Relationship between common measures of training stress and maximum mean power during road cycling races
In preparation for endurance cycle races, cyclists carry out a large volume of training to attain the necessary fitness to perform. These loads must be managed wisely to be optimally prepared for race day. In the early 1970s Banister and colleagues introduced empirical models that describe the relationship between training load and performance ability. Banister suggested that Performance = Fitness – Fatigue and proceeded to introduce mathematical sophistication to this underlying premise by incorporating decay constants for both fitness and fatigue. Banister suggested that training load rapidly influenced fatigue but only slowly influenced fitness. However, with recovery fitness was well maintained while fatigue quickly dissipated. More recently, commercially available software packages have made it easier for coaches and cyclists to engage in these concepts.
The TrainingPeaksTM software incorporates a performance manager, which is based on an impulse-response model for managing the training loads of cyclists based on data recorded by on board cycle ergometers called power meters. The aim of this study was to determine how well the performance manager model predicts the performance ability of competitive cyclists in road time trials, individual road races and multi-day events.
Nationally and Internationally competitive cyclists (20M, 5F) submitted power meter files for a six- to eight-month period. Measures of fitness, fatigue and freshness were derived in the performance manager from the day before each competition. Maximum mean powers (MMPs) for 5-s, 60-s, 5-min and 20-min durations were recorded from each race. Mixed modelling was used to estimate the linear relationship between changes in fitness, fatigue and freshness, and changes in the MMPs during competition.
Expressed as coefficients of variation (CV), within-cyclist variation in MMP from competition to competition ranged from 15% (5-s MMP) to 4.1% (20-min MMP). These CVs were too large for the MMPs to track the usual changes in performance that cyclists would show between competitions. When the bottom half of each cyclist's MMPs were discarded, only 5- and 20-min MMPs in time trials had CVs that could track reasonable changes in performance (~2.5%). However, the mixed models showed effects of fitness, fatigue and freshness on MMPs that were either unclear or too weak to be useful.
This study casts doubt on the use of fitness, fatigue and freshness measures to assess training load and the use of MMPs to assess performance in road cycling. Different models of measuring training loads should be investigated. Contextual information about each competition ride might reduce the error in MMPs by allowing filtering or adjusting for poor performances, but other measures of performance from competitions may be needed to determine whether fitness, fatigue and freshness are worth monitoring.