the authors did not performance.t^'^"*' The Concept II air-braked provide competitive performance times that rowing ergometer (Morrisville, VT, USA) has would allow computation of a standard error of led the market since the development of the lib the estimate. model in 1986. The three subsequent models (c, d In a study of ten junior males whose on-water and e) have maintained the same rowing motion tests were single-scull competitions,!'*"^ the standard and method for calculating work output, but error of the estimate was 2.6%, which in our scale of have made changes to improve comfort, safety, magnitudes is a very large error. Although the robustness, damper settings and display options. 'competition results' were obtained on a 'windless A study to compare two Concept II models'^^' day', it is not clear whether the results were ob- was too underpowered to make meaningful con- tained from a single race. We therefore suspect en- clusions, but it is reasonable to assume that the vironmental conditions contributed to the error. only differences between the models are cosmetic The limited competition experience of the young (Smith TB, personal observations). rowers (aged 18.9+ 1.7 years) may also have con- In a recent development, the Concept II has tributed to the error, along with the substantial been placed on a slide to allow back and forth uncertainty in the estimate (90% confidence limits motion that simulates more closely the dynamics x/-^1.54) arising from the small sample size. of on-water rowing.P-^*-^^^ Comparisons of the In another study,^'*'' 49 junior elite males static version of the Concept II with the new dy- completed two 1000 m time trials in single sculls, namic 'sHder' version, suggested negligible dif- which were combined to give a 2000 m time and ferences in mean power output in time trials of a standard error of the estimate of 7.2%. All tests 2000 m and 6 minutes, but, on the shder, peak were undertaken at a national training camp, and mean stroke force were lower and stroke rate which presumably ensured high motivation. Both was higher.I^'^^J The slider is becoming increas- single-scull time trials were conducted on the same ingly popular in training, as there is evidence that day with winds of "approximately 2-3m/sec, dynamic rowing ergometry puts less strain on the the direction being predominately a headwind" lower back, which is beneficial to rowers who (p. 123).''*'' Although there are various potential commonly suffer back injury.P^l Other advantages sources of error, we believe that change in wind include a better 'on-water feel' and the capacity to speed and direction between trials for the different link devices together to simulate crew rowing. rowers was the main source of the extremely large The few studies comparing the static and slider error.

Table I. Standard error of the estimate of 2000 m single-scull performance time derived from correiations of this performance measure with measures from tests on a Concept II rowing ergometer. Measures shown in order of the adjusted standard error of the estimate (lowest to highest)

Test measures Rowers" Test measure Correlation SEE (%) 90% CI Adjusted References (mean + SD) SEE(%) 2000 m time single-sculi time 10M 7:2810:13" 0.72 2.1 1.4,3.3 2.6 Jurimae et ai.i""! Peak incrémental power (W) 10M 369 ±37 -0.70 2.2 1.4,3.4 2.8 VO2 at 4 mmoi/L lactate (L/min) 10M 4.13±0.63 -0.69 2.2 1.4,3.4 2.9 VOaâx (L/min) 10M 4.85 + 0.63 -0.64 2.4 1.5,3.6 3.3 Lactate at 350 W (mmoi/L) 10M 11.8±4.8 0.64 2.4 1.5,3.6 3.3 Power at 4 mmol/L iactate (W) 10M 275141 -0.61 2.4 1.6,3.8 3.5 40 sec ail-out mean power (W) 10M 614±82 0.60 2.5 1.6,3.8 3.6 VOsâ, (mL/kg/min) 10M 61.515.6 -0.33 2.9 1.9,4.5 7.8 2000 m time and body mass" 48 M 9 0.77 3.1 2.7, 3.7 ? Nevili et al.l"'! 2000 m time single-sculi time 48 M 9:0810:26" 0.54 4.1 3.5, 4.9 7.2 For other subject characteristics, see tabie il. b Mean 1SD test measure data for 2000 m singie-scuil times in Jurimae et al.'""I and Neviil et al.'"'! were presented in min:sec. c Measures combined via multiple linear regression. 90% CI = 90% confidence interval for the SEE; M = male; SEE = standard error of the estimate; ÔÎ=oxygen uptake; \IO2rmx = maximal VO2; ? indicates not provided or estimable.

These large standard errors of the estimate do needs to be obtained with a good sample size of not necessarily mean that performance on the top rowers under ideal environmental conditions. Concept II is invalid; more likely, there is large random error in the criterion measure of 2000 m 3,2 Reiiabiiity of tine Off-Water Time Trial on-water time, most of which is due to environmental factors.'^! When body mass was taken into If we accept that the performance on a Concept account in a multiple linear allometric regression II has adequate validity, at least for physiological equation, the observed standard error of the es- power output, an important issue is whether this timate decreased from 4.1% to 3.1% (see table I), ergometer has adequate rehability for tracking which is consistent with the observation that changes in performance. Reliability for tests on a body mass provided a substantial contribution to Concept II has been reported in two studies. Concept II time (r = 0.68) but a negligible con- Schabort et al.'"'''' examined 2000 m time-trial tribution to single-scull time (r = 0.04).['*'l The speed on a Concept II for eight well trained exponent of body mass in the allometric equation rowers who rowed on three occasions at 3-day was approximately -0.8 so the widespread prac- intervals and reported a 0.6% standard error of tice of expressing Concept II performance as measurement. In the second study, 15 elite rowers mean power per kilogram must produce close to performed five 500 m time trials each on a Con- the optimal measure for combining body mass cept II and Rowperfect ergometer and later per- with ergometer performance. formed a 2000 m time trial on 3 consecutive days Even when environment and body mass are on one ergometer.f"*^' The standard error of mea- taken into account, it is inevitable that some surement for 2000 m time were 0.4% and 1.1 % on rowers perform better on the Concept II than on the Concept II and Rowperfect, respectively, water and vice versa, so the Concept II cannot while the standard error of measurement for the predict on-water abihty perfectly. For a better 500 m trials were 0.7% and 1.1%. Combining estimate of the validity of performance on a these studies, the standard error of measurement Concept II, the standard error of the estimate of -0.5% is half the standard error of measure-

ment for competitive on-water performance (our other measures derived from the performance test 1% yardstick) and is only just outside the 0.3% on the Concept II have been investigated for their threshold to quantify smallest meaningful changes ability to predict rowing performance. In one in competitive performance. Although this reli- study,'^'' the relationship of these measures to on- ability is not ideal, it is unusual for tests of athletic water 2000 m performance has been quantified, performance to be this good.'''*' This higher reli- but, by far, the majority of the studies has used ability of performance on the Concept II is likely a the 2000 m Concept II time trial as the criterion result of less technical demands of the ergometer measure. and environmental effects causing less variability In the one study that used 2000 m single-scull to performance on a rowing ergometer, compared perfonnance,''*"^ the standard error of the estimate with on-water rowing. In comparison, the Row- was moderate to large (2.1-2.8%), probably be- perfect was inferior, and it is unclear whether its cause of the different effects of environment, tech- reliability would reach that of the Concept II with nique and body mass on performance on water enough familiarization. versus on the ergometer (table I). Approximately half of the remaining studies had enough data to calculate the standard error of the estimate. For 3.3 Other Off-Water Test Measures these studies the subject characteristics are in- The rowing community was aware of the value cluded in table II, while the standard errors of the of performance testing oti the Concept II long estimate for the various measures are in table III. before any reliability and validity studies were The measures that come close to the -1.0% yard- performed."-^ Indeed, the 2000m time trial on the stick are peak incremental power, maximal oxygen Concept II has become the most commonly used consumption (VO2max)' some measures of lactate selection tool for national rowing organizations power and power in the 30-second modified (Smith TB, personal observations). Furthermore, Wingate. These measures have adequate validity

Table II. Subject characteristics for studies used to caiculate the standard error of the estimate in 2000 m ergometer performance time for measures from rowing tests Study Rowers Body mass 2000 m (kg)" (L/min)" time (minisec)" Bourdinetai.l"'! 31 nationai and internationai heavyweight M 88.6 ±5.1 5.68 ±0.32 6:04±0:10 23 national and internationai iightweight M 74.0 ±1.8 5.05±0.20 6:21 ±0:07 54 combined 82.4±8.3 5.41 ±0.42 6:12±0:12 Cosgrove et ai.'"^! 13ciub-ievei M 73.1 ±6.6 4.5 ±0.4 7:05±0:11 Faff et al.i"'! 8 iVI (?) teenagers 85±14 4.97 ±0.48 6:45±0:14 Gillies and 6eiil^°l 10 competitive M 82.3±7.5 4.38 ±0.42 7:07±0:14 22 competitive F 71±10 3.19±0.57 8:17±0:30 32 combined 75±11 3.62 ±0.84 7:55 ±0:42 Jurimae et ai.'""' 10 experienced M 79.3 ±7.3 4.85 ±0.63 6:38±0:18 Nevili et ai.[='i 48 current/former senior A/B iVI 88±11 5.60 ±0.56 6:07 ±0:13 28 current/former senior A/B F 71.7 ±8.5 4.03 ±0.33 7:01 ±0:17 76 combined 82±13 5.02±0.91 6:27 ±0:30 Nevill et ai.l'i'i 48 eiite junior M 83±7 ? 6:44 ±0:11 Riechman et ai.'^^' 12 competitive F 67±12 3.18±0.35 7:47 ±0:12 Russeii et ai.l"l 19 eiite schooiboys 85±8 4.6 ±1.5 6:43±0:16 Womack et ai.P"! 10 coiiege M (pre-Fall) 86.1 ±7.3 5.25 ±0.69 6:48±0:18 10coiiegei\/l(post-Fali) 86.117.3 5.28 ±0.62 6:42±0:18 a Data are presented as mean ± SD. F = female; M = male; V02max = f^iaximai oxygen uptai

< i III I. 5 Ill a. IB 'S .[51 ] IBe l îtal . CD 0) CO vu c CO "5) O io nj (D I 1 w EI

¡Hi e o evi i g 5 Í ^ CC

CO CO CO CO T- in ^ CD r^ ^ ai cO CO CO in CD A ^ ^ c\i

h-.T-;Ti;T^cMi^T-inT-cqcDLnpin •1— oicMCOCOCOCOCM'^CO'^COlricO qLor^''~oî^^*^cDOîh^cococM^^ f •T-'T-^T-^CMT-^^-^CMI-^T-T-^-I-^CJCMT-^

a CO CO tri

OJ CO OJ O) Q} CO CO O) CO CO Ol CO CO

o C\J CD C7) OJ CM CM CO CO CO m CM in CO in CD o CO CD CO CD CO O O o o o O o o O O o o Ö Ö +1 +1 -H -^i -Í-I +\ -H o +1 +1 +1 +1 +1 +1 +1 +1 +1 -H +\ -hl +1 +1 +1 CD in CO CO CO CO LO CO OJ T- CO OJ CO 5 S f; tn ,0 5 ,4 1 g CJ o CO LO LO in sCO CO in sin in m '^ CO c\j cy CO

5 CO u. 2 2 u: 2 2 2 in CO SI ? ? CO 1- o o

Q. â S [54 ] g- 154 ] S i i. 1 40 ] "to CO 'S c co a 'S a ra "cô c CD •H-' 'S 'S 'S 1 I ac k e t al . ima n e t £ 1 rov e t a ac k e t al . 'ov e t a ro § .9 Riec h Fatt e Riec h Bour c Cosg i Nevil l Nevil l Nevil l Cosg i ir ce i 1 1 Juri m

NtOCO A A i--^T--^-r-^^T--^T-^CNÍC\ÍC-OJ04C^-OJC\jCvjC\jcrDCOCÓCOCOCÓCOC--

OJCJT-; C\J h-. T-T-cpcoh-Oïi- CD r^coinoïcqo) ocqoîcoco Oicocp ÎDN- O CDiriiri •^ r-^ oJc\ic\jc\ic\ic\ic\i w oJcocsJ oJ csico cowojco'^ cücócó coco h^ô" 1— in oíoíoJi— csTcNÎ^" co ^"uioo LO oqrôo"i—T-ô ococq co a^ cdcvJCO oJ co ÖÖT-^T-^T-^T-^T-^ T-^ T-^T-^T-^ 1-^ T-^T-^ cjcsjcjc\ic\j CNJOJCM C\JT-^

CD (NI h-. CO CO CD LO h- •<^ co co OJCM CO CUCM CvjOJCgOJOJ OJCOCO COCM

OJ OJ OJ 00 CO 00 CO s s Si o o o o o

h- h- CO CO 00 00 OJ cy OJ +i UJ . (f^ LU IVJ o r^ T- ^ CO CO OJ t- •^ 1- oo CO CO CO

5 CO

u_ S 5 a. CM CN o CJî OJ o CM o s OJ CO CO 00 OJ OJ T— 1— T— T— CO CO •

I I'S (5 1 (5 1 CO to

to 'S ra œ ai . tD 'S 'S ii n 'S ra = c C3) = > ^ Uï > E = evil l evii l our c (Doom z CD z Z £D Ü Z

CD o CM CO CO CX) in OJ T- ^ •«-•,- CO cri iri cd oJ CO CÓ N! CO A in CO CO

CO CO CO o CO CO in O) CO CO cô c\i CVJ CO CO CO CO O) CD' C\J CO r^ T- CO O CO C3Î CD CD CO c\j CNJ .^ .^ co CO Ö 1-^ -r-^ T-^ T-^ OJ CM -^

LU UJ m CO CO LO T- T- CO t^ o CD CD q q q -^ CO CM CO CM CO CM CM CM CO CÓ CM CO iâ CO CO CO

O) 1- O) CO en lO CO Oî h- in cy CO CO G) iñScMi-COCOOCMCO CO CO h* CO 00 CO CO CO CD d d d d Q d P I I CI) d d I Ö Ö Ö Ç> -0 , 1 ? 1 T 1

d °î H C3 C^ t 1 0 0 OOCOOJini- CM CM "ï o 1— o o -)-i o o -|-| +1 0 CO CÖ cDcocDr^t^T- T-coin -H +1 -»-I +1 CO +' +1 1- O +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +( +1 CDinO)O)N-LO CO CM ^ m m 0 in COOCDOOCOCM-^CCM M rricri^^irifiri ^ iri iri C-O C~O - CO CM CO T - CO CO y— CM CO CO CM' m CD CD in

S u- li.' 5 LL u.' 2 Tj- CO COi— COCOOQ ^ COCO 00 LO CM C\JC0CJ 1OT--^CM SI ? SI SI e

E III .tí ó J¿ lî II 1 S E E d) •-*-*- ^ E E ¿ ra rà râ fil Q> Ol O) O =: CO CO s 5= 5S ' 5 5 5 í/i ir-. (D o o o .g .g cococoinioinin

for assessing moderate differences in rowing performance between rowers. In four studies, multiple linear stepwise re- lé S' gression analyses provided best combinations of ? CD measures to predict 2000 m, Concept II time-trial g g ^ I I performance. The sample sizes were far too low in III three of these studies to perform such analyses, so £ O ce the relatively low standard error of the estimate of 0.5-1% we obtained from their data must be substantial underestimates of the true error. In the other study, Neviil et al.'^'l combined data from 48 males and 28 females to obtain a rea-

CO CO CD sonable sample size, but the result is effectively a I" CO iri CO prediction equation for distinguishing between genders. Nevertheless, the resulting standard CO cq T- CD CM co ^ h- aq error of the estimate was relatively low (1.6%), so o co ^* iri CÓ co -»t -^ co cd ^ h-" 1^" oá 1-" co' there may still be some value in combining several uf p co' E S cvi c\j CÓ ^ ^ ^ ^ ^ measures for predicting 2000 m performance on S Z . - 03 m > the ergometer, and especially on water. Definitive (D M studies need to be performed. CM ai A low standard error of the estimate for a per- CM CM CO to oi formance measure is desirable, but measures with higher standard error of the estimate may still be useful if they have standard error of measurement low enough to reliably track an athlete's change ? S 2 i in performance. Unfortunately, only one rowing study provided data to calculate the standard error of measurement for performance measures S other than the 2000 m time trial.'^^1 Owing to a complex research design in this study, we were +1 CM only able to calculate the standard error of mea- CO (D OJ T- C\J 1- T- surement for various measures of lactate thresh- I c old and for peak lactate in an incremental test. The I t standard error of measurement for lactate threshold were relatively low (0.5-1.8%) but there is co co considerable error arising from the small sample size of ten elite male rowers. The lack of reliability studies in rowing led us m S <1> to examine the reliability of similar performance Si ^ £ I measures for other modes of exercise in a com- prehensive meta-analytic review.t''^^ The measure of reliability in the review was the standard error CJ CO .§ of measurement of power output; we have therefore divided the standard error of measurement .E .£ 03 -^ o by 3 to obtain an equivalent standard error of E E lia 1} in LD 'S measurement for performance time, as explained 0) 2 2 ^ ^ (0 CO CO Ü £ "o ^ co in that review. The most reliable tests that might CD < O ij Jj I be applicable to rowing were peak incremental

power (standard error of measurement for time Acknowledgements -0.3%), V02max and lactate threshold (-0.5%). While all tests are well within the 1% yardstick, No sources of funding were used to assist in the prepara- tion of this review. The authors have no conflicts of interest only peak incremental power could track smallest that are directly relevant to the content of this review. worthwhile changes. These measures also have the lowest standard error of the estimate for predicting 2000 m time-trial performance on a Appendix Concept II (see table III). The only other measure Calculation of the standard error of the with a low standard error of the estimate in our estimate, f^^l review is the modified 30-second Wingate test on If X and Y are the practical and criterion in the the Concept II, but the standard error of mea- validity study, r is their correlation, ex and ey are surement of Wingate measures was somewhat their random errors, n is the sample size, SD is larger (-1.2%) than that of the other two mea- standard deviation, and SEE is the standard error sures in the meta-analytic review.'"'^l Thus, it is of the estimate, then: possible that Wingate performance is more reliable on the Concept II than on other ergometers. ex = .^[SOx^ (l-r^ SD^^ / (SDy^-ey^))]; In summary, peak incremental power, VO2max' some measures of lactate threshold power and observed slope of regression line = r(SDY / SDx); possibly 30-second power, have measurement observed SEE = SDy ^/[(l-r^) (n-l)/(n-2)]; properties that make them potentially valuable iFor assessing rowing performance. In our view, true slope = (observed slope) / (l -e V02niax provides no information additional to that provided by peak incremental power, which true SEE without criterion error = (true slope) ex; along with 30-second power has the advantage of true SEE with criterion error = W [(true SEE) +eY^]. requiring no equipment other than the Concept II. These tests can be performed weekly at any The adjusted SEE shown in the tables is the true time of the year, with little impact on the training SEE with criterion error. The random error in the programme. Whether the measures can track criterion, ey, was assumed to be 1% for 2000 m performance adequately on the rowing ergometer, single-scull performance time (see table I) and and more importantly on water, is a question that 0.5% for 2000 m Concept II performance time (see needs to be addressed with further research. table III). This approach cannot be used for measures derived by multiple linear regression unless the 4. Conclusion authors provide the mean and standard deviation of the predicted values. Measures of on-water rowing performance are very noisy, owing to the effects of environment, and they do not measure performance of an individ- References 1. Hagerman FC, Hagerman GR, Mickelson TC. Physiologi- ual in a crew. Performance testing on the Concept cal profiles of elite rowers. Phys Sportsmed 1979; 7: 74-81 II eliminates these problems. Peak incremental 2. De Campos Mello F, De Moraes Bertuzzi RC, Grangeiro power and 30-second power on this ergometer are PM, et al. Energy systems contributions in 2,000m race simulation: a comparison among rowing ergometers and likely to be useful for frequent monitoring of a water. Eur J Appl Physiol 2009; 107: 615-9 rower's physiological power output. However, the 3. Hagerman FC. Connors iVIC, Gault JA, et al. Energy ex- Concept II does not adequately address the skill penditure during simulated rowing. J Appl Physiol 1978: component of performance on water. Instrumen- 45: 87-93 tation to measure each rower's on-water power 4. Sécher NH. The physiology of rowing. J Sports Sei 1983; 1:23-53 output should provide the best measure of rowing 5. Mikulic P. Anthropométrie and physiological profiles oi" rowers of varying ages and ranks, kinesiology 2008; 40: 80-8 performance, but it remains to be seen whether the 6. Shephard RJ. Science and medicine of rowing: a review. errors are acceptably low. J Sports Sei 1998; 16: 603-20

7. Mäestu J, Jürimäe J, Jürimäe T. Monitoring of performance 27. Vogler A, Lindg A, Riee A. Accuracy and reliability of Mini- and training in rowing. Sports Med 2005; 35: 597-617 maxx GPS technology in rowing [abstract]. In: Cabri J, Alves 8. Smith TB, Hopkins 'WG. Variability and predictability of F, Araújo D, et al., editors. Proceedings of the 13th Annual finals times of elite rowers, Med Sei Sports Exerc 2011; 43 Congress of the European College of Sport Science; 2008 Jul (11)2155-60 9-12. Gamlebyen grafiske; 2008: 413 [online]. Available from URL: http://www.eess.mobi/index.php?option=com_content 9. Hopkins WG, Hawley JA, Burke LM. Design and analysis &view=article&id=323&Itemid=109 [Accessed 2012 Feb 21] of research on sport performance enhancement. Med Sei Sports Exerc 1999; 31: 472-85 28. Kleshnev V. Comparison of measurements of the force at the oar handle and at the gate or pin. Rowing Bio- 10. Hopkins WG, Marshall SW, Batterham AM, et al. Pro- mechanics Newsletter 2010: 1-3 gressive statistics for studies in sports medicine and exercise science. Med Sei Sports Exerc 2009; 41:3-13 29. Baudouin A, Hawkins D. Investigation of biomechanical factors affecting rowing performance. J Biomech 2004; 37: 11. Kleshnev V. Estimation of biomechanical parameters and 969-76 propulsive efficiency of rowing. 1998 [online]. Available from URL: http://www.biorow,eom/Papers.htm [Accessed 30. Smith R, Galloway M, Patton R, et al. Analysing on-water 2011 Jan 17] rowing performance. Sports Coach 1994; 17: 37-40 31. Lamb DH. A kinematic comparison of ergometer and on- 12. Kleshnev V. Propulsive efficiency of rowing In: Sanders RH, water rowing. Am J Sports Med 1989; 17: 367-73 Gibson NR, editors. Proceedings of the XVII International Symposium on Biomechanics in Sports; 1999 Jun 30-JuI 6; 32. Kleshnev V. Comparison of on-water rowing with its simu- Perth (WA): Perth (WA): School of Biomédical and Sport lation on Concept2 and Rowperfect machines. In: Wang Q, Science, Edith Cowan University, 1999: 224-8 editor. Scientific proceedings of the XXIlIth International Symposium on Biomechanics in Sports; 2005 Aug 22-7. 13. McArthur J. High performance rowing. Wiltshire: The Cro- Beijing. Konstanz: International Society of Biomechanics wood Press, 1997 in Sports, 2005: 130-3 14. Kleshnev V. Prognostic times. Rowing Biomechanics 33. Dawson RG, Lockwood RJ, Wilson JD, et al. The rowing Newsletter; 2005. 1 cycle: sources of variance and invariance in ergometer 15. Patterson A. Review of rowing technology. World Rowing and on-the-water performance. J Motor Behav 1998; 30: Coaches Conference; 2011 Jan 22 Windsor. Lausanne: In- 33-43 ternational Rowing Federation, 2011; 1-30 34. Hahn A. Bourdon P, Tanner R. Protocols for the physiolo- 16. Langmuir I. Surface motion of water induced by wind. gical assessment of rowers. In: Gore CJ, editor. Physiologi- Science 1938; 87: 119-23 cal tests for elite athletes. Champaign (IL): Human Kinetics, 17. Wu J. Wind-induced drift currents. J Fluid Mech 1975; 68: 2000:311-26 49-70 35. Vogler AJ, Rice AJ, Withers RT. Physiological responses to 18. Kleshnev V. Weather and boat speed. Rowing Biomechanics exercise on different models of the concept II rowing erg- Newsletter; 2009: 1-3 ometer. Int J Sports Physiol Perform 2007; 2: 360-70 19. Pulman C. The physics of rowing. Ithaca: University of 36. Benson A, Abendroth J, King D, et al. Comparison of Cambridge, 2005 rowing on a Concept 2 stationary and dynamic ergometer. J Sports Sei Med 2011; 10: 267-73 20. Witte TH, Wilson AM. Accuracy of non-differential GPS for the determination of speed over ground. J Biomech 37. Holsgaard-Larsen A, Jensen K. Ergometer rowing with and 2004; 37: 1891-8 without slides. Int J Sports Med 2010; 31: 870-4 38. CoUoud F, Bahuaud P, Doriot N, et al. Fixed versus free- 21. Duffield R, Reid M, Baker J, et al. Accuracy and reliability fioating stretcher mechanism in rowing ergometers: me- of GPS devices for measurement of movement patterns in chanical aspects. J Sports Sei 2006; 24: 479-93 confined spaces for court-based sports. J Sei Med Sport 2010; 13: 523-5 39. Elliott E, Lyttle A, Birkett O. The RowPerfect Ergometer: a training aid for on-water single scull rowing. Sports Bio- 22. Jennings D, Cormack S, Coutts AJ, et al. The validity and mech 2002; 1: 123-34 reliability of GPS units for measuring distance in team sport specific running patterns. Int J Sports Physiol Per- 40. Jurimae J, Maestu J, Jurimae T, et al. Prediction of rowing form 2010; 5: 328-41 performance on single sculls from metabolic and anthropo- métrie variables. J Hum Movement Stud 2000; 38: 123-36 23. Petersen C, Pyne D, Portus M, et al. Validity and reliability of GPS units to monitor cricket-specific movement pat- 41. Nevill AM, Beech C, Holder RL, et al. Scaling concept II rowing ergometer performance for differences in body terns. Int J Sports Physiol Perform 2009; 4: 381-93 mass to better reflect rowing in water. Scand J Med Sei 24. Pyne DB, Petersen C, Higham DG, et al. Comparison of 5- Sports 2010; 20: 122-7 and 10-hz gps technology for team sport analysis [abstract]. 42. Mikulic P, Smoljanovi T, Bojani I, et al. Relationship be- Med Sei Sports Exerc 2010; 42: 78 tween 2000-m rowing ergometer performance times and 25. Caplan N, Gardner T. Modeling the infiuence of crew World Rowing Championships rankings in elite-standard movement on boat velocity fiuetuations during the rowing rowers. J Sports Sei 2009; 27: 907-13 stroke. Int J Sports Sei Engng 2007; 1: 165-76 43. Mikulie P, Smoljanovic T, Bojanic I, et al. Does 2000-m 26. Hill H, Fahrig S. The impact of fiuetuations in boat velocity rowing ergometer performance time correlate with final during the rowing cycle on raee time. Scand J Med Sei rankings at the World Junior Rowing Championship? A case Sports 2009; 19: 585-94 study of 398 elite junior rowers. J Sports Sei 2009; 27: 361-6

44. Sehabort EJ. Hawley JA, Hopkins WG, et al. High relia- 52. Rieehman SE, Zoeller RF, Balasekaran G, et al. Prediction bility of performance of well-trained rowers on a rowing of 2000 m indoor rowing performance using a 30 s sprint ergometer. J Sports Sei 1999; 17: 627-32 and maximal oxygen uptake. J Sports Sei 2002; 20: 681-7 45. Soper C, Hume PA. Reliability of power output during 53. Russell AP, Le Rossignol PF, Sparrow WA. Prediction of rowing changes with ergometer type and race distance. elite schoolboy 2000-m rowing ergometer performance Sports Biomech 2004; 3: 237-48 from metabolic, anthropométrie and strength variables. 46. Hopkins WG, Schabort EJ, Hawley JA. Reliability of power J Sports Sei 1998; 16:749-54 in physical performance tests. Sports Med 2001; 31: 211-34 54. Womaek CJ, Davis SE, Wood CM, et al. Effects of training 47. Bourdin M, Messonnier L, Hager JP, et al. Peak power on physiological correlates of rowing ergometry perfor- output predicts rowing ergometer performance in elite mance. J Str Cond Res 1996; 10: 234-8 male rowers. Int J Sports Med 2004; 25: 368-73 55. Bourdon PC, David AZ, Buekley JD. A single exereise test for 48. Cosgrove MJ, Wilson J, Watt D, et al. The relationship be- assessing physiologieal and performance parameters in elite tween selected physiological variables of rowers and row- rowers: The 2-in-l test. J Sei Med Sport 2009; 12: 205-11 ing performance as determined by a 2000 m ergometer test. 56. Hopkins WG. Adjusting regression statistics to assess va- J Sports Sei 1999; 17:845-52 lidity of measures in different populations. In: Loland S, 49. Faff J, Bienko A, Burkhard-Jagodzinska K, et al. Diagnostic B0 K, Fasting K, et al., editors. Proceedings of the 14th value of indices derived from the critical power test in as- Annual Congress of the European College of Sport sessing the anaerobic work capacity of rowers. Biol Sport Science; 2009 Jun 24-7; Oslo. Oslo: Gamlebyen grafiske, 1993; 10:9-14 2009:413 50. Gillies EM, Bell GJ. The relationship of physical and physiological parameters to 2000 m simulated rowing performance. Res Sports Med 2000; 9: 277-88 Correspondence: Dr T. Brett Smith, Department of Sport & 51. Nevill AM, Allen SV, Ingham SA. Modelling the determi- nants of 2000 m rowing ergometer performance: a pro- Leisure Studies, University of Waikato, Private Bag 3105, portional, curvilinear allometric approach. Scand J Med Hamilton, New Zealand. Sei Sports 2011; 21: 73-8 E-mail: [email protected]

© 2012 Adis Data Intormation BV. Aii rights reserved. Sports Med 2012; 42 (4) Copyright of Sports Medicine is the property of ADIS International Limited and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.