I apologize if I’m wrong, but I think there is some kind of inaccuracy in how the power curves are calculated.
When I’m looking at my power curve for the last 42 days, I see values of 251 W for 6 min, 253 W for 7 min, 245 W for 8 min. I don’t understand the spike at 7 minutes. If I managed to do 253 W for 7 min, I obviously did it over 6 min as well. The power curve should just be monotonously decreasing. I see such spikes at multiple locations in my power curve.
I assume that parsing activities to put values into the power curve happens on a very coarse time grid and different values (for 1, 2, 3 min etc) are populated independently of each other?
How about a very simple adjustment if you roll up the power curve from behind, whenever it is updated? If at a longer time span a higher value was detected than at a shorter time span, the value for the shorter time span should be automatically be set to the higher value (i.e., in my example the 6 min value should be set to 253 W).
It might be a bit pedantic but I’m somewhat bothered with those spikes
It is possible for your power for 7m to be better than for 6m. Consider a 10s sprint, 10s easy and another 10s sprint. The 30s average power will be higher than the 20s average power because it includes both sprints.
The power curves do not store every second (too much space) but they are much more granular than only every minute. It is more like every second for first minute, every 5s until 2m, every 10s until 5m etc…
Thank, you David. It makes sense, if you consider it that way.
I was just thinking about the power curve slightly differently… mainly in terms of continuous efforts.
Based on your example I thought if I achieved an average power over 30s, I would also be able to achieve the same average power over 20 one way or the other. Maybe this idea breaks down with very high anaerobic intervals.
Anyhow, I understand now, how these spikes are created and how to interpret them. Sorry for the inconvenience and please don’t see this as a bug report anymore. Thanks!
You might want to go look for spurious data in your workouts? I had my power meter flaking out on me recently so I had a bunch of spikes above my 1s power*. It was screwing up my power curve at the very high end.
It’s less likely to be the case for your 6 vs 7 minute power, but as David says the calculation takes in a number of variables, so maybe you have some weird spikes after minute 6 in one workout.
Power curve will show you the date of the workout, and then you can go and look for any jumps in the power curve. You can then use “actions-Ignore part” to filter it out.
You can add the “raw data” chart as well, might help to pick it out.
You may also have the inverse problem, which is the power cutting out earlier when it was counting the 6 minute power. I guess you could exclude that as well? It only takes one day of weird power meter readings to throw off the power curve.
*for those interested - make sure you tighten that SRAM/Quarq battery cover plenty tight!
edit: and kudos to @David for making this easy to track down and fix!
