Monitoring the load placed on athletes in both training and competition has become a very hot topic in sport science. Both scientists and coaches routinely monitor training loads using multidisciplinary approaches, and the pursuit of the best methodologies to capture and interpret data has produced an exponential increase in empirical and applied research. Indeed, the field has developed with such speed in recent years that it has given rise to industries aimed at developing new and novel paradigms to allow us to precisely quantify the internal and external loads placed on athletes and to help protect them from injury and ill health. In February 2016, a conference on “Monitoring Athlete Training Loads—The Hows and the Whys” was convened in Doha, Qatar, which brought together experts from around the world to share their applied research and contemporary practices in this rapidly growing field and also to investigate where it may branch to in the future. This consensus statement brings together the key findings and recommendations from this conference in a shared conceptual framework for use by coaches, sport-science and -medicine staff, and other related professionals who have an interest in monitoring athlete training loads and serves to provide an outline on what athlete-load monitoring is and how it is being applied in research and practice, why load monitoring is important and what the underlying rationale and prospective goals of monitoring are, and where athlete-load monitoring is heading in the future.
Pitre C. Bourdon, Marco Cardinale, Andrew Murray, Paul Gastin, Michael Kellmann, Matthew C. Varley, Tim J. Gabbett, Aaron J. Coutts, Darren J. Burgess, Warren Gregson and N. Timothy Cable
While historically adolescents were removed from their parents to prepare to become warriors, this process repeats itself in modern times but with the outcome being athletic performance. This review considers the process of developing athletes and managing load against the backdrop of differing approaches of conserving and maximizing the talent available. It acknowledges the typical training “dose” that adolescent athletes receive across a number of sports and the typical “response” when it is excessive or not managed appropriately. It also examines the best approaches to quantifying load and injury risk, acknowledging the relative strengths and weaknesses of subjective and objective approaches. Making evidence-based decisions is emphasized, while the appropriate monitoring techniques are determined by both the sporting context and individual situation. Ultimately a systematic approach to training-load monitoring is recommended for adolescent athletes to both maximize their athletic development and allow an opportunity for learning, reflection, and enhancement of performance knowledge of coaches and practitioners.
William A. Sands, Ashley A. Kavanaugh, Steven R. Murray, Jeni R. McNeal and Monèm Jemni
Athlete preparation and performance continue to increase in complexity and costs. Modern coaches are shifting from reliance on personal memory, experience, and opinion to evidence from collected training-load data. Training-load monitoring may hold vital information for developing systems of monitoring that follow the training process with such precision that both performance prediction and day-to-day management of training become adjuncts to preparation and performance. Time-series data collection and analyses in sport are still in their infancy, with considerable efforts being applied in “big data” analytics, models of the appropriate variables to monitor, and methods for doing so. Training monitoring has already garnered important applications but lacks a theoretical framework from which to develop further. As such, we propose a framework involving the following: analyses of individuals, trend analyses, rules-based analysis, and statistical process control.
Dean Ritchie, Will G. Hopkins, Martin Buchheit, Justin Cordy and Jonathan D. Bartlett
Load monitoring in Australian football (AF) has been widely adopted, yet team-sport periodization strategies are relatively unknown. The authors aimed to quantify training and competition load across a season in an elite AF team, using rating of perceived exertion (RPE) and GPS tracking.
Weekly totals for RPE and GPS loads (including accelerometer data; PlayerLoad) were obtained for 44 players across a full season for each training modality and for competition. General linear mixed models compared mean weekly load between 3 preseason and 4 in-season blocks. Effects were assessed with inferences about magnitudes standardized with between-players SD.
Total RPE load was most likely greater during preseason, where the majority of load was obtained via skills and conditioning. There was a large reduction in RPE load in the last preseason block. During in-season, half the total load came from games and the remaining half from training, predominantly skills and upper-body weights. Total distance, high-intensity running, and PlayerLoad showed large to very large reductions from preseason to in-season, whereas changes in mean speed were trivial across all blocks. All these effects were clear at the 99% level.
These data provide useful information about targeted periods of loading and unloading across different stages of a season. The study also provides a framework for further investigation of training periodization in AF teams.
Jonathan D. Bartlett, Fergus O’Connor, Nathan Pitchford, Lorena Torres-Ronda and Samuel J. Robertson
The aim of this study was to quantify and predict relationships between rating of perceived exertion (RPE) and GPS training-load (TL) variables in professional Australian football (AF) players using group and individualized modeling approaches.
TL data (GPS and RPE) for 41 professional AF players were obtained over a period of 27 wk. A total of 2711 training observations were analyzed with a total of 66 ± 13 sessions/player (range 39–89). Separate generalized estimating equations (GEEs) and artificial-neural-network analyses (ANNs) were conducted to determine the ability to predict RPE from TL variables (ie, session distance, high-speed running [HSR], HSR %, m/min) on a group and individual basis.
Prediction error for the individualized ANN (root-mean-square error [RMSE] 1.24 ± 0.41) was lower than the group ANN (RMSE 1.42 ± 0.44), individualized GEE (RMSE 1.58 ± 0.41), and group GEE (RMSE 1.85 ± 0.49). Both the GEE and ANN models determined session distance as the most important predictor of RPE. Furthermore, importance plots generated from the ANN revealed session distance as most predictive of RPE in 36 of the 41 players, whereas HSR was predictive of RPE in just 3 players and m/min was predictive of RPE in just 2 players.
This study demonstrates that machine learning approaches may outperform more traditional methodologies with respect to predicting athlete responses to TL. These approaches enable further individualization of load monitoring, leading to more accurate training prescription and evaluation.
Shaun J. McLaren, Michael Graham, Iain R. Spears and Matthew Weston
To investigate the sensitivity of differential ratings of perceived exertion (dRPE) as measures of internal load.
Twenty-two male university soccer players performed 2 maximal incremental-exercise protocols (cycle, treadmill) on separate days. Maximal oxygen uptake (V̇O2max), maximal heart rate (HRmax), peak blood lactate concentration (B[La]peak), and the preprotocol-to-postprotocol change in countermovement-jump height (ΔCMJH) were measured for each protocol. Players provided dRPE (CR100) for breathlessness (RPE-B) and leg-muscle exertion (RPE-L) immediately on exercise termination (RPE-B0, RPE-L0) and 30 min postexercise (RPE-B30, RPE-L30). Data were analyzed using magnitude-based inferences.
There were clear between-protocols differences for V̇O2max (cycle 46.5 ± 6.3 vs treadmill 51.0 ± 5.1 mL · kg−1 · min−1, mean difference –9.2%; ±90% confidence limits 3.7%), HRmax (184.7 ± 12.7 vs 196.7 ± 7.8 beats/min, –6.0%; ±1.7%), B[La]peak (9.7 ± 2.1 vs 8.5 ± 2.0 mmol/L, 15%; ±10%), and ΔCMJH (–7.1 ± 4.2 vs 0.6 ± 3.6 cm, –23.2%; ±5.4%). Clear between-protocols differences were recorded for RPE-B0 (78.0 ± 11.7 vs 94.7 ± 9.5 AU, –18.1%; ±4.5%), RPE-L0 (92.6 ± 9.7 vs 81.3 ± 14.1 AU, 15.3%; ±7.6%), RPE-B30 (70 ± 11 vs 82 ± 13 AU, –13.8%; ±7.3%), and RPE-L30 (86 ± 12 vs 65 ± 19 AU, 37%; ±17%). A substantial timing effect was observed for dRPE, with moderate to large reductions in all scores 30 min postexercise compared with scores collected on exercise termination.
dRPE enhance the precision of internal-load measurement and therefore represent a worthwhile addition to training-load-monitoring procedures.
Mathieu Lacome, Ben Simpson, Nick Broad and Martin Buchheit
Purpose: To examine the ability of multivariate models to predict the heart-rate (HR) responses to some specific training drills from various global positioning system (GPS) variables and to examine the usefulness of the difference in predicted vs actual HR responses as an index of fitness or readiness to perform. Method: All data were collected during 1 season (2016–17) with players’ soccer activity recorded using 5-Hz GPS and internal load monitored using HR. GPS and HR data were analyzed during typical small-sided games and a 4-min standardized submaximal run (12 km·h−1). A multiple stepwise regression analysis was used to identify which combinations of GPS variables showed the largest correlations with HR responses at the individual level (HRACT, 149  GPS/HR pairs per player) and was further used to predict HR during individual drills (HRPRED). Then, HR predicted was compared with actual HR to compute an index of fitness or readiness to perform (HRΔ, %). The validity of HRΔ was examined while comparing changes in HRΔ with the changes in HR responses to a submaximal run (HRRUN, fitness criterion) and as a function of the different phases of the season (with fitness being expected to increase after the preseason). Results: HRPRED was very largely correlated with HRACT (r = .78 [.04]). Within-player changes in HRΔ were largely correlated with within-player changes in HRRUN (r = .66, .50–.82). HRΔ very likely decreased from July (3.1% [2.0%]) to August (0.8% [2.2%]) and most likely decreased further in September (−1.5% [2.1%]). Conclusions: HRΔ is a valid variable to monitor elite soccer players’ fitness and allows fitness monitoring on a daily basis during normal practice, decreasing the need for formal testing.
Alistair P. Murphy, Rob Duffield, Aaron Kellett and Machar Reid
To investigate the discrepancy between coach and athlete perceptions of internal load and notational analysis of external load in elite junior tennis.
Fourteen elite junior tennis players and 6 international coaches were recruited. Ratings of perceived exertion (RPEs) were recorded for individual drills and whole sessions, along with a rating of mental exertion, coach rating of intended session exertion, and athlete heart rate (HR). Furthermore, total stroke count and unforced-error count were notated using video coding after each session, alongside coach and athlete estimations of shots and errors made. Finally, regression analyses explained the variance in the criterion variables of athlete and coach RPE.
Repeated-measures analyses of variance and interclass correlation coefficients revealed that coaches significantly (P < .01) underestimated athlete session RPE, with only moderate correlation (r = .59) demonstrated between coach and athlete. However, athlete drill RPE (P = .14; r = .71) and mental exertion (P = .44; r = .68) were comparable and substantially correlated. No significant differences in estimated stroke count were evident between athlete and coach (P = .21), athlete notational analysis (P = .06), or coach notational analysis (P = .49). Coaches estimated significantly greater unforced errors than either athletes or notational analysis (P < .01). Regression analyses found that 54.5% of variance in coach RPE was explained by intended session exertion and coach drill RPE, while drill RPE and peak HR explained 45.3% of the variance in athlete session RPE.
Coaches misinterpreted session RPE but not drill RPE, while inaccurately monitoring error counts. Improved understanding of external- and internal-load monitoring may help coach–athlete relationships in individual sports like tennis avoid maladaptive training.
Bart Roelands and Kevin De Pauw
on nutritional manipulations that aim to get athletes at the start of a race in the best possible shape 1 ; training strategies and training-load-monitoring tools to avoid having athletes crossing the thin line between training and recovery, making them vulnerable to nonfunctional overreaching and
Javier Raya-González, Fabio Yuzo Nakamura, Daniel Castillo, Javier Yanci and Maurizio Fanchini
participated. The results of the study showed that internal load markers were neither associated with injuries nor had predictive capacity to identify soccer players that will incur a noncontact injury. Internal load monitoring is increasingly popular in high-performance sport to ensure athletes achieve an