A case study demonstrating more accurate assessment of running fitness and race time prediction by TrainAsONE’s AI running coach when compared to traditional VO2max focused platforms.
Abstract
Here we present a case study of assessing the running fitness of a 50-year-old male over a period of approximately five weeks using two consumer devices / platforms that focus on VO2max, and a novel approach delivered by TrainAsONE’s AI running coach.
We objectively assessed the running fitness using two 10 km road races at the beginning and end of the study. The results showed an improvement in running fitness and a new personal record (PR) achieved by the subject in the second race.
The VO2max estimates and race performance predictions made by the consumer platforms exhibited negative trends and did not align with the actual race results.
However the TrainAsONE AI approach provided a novel fitness marker, Goal Score, which showed a fluctuating pattern with increasing peaks, with the greatest peak immediately prior to the subject achieving the PR at the second race.
Additionally, the 10 km race predictions provided by TrainAsONE were the most accurate, and demonstrated 0.7 % and 1.7 % errors for the two races respectively.
Introduction
VO2max is known to be a very poor predictor of race performance.
VO2max has been a mainstay of monitoring running fitness for many years. Traditionally, its measurement is performed in a laboratory as part of a formal peak-performance assessment exercise. However, many health and fitness platforms employ a variety of approaches to estimate its value from ‘in the field’ sub-maximal exercise. This is then used to gauge and monitor fitness, and predict race performance. Despite the widespread acknowledgement that VO2max is a very poor predictor of peak performance, consumer platforms still rely on it for assessing running fitness.
Methods
In this study, we evaluated the running fitness of a 50-year-old male using two consumer devices/platforms that focus on VO2max, RW and DAP. RW is a Running Watch sold by a very popular sports watch manufacturer. And DAP, a web application marketed as a Data Analysis Platform for athletes. We compare these to a novel metric, Goal Score, provided by the AI running coach, TrainAsONE (TAO).
The running fitness was objectively assessed using two 10 km road races at the beginning and end of the study. The daily training plan was generated by TrainAsONE specifically for the subject, targeting to achieve a new PR (race target time training being another unique and novel feature provided by TrainAsONE).
Results
Not only were the TrainAsONE predictions more accurate … but they trended upwards, matching the improved race times.
The VO2max estimates from RW and DAP decreased over the course of the study. The RW estimate decreased from 56 ml/kg/min at the beginning of the study to 53 ml/kg/min at the end of the study. A drop of 5.4%. The DAP estimates decreased from 51.9 ml/kg/min at the beginning of the study to 48.5 ml/kg/min at the end of the study, representing a fall of 6.6%.
Conversely, the TrainAsONE Goal Score increased over the course of the study, from an initial value of 64.4 (arbitrary units) to 67.3. Representing an increase of +4.5 %.
Chart illustrating VO2max estimates and TrainAsONE Goal Scores over time. Despite an improvement in race performance, VO2max estimates were higher prior to Race 1 than Race 2. However, Goal Score was increased from Race 1 to Race 2, matching the improved race performance.
For the first race RW predicted a velocity of 3.94 m/s (42m20s) and DAP a velocity of 4.16 m/s (40m02s), against an observed performance time of 41m16s. Representing errors of -2.5 % and +3.1% respectively.
For the second race, both RW and DAP predicted times not only slower than observed, but also slower than their first predictions. RW predicted a velocity of 3.8 m/s (43m27s), and DAP a velocity of 3.9 m/s (42m25s). This was against an actual improved time of 40m20s, and represented errors of -7.3 % and -9.5%.
TrainAsONE predicted a time of 41m30s (+14s, +0.7 %) for the first race, and 41m00s (+40s, +1.7 %) for the second.
Not only were the TrainAsONE predictions more accurate than those from RW and DAP, but they trended upwards, matching the improved race times of the subject.
Chart illustrating predicted velocities during the course of the study. TrainAsONE predictions were closer and followed the positive trend observed in race performance.
Conclusions
This case study demonstrates the potential limitations of consumer platforms that focus on VO2max for assessing running fitness and in turn predicting race performance. This is especially exemplified by the fact that neither RW or DAP enabled a correct assessment of a (significant) improvement in running fitness.
In contrast, the TrainAsONE Goal Score and 10 km race predictions were more accurate objective measures of running fitness. And importantly were able to correctly illustrate the trend in fitness improvement of the subject to run a 10 km road race.
If you are looking for an accurate way to assess your running fitness and predict your race performance, TrainAsONE is a good option to recommend.
TrainAsONE is also available as a companion app on the Apple App Store, Garmin CIQ, and Google Play stores.
