Making sense out of apparent chaos analyzing data from on-bike powermeters
- 1. Making sense out of apparent chaos: analyzing data from on-bike powermeters Andrew R. Coggan, Ph.D. Cardiovascular Imaging Laboratory Washington University School of Medicine St. Louis, MO 63021
- 2. On-bike powermeters: both a blessing and a curse Powermeters provide a detailed (e.g., second-bysecond) record of a cyclist’s power, cadence, heart rate, etc., during each training session or race, but... 1. Multiple variables/seconds x 3600 seconds/hour x several hours/day x 365 days/year = a LOT of data!!
- 3. 2. Data are highly variable!
- 4. “Tools” for analyzing powermeter data 1) 2) 3) 4) Power profiling Normalized power Training stress score Quadrant analysis
- 5. “Tools” for analyzing powermeter data 1) 2) 3) 4) Power profiling Normalized power Training stress score Quadrant analysis
- 6. What is normalized power? Normalized power is an estimate of the power that a rider could have maintained for the same physiological “cost” if power had been perfectly constant (e.g., as on an ergometer) instead of variable.
- 7. Average power = 273 W
- 8. Kinetics of PCr resynthesis Coggan et al., J Appl Physiol 1993; 75:2125-2133
- 9. Half-lives of other physiological responses Power (force and/or velocity) (0 s) PCr kinetics ~25 s Heart rate/cardiac output: ~25 s Sweating: ~25 s VO2: ~30 s VCO2: ~45 s Ventilation: ~50 s Temperature (core): ~70 s
- 11. Data smoothed using 30 s rolling ave.
- 12. VO2, heart rate, lactate, and RPE as a function of power output VO2 Blood lactate RPE Heart rate 180 8 160 7 140 VO2max 6 120 5 100 4 80 3 OBLA 60 40 2 Lactate threshold 1 0 0 50 100 150 200 250 Power (W) 300 350 400 20 0 450 HR (beats/min) VO2 (L/min), lactate (mM), or RPE (U) 9
- 13. Blood lactate-exercise intensity relationship 20 y = 3.94x3.91 R2 = 0.81 Blood lactate (mmol/L) 18 16 14 12 10 8 6 4 2 0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 Power/power at lactate threshold Coggan, unpublished observations
- 16. Steps to calculate normalized power 1) smooth the data using a 30 s rolling average to take into account the time course of physiological responses 2) Raise the data obtained in step 1 to the 4th power take into account the non-linear nature of physiological responses 3) take the average of the values obtained in step 2 4) reverse step 2 to obtain the normalized power
- 17. Normalized power = 301 W
- 18. Relationship of average and normalized power to maximal steady state power Average power Normalized power Power during ~1 h race (W) 500 400 y = 0.93x + 27 R2 = 0.93 300 y = 1.27x - 126 R2 = 0.73 200 100 0 0 100 200 300 400 500 Maximal steady state power (W) Coggan, unpublished observations
- 19. Normalized power for 1 h (W) Relationship of normalized power to power at lactate threshold (Dmax method) 500 400 y = 0.88x + 51 R2 = 0.91 300 200 100 0 0 100 200 300 400 500 Power at lactate threshold (Dmax method) (W) Edwards et al., unpublished observations
- 20. Advantages of/uses for normalized power • Allows more valid comparison of races or training sessions with differing demands – e.g., hilly vs. flat training rides, criteriums vs. TTs, outdoor vs. indoor training • Helpful in the design of novel interval workouts – if normalized power for session (intervals plus recovery periods combined) exceeds athlete’s power-duration curve, unlikely that they will be able to complete workout as planned
- 21. Advantages of/uses for normalized power (con’t) • Can be used to assess changes in fitness w/o need for formal testing – normalized power from hard ~1 h race provides estimate of maximal steady state power • May prove to be useful constraint when attempting to model performance – e.g., to determine optimal TT pacing strategy
- 22. Limitations of normalized power • Essentially assumes that the net contribution from anaerobic ATP production is negligible – therefore not valid during shorter efforts in which contribution from anaerobic capacity is significant (e.g., individual pursuit) • Occasionally overestimates sustainable power – is the algorithm biased, or are such data just statistical outliers?