As a followup to a Garmin 705 post earlier, I discovered a serious flaw in the way TrainingPeaks calculates the average speed, average power and average cadence when using a Garmin 305 (obviously I can't calculate out the power discrepancy but I'm positive it would do it the same way). The issue is that TrainingPeaks CANNOT handle variable time-rate data like the Garmin devices produce when they are set to the "Smart Recording" feature. The best way to illustrate this is just with some numbers, taken from a recent ride. I took some data right after starting from a dead stop at a light, and the Garmin had a widely varying timebase for the first 30 seconds or so. Here's the data:
8:58 2.152mi 0mph 0cad 2sec
9:00 2.157mi 12.7mph 51cad 2sec
9:02 2.164mi 11.4mph 56cad 2sec
9:04 2.174mi 14.8mph 78cad 2sec
9:12 2.207mi 16.8mph 24cad 8sec
9:22 2.255mi 17.3mph 83cad 10sec
Now if you look at the data, the correct way to calculate average speed would be to take the speed for each interval, multiply by the length of the interval and then divide by the total number of seconds. If you do this you end up with an average speed of 14.82mph. TrainingPeaks however reports 12.2mph!!! It's off by a whopping 21.4%! Likewise calculated correctly the average cadence is 53.5 and TrainingPeaks reports 58. This is a big screwup on their part, especially when the Garmin reports data on anywhere from 2 to 15 seconds...TrainingPeaks treats these as though they are all on the SAME timescale.
The other issue that is causing problems is that due to the above f'up, TP also treats every single point the same in timescale. This isn't a problem if you are recording in 1 second intervals, because the "zeros" when you stop at a light are a very, very small proportion of the total ride. In the case of the above example ride, I had 12 datapoints at 0mph due to stops at beginning and end, and lights along the way. Out of a total of 638 recorded datapoints this effectively runs down the average speed by 1.9% (12/638). When recording at a 1 second interval it would be 12/3860 = 0.3% change.
No wonder TP reports my average speed on a 100 miler as 18.3mph while Garmin's Training Center reports it as 19.5mph...other datapoints are 18.1/17.6; 16.8/16.4; 18.8/18.0. It's consistently but irregularly off by a pretty significant margin, anywhere from 2.4% to 6.6% from just a few rides. My conclusion from this is that TP is totally useless for Garmin data when taken at the "Smart Recording" feature. TrainingPeaks needs to fix this bug immediately, they have lost my business until they do.
8:58 2.152mi 0mph 0cad 2sec
9:00 2.157mi 12.7mph 51cad 2sec
9:02 2.164mi 11.4mph 56cad 2sec
9:04 2.174mi 14.8mph 78cad 2sec
9:12 2.207mi 16.8mph 24cad 8sec
9:22 2.255mi 17.3mph 83cad 10sec
Now if you look at the data, the correct way to calculate average speed would be to take the speed for each interval, multiply by the length of the interval and then divide by the total number of seconds. If you do this you end up with an average speed of 14.82mph. TrainingPeaks however reports 12.2mph!!! It's off by a whopping 21.4%! Likewise calculated correctly the average cadence is 53.5 and TrainingPeaks reports 58. This is a big screwup on their part, especially when the Garmin reports data on anywhere from 2 to 15 seconds...TrainingPeaks treats these as though they are all on the SAME timescale.
The other issue that is causing problems is that due to the above f'up, TP also treats every single point the same in timescale. This isn't a problem if you are recording in 1 second intervals, because the "zeros" when you stop at a light are a very, very small proportion of the total ride. In the case of the above example ride, I had 12 datapoints at 0mph due to stops at beginning and end, and lights along the way. Out of a total of 638 recorded datapoints this effectively runs down the average speed by 1.9% (12/638). When recording at a 1 second interval it would be 12/3860 = 0.3% change.
No wonder TP reports my average speed on a 100 miler as 18.3mph while Garmin's Training Center reports it as 19.5mph...other datapoints are 18.1/17.6; 16.8/16.4; 18.8/18.0. It's consistently but irregularly off by a pretty significant margin, anywhere from 2.4% to 6.6% from just a few rides. My conclusion from this is that TP is totally useless for Garmin data when taken at the "Smart Recording" feature. TrainingPeaks needs to fix this bug immediately, they have lost my business until they do.