Critical Power-10 minute and 20 minute discrepancy

[whoawhoa]

Over the past few months I have noticed a significant (usually around 10+ watts) difference in my predicted and actual 20 minute power using a 3 and 10 minute critical power test. Or, put another way, a difference between my predicted 10 minute power and actual 10 minute power when using a 3 and 20 minute critical power test. My actual 10 minute power always leads to a predicted 20 minute power that is unsustainable.

Example:
3-min: 450 watts
10 minute: 390 watts (predicted 10 using 3 and 20=378)
20 minute: 360 watts (predicted 20 using 3 and 10=380)

My theory is that the amount of standing up and sprinting I do towards the end of the tests effect it. In both the 10 and 20 minute tests I began to stand up more towards the end of the test as I fatigue, and finish the test with 20 or 30 seconds all out sprinting. I feel like this affects the average of the shorter test way more and throws off the Critical Power calculations. Anyone have a beter idea?

 

[Alex Simmons]

I'd say that 3 min is at the low end of the CP test duration range and 10 min is not long enough and if you can sprint for 30 seconds at the end, you haven't gone hard enough.

 

[daveryanwyoming]

Over the past few months I have noticed a significant (usually around 10+ watts) difference in my predicted and actual 20 minute power using a 3 and 10 minute critical power test. Or, put another way, a difference between my predicted 10 minute power and actual 10 minute power when using a 3 and 20 minute critical power test. My actual 10 minute power always leads to a predicted 20 minute power that is unsustainable.

Example:
3-min: 450 watts
10 minute: 390 watts (predicted 10 using 3 and 20=378)
20 minute: 360 watts (predicted 20 using 3 and 10=380)

My theory is that the amount of standing up and sprinting I do towards the end of the tests effect it. In both the 10 and 20 minute tests I began to stand up more towards the end of the test as I fatigue, and finish the test with 20 or 30 seconds all out sprinting. I feel like this affects the average of the shorter test way more and throws off the Critical Power calculations. Anyone have a beter idea?

I take it you mean a Monod CP test. If so, I'd take at least three points not two. With two points you'll always get an R^2 of 1.000 which doesn't give you any sanity check on the fit of your data points. Add a third or fourth point and you at least see how well the data fits the model.

In general if you underestimate your short power during a Monod test you'll overestimate your CP or power at long durations. That's counterintuitive, but how the model works. From the example you listed it looks as if your AWC is underestimated and your CP is overestimated. I'd start by making your long test longer than 10 minutes and adding a third test point.

Can't say if standing up is altering your data, but I wouldn't be surprised, you're drawing on additional muscle groups and delivering power that won't be sustainable for long efforts(unless you can stand and deliver power for a long time, most of us can't) but using that data to predict long efforts. Gotta believe that alters things a bit.

-Dave

 

[Piotr]

Over the past few months I have noticed a significant (usually around 10+ watts) difference in my predicted and actual 20 minute power using a 3 and 10 minute critical power test. Or, put another way, a difference between my predicted 10 minute power and actual 10 minute power when using a 3 and 20 minute critical power test. My actual 10 minute power always leads to a predicted 20 minute power that is unsustainable.

Example:
3-min: 450 watts
10 minute: 390 watts (predicted 10 using 3 and 20=378)
20 minute: 360 watts (predicted 20 using 3 and 10=380)

My theory is that the amount of standing up and sprinting I do towards the end of the tests effect it. In both the 10 and 20 minute tests I began to stand up more towards the end of the test as I fatigue, and finish the test with 20 or 30 seconds all out sprinting. I feel like this affects the average of the shorter test way more and throws off the Critical Power calculations. Anyone have a beter idea?

Could it be that you're more anaerobically inclined then the Monod model assumes? That would be my guess. How does your tested vs. predicted FTP stack up?

 

[whoawhoa]

I'd say that 3 min is at the low end of the CP test duration range and 10 min is not long enough and if you can sprint for 30 seconds at the end, you haven't gone hard enough.

Not going hard enough is definitely not the issue-I usually begin to stand when I am too exhausted to produce power while seated. So not a sprint, per se-just 20-30 seconds in the 500 watt range completely and utterly crosseyed.

 

[whoawhoa]

maybe this would be a better way to illustrate my point:

using my peak 10 minute value of 390 watts, and using my peak 20 minute value of 360 watts:

my 3 minute power would have to be 550 watts, giving me an AWC of 629, to line all three points up. Clearly unrealistic.

Using my 3 and 20 minute values, my ftp and AWC look a whole lot more realistic. Only problem is, my 10 minute value is estimated 378-so could the added muscle recruitment of standing throw in an extra 12 watts?

 

[daveryanwyoming]

...-so could the added muscle recruitment of standing throw in an extra 12 watts?

Well, if you take say 2:20 at 450 watts and the last 40 seconds at 500 watts the average will be ~460 watts or up about 10 watts from your average for the rest of the effort. Move those times and power numbers around a bit and it doesn't seem tough to get a 12 watt boost by standing. The Monod tests should be steady sustainable power for each time duration. Not sure how well the model holds up to submaximal efforts with a big anaerobic effort tacked onto the end.

 

[frenchyge]

Not going hard enough is definitely not the issue-I usually begin to stand when I am too exhausted to produce power while seated. So not a sprint, per se-just 20-30 seconds in the 500 watt range completely and utterly crosseyed.

Hopefully you're doing the same for the 3-min test, except I'd give it a good minute or more of coss-eyed standing on that one. That would be the point that's off by the greatest amount if you're not draining the tank completely.

 

[waterrockets]

If so, I'd take at least three points not two.

+1 put the three values for 3m, 10m, and 20m into the model and tell us how it looks for 5m, 15m, and 60m compared to actual or expected.

 

[RChung]

I take it you mean a Monod CP test. If so, I'd take at least three points not two.

I'm not a fan of Monod estimates because they can be pretty sensitive to how one collects the data (as the OP has found). Nonetheless, if one is intent on using the method, I'd say it's better to have two carefully collected data points than three not-so-carefully collected ones. Thus, rather than going out and collecting another data point and adding it to two suspicious points, I'd push both the left and right hand points out a bit and try to collect them as consistently as possible.

 

[rmur17]

I'm not a fan of Monod estimates because they can be pretty sensitive to how one collects the data (as the OP has found). Nonetheless, if one is intent on using the method, I'd say it's better to have two carefully collected data points than three not-so-carefully collected ones. Thus, rather than going out and collecting another data point and adding it to two suspicious points, I'd push both the left and right hand points out a bit and try to collect them as consistently as possible.

just a point: IIRC Dr. Coggan and possibly others have cautioned against dropping less than 3-min on the short e/o the tests and going beyond 30-min on the long end.

Go too short and you may be unable to fully exhaust AWCover the test duration; go too long and the assumed linearity between work and time may not hold.

I don't test CP/AWC regularly but when I do I stick to maximal 5 and 20 min tests. As a bonus, the 5 min test gives me another solid point on the power-profiling charts.

 

[RChung]

just a point: IIRC Dr. Coggan and possibly others have cautioned against dropping less than 3-min on the short e/o the tests and going beyond 30-min on the long end..

Yeah, I think I was one of them. That's what I meant by "push both the left and right hand points out a bit"; I think 3 minutes and 10 minutes are a tad short for this detestable kind of test. Four or five minutes is probably as low as I'd go for the left hand point, and beyond 18 - 25 minutes on the right hand point you get diminishing returns. Your 5 and 20 fit right in. The key is to do it the same way each time, e.g., don't do 3 and 10 minute tests one time and 5 and 20 the next.

 

[frenchyge]

That's what I meant by "push both the left and right hand points out a bit"

Ahhh, I had also misinterpreted you to say that the points should be pushed further away from each other.

 

[RChung]

Ahhh, I had also misinterpreted you to say that the points should be pushed further away from each other.

Yeah, on re-reading it I realized that I wasn't very clear.

The rules for getting good results from Monod testing are:

1. Whatever you do, do it the same way every time.

2. Adding a dubious third point to two carefully-collected points doesn't improve accuracy.

3. Two points farther apart from each other give more stable estimates than two points closer together.

4. Contingent on #3, the left hand point shouldn't be too short.

5. Even taking into account #3, there are diminishing returns to pushing the right hand point too far to the right.

 

[daveryanwyoming]

.... 2. Adding a dubious third point to two carefully-collected points doesn't improve accuracy. ....

Yes, but as it's difficult to differentiate between a "carefully collected" point and a "dubious point" wouldn't a third "carefully collected point" add to the accuracy? I don't see how a third point implies sloppy protocols but two points doesn't.

-Dave

 

[RChung]

Yes, but as it's difficult to differentiate between a "carefully collected" point and a "dubious point" wouldn't a third "carefully collected point" add to the accuracy? I don't see how a third point implies sloppy protocols but two points doesn't.

It may not. However, in your original suggestion to collect a third point, I didn't see anything that warned that collection protocols are critical. In my field, collecting good data is both hard and expensive so we do everything possible to make sure that the few data we do collect are collected consistently (even if "consistently" doesn't always mean the same thing as "carefully"). In essence, collecting more data is something you do when you're worried about sampling error. However, in a case like this I'd be a lot more concerned about non-sampling error. Consistency of collection means that even if the estimate is biased, you hope that you can make it biased in the same direction each time so you can still detect changes in the estimate. In a numerical estimation sense, the increase from two observations to three does almost nothing to reduce sampling error (you'd probably need an order of magnitude more observations to do anything about that) so you're better off putting that effort into better collection.

 

[daveryanwyoming]

... the increase from two observations to three does almost nothing to reduce sampling error (you'd probably need an order of magnitude more observations to do anything about that) so you're better off putting that effort into better collection.

Great answer, thanks. Your explanation was really helpful, n = 3 isn't appreciably better than n = 2 and you might as well put your efforts into collecting good data. That still remains problematic as it's hard to know if you've really put out your "best" effort for a given duration or just lost focus or couldn't push yourself to the end. I guess I see why you're not a big fan of the Monod approach.

Thanks,
Dave