Distorted max 20 min in ride

peterwright said:

Had an interesting one today where I looked at my max 20 mins in a long hard ride - it seemed very high at 317NP but with an AP of only 255. When I looked in more detail it had taken the last 9 mins of a climb then the downhill part of 4 mins followed by another hard climb of 6.5 mins and bundled this all together and estimated this very high NP level.

What do you make of this - it has given me a peak 20 of around 15w better than I have done before but I assume it to be meaningless due to it taking the rest period between the climbs and assuming it was climbing.

Any thoughts ?

Peter

No, that sounds reasonable, assuming the AP is correct. Still only a variability index of 1.25-not unlikely for that terrain.

peterwright said:

Had an interesting one today where I looked at my max 20 mins in a long hard ride - it seemed very high at 317NP but with an AP of only 255. When I looked in more detail it had taken the last 9 mins of a climb then the downhill part of 4 mins followed by another hard climb of 6.5 mins and bundled this all together and estimated this very high NP level.

What do you make of this - it has given me a peak 20 of around 15w better than I have done before but I assume it to be meaningless due to it taking the rest period between the climbs and assuming it was climbing.

Any thoughts ?

Peter

I think you might be expecting too much of the normalized power algorithm. 🙂

Look at it this way: you rode really hard for 20 min, but your average power was ~50 W, or ~16%, lower than what you know you can sustain for that duration. The normalized power, OTOH, was ~15 W, or ~5%, higher than your 20 min best. Which do you feel is a better estimate of how hard you were riding for that period of time?

acoggan said:

I think you might be expecting too much of the normalized power algorithm. 🙂

Look at it this way: you rode really hard for 20 min, but your average power was ~50 W, or ~16%, lower than what you know you can sustain for that duration. The normalized power, OTOH, was ~15 W, or ~5%, higher than your 20 min best. Which do you feel is a better estimate of how hard you were riding for that period of time?

I am just not sure that 2 hard hills with a rest in between represents a steady hard 20 min best effort ..especially when I knew it was 2 very distinct segments as opposed to one hard effort. It seems like the NP algorithm just took the power for the 2 hills and because they were close together it also bumped up the rest period in between and gave me a very nice 20 min NP which is not (imo) representative.

peterwright said:

I am just not sure that 2 hard hills with a rest in between represents a steady hard 20 min best effort ..especially when I knew it was 2 very distinct segments as opposed to one hard effort. It seems like the NP algorithm just took the power for the 2 hills and because they were close together it also bumped up the rest period in between and gave me a very nice 20 min NP which is not (imo) representative.

The NP algorithm doesn't "bump" anything up. There are two key elements at the heart of the NP algorithm. One is the 30sec averaging of power. This is key because of the half-life of the physiological response. This dampens out a lot of the highly stochastic real-time power curve. The second key element is the 4th power weighting of the 30s averaged power. AP is a straight arithmetic weighting formula in which every observation is weighted equally. So, averaging 400w and 200w results in 300w. The 4th power weighting algorithm weights the 400w and the 200w at the 4th power, so if you had an equal duration (e.g., 10min) at 400w and at 200w you would end up with an NP of 341w ((400^4 + 200^4)/2)^0.25, or 13% higher than the AP. As to the validity of the 4th power weighting, it comes directly from the blood lactate curve (as a function of power). So, I guess it's valid if you believe that the blood lactate curve is a reasonable proxy for physiological response to intensity (power). I think it is very robust.

peterwright said:

Had an interesting one today where I looked at my max 20 mins in a long hard ride - it seemed very high at 317NP but with an AP of only 255. When I looked in more detail it had taken the last 9 mins of a climb then the downhill part of 4 mins followed by another hard climb of 6.5 mins and bundled this all together and estimated this very high NP level.

What do you make of this - it has given me a peak 20 of around 15w better than I have done before but I assume it to be meaningless due to it taking the rest period between the climbs and assuming it was climbing.

Any thoughts ?

Peter

Peter,
How far was this into the ride and how hard had you ridden before that 20min NP record? Andy has already commented ... essentially that there are certain cases where the algorithm may overestimate power. I'm always interested in there as I spent some time looking for ways to improve the calculation -- without any success at all I should add.

i think it will come down to: what was the contribution of AWC to your power over that 20 min period? And how much and and what speed can we restore AWC during sub-threshold riding?

From my own data my all-time 20min NP was only 10W over my best 20min AP. The same goes for longer durations - I can't say I've ever encountered a large disconnect between NP and perceived power over an extended interval. But I'm always interested to hear about them.

rmur

rmur17 said:

Peter,
How far was this into the ride and how hard had you ridden before that 20min NP record? Andy has already commented ... essentially that there are certain cases where the algorithm may overestimate power. I'm always interested in there as I spent some time looking for ways to improve the calculation -- without any success at all I should add.

i think it will come down to: what was the contribution of AWC to your power over that 20 min period? And how much and and what speed can we restore AWC during sub-threshold riding?

From my own data my all-time 20min NP was only 10W over my best 20min AP. The same goes for longer durations - I can't say I've ever encountered a large disconnect between NP and perceived power over an extended interval. But I'm always interested to hear about them.

rmur

Thanks for the insight all. Ric - it was 60 mins into the ride but the previous 60 mins had been moderate and this was the first climb of the ride - 12.5 mins @ 96% FTP (the last 6 forming part of the peak) then we pretty much cruised for 3.5 mins then we hit a real mother of a climb and I really hit it very hard (in fact I underestimated the length of it and rode the first 2 mins at 137% FTP and the whole lung bursting 6.45 mins of it at 126% FTP)

So I just have a problem understanding how this 20 min period can be a peak when there were 2 distinct high peaks with a distinct rest period in between.

If I look at my MP curve then the 20 min is a "blip" and I do not think that I could ride for 20 mins at an NP of 110% FTP.

peterwright said:

Thanks for the insight all. Ric - it was 60 mins into the ride but the previous 60 mins had been moderate and this was the first climb of the ride - 12.5 mins @ 96% FTP (the last 6 forming part of the peak) then we pretty much cruised for 3.5 mins then we hit a real mother of a climb and I really hit it very hard (in fact I underestimated the length of it and rode the first 2 mins at 137% FTP and the whole lung bursting 6.45 mins of it at 126% FTP)

So I just have a problem understanding how this 20 min period can be a peak when there were 2 distinct high peaks with a distinct rest period in between.

If I look at my MP curve then the 20 min is a "blip" and I do not think that I could ride for 20 mins at an NP of 110% FTP.

Correct me if I'm wrong, but it sounds like there are really two separate issues that have you scratching your head:

1) the fact that CyclingPeaks picked that double-humped effort as a single peak, and

2) the fact that the normalized power value for that effort overestimates (by about 5%) what you could actually sustain for that duration.

#2 is explained by the fact that, while the algorithm is surprisingly robust, it clearly ain't perfect. #1, OTOH, is simply the way CyclingPeaks works: the highest (peak) value for any duration is used, e.g., to draw the mean maximal power curve, regardless of how that value is obtained. So, a double-humped effort such yours may result in a "blip" on the curve regardless of whether it is based on average or normalized power.

acoggan said:

Correct me if I'm wrong, but it sounds like there are really two separate issues that have you scratching your head:

1) the fact that CyclingPeaks picked that double-humped effort as a single peak, and

2) the fact that the normalized power value for that effort overestimates (by about 5%) what you could actually sustain for that duration.

#2 is explained by the fact that, while the algorithm is surprisingly robust, it clearly ain't perfect. #1, OTOH, is simply the way CyclingPeaks works: the highest (peak) value for any duration is used, e.g., to draw the mean maximal power curve, regardless of how that value is obtained. So, a double-humped effort such yours may result in a "blip" on the curve regardless of whether it is based on average or normalized power.

Andy - you are correct on both counts. In terms of the algorithm, I do not think it is that far out - if I use Raps ready reckoner of 90% of max for any duration then I would be doing my 2x20 @ 285 - in fact I can do this on a good day and fresh legs.

Re the way CP collects data - it was just interesting to see how it collected this segment - thanks for clearing it up.

Cheers

Peter

peterwright said:

Andy - you are correct on both counts. In terms of the algorithm, I do not think it is that far out - if I use Raps ready reckoner of 90% of max for any duration then I would be doing my 2x20 @ 285 - in fact I can do this on a good day and fresh legs.

Re the way CP collects data - it was just interesting to see how it collected this segment - thanks for clearing it up.

Cheers

Peter

Peter,
thanks for the feedback. Sounds like you were pretty fresh when you hit those climbs and the effort was unusually high, i.e. a typical 'stretch' case for NP.

Andy, is it theoretically possible to adjust NP for these cases? Practically possible? Anything in the pipeline?

I have no ideas - only questions 🙂

rmur

peterwright said:

It seems like the NP algorithm just took the power for the 2 hills and because they were close together it also bumped up the rest period in between and gave me a very nice 20 min NP which is not (imo) representative.

That's not the NP algorithm, per se. 'Averaging' the power over that 20-min period produces a similar effect of flattening out the peaks and filling in the valley. Certainly you lose some resolution when you look at a single number (whether it be AP or NP) to represent what happened over an entire ride or portion thereof.

OTOH, as has been discussed, the body draws from its reserves during the peaks and replenishes them during the valleys. So, some degree of averaging better reflects what is truly happening to the rider (ala NP).

rmur17 said:

i think it will come down to: what was the contribution of AWC to your power over that 20 min period? And how much and and what speed can we restore AWC during sub-threshold riding?

If NP can be "tricked," this seems the most likely candidate. If a rider has a power profile with unusually high MPs at shorter durations relative to his FT, I can imagine a ride profile with many short anaerobic efforts greater than 15secs, followed by sub-FT efforts for recovery of anaerobic energy, thus maximizing his anaerobic capacity resulting in an NP above FT (assume 1hr ride). The only modification I can envision is to have a sliding scale weighting of efforts >FT as a function of duration. So, a 30s effort would not be given full weight but a 5min effort would be given full weight. This is testable.

RapDaddyo said:

If NP can be "tricked," this seems the most likely candidate. If a rider has a power profile with unusually high MPs at shorter durations relative to his FT, I can imagine a ride profile with many short anaerobic efforts greater than 15secs, followed by sub-FT efforts for recovery of anaerobic energy, thus maximizing his anaerobic capacity yet maintaining an NP at or below FT.

But the degree to which anaerobic capacity can be restored during sub-threshold efforts will depend on just how long and how far below threshold they are...which means that somebody with a high anaerobic capacity will have to recover easier/longer, which in turn will drag down the normalized power. The net effect is that riders with a particularly high anaerobic capacity don't seem to be any better at "busting the algorithm" than those with more average anaerobic capacity. Indeed, of the 30 or so such files I've collected so far, I can't identify any common characteristics (other than the fact that people were riding pretty hard).

rmur17 said:

Andy, is it theoretically possible to adjust NP for these cases? Practically possible? Anything in the pipeline?

There's always stuff in the pipeline, but whether it will ever come out the other end is another story. 😄

acoggan said:

But the degree to which anaerobic capacity can be restored during sub-threshold efforts will depend on just how long and how far below threshold they are...which means that somebody with a high anaerobic capacity will have to recover easier/longer, which in turn will drag down the normalized power. The net effect is that riders with a particularly high anaerobic capacity don't seem to be any better at "busting the algorithm" than those with more average anaerobic capacity. Indeed, of the 30 or so such files I've collected so far, I can't identify any common characteristics (other than the fact that people were riding pretty hard).

Andy

As a point of interest - how can I get a feel for or measue my AWC ? I have always seemed to have an ability to push deep into the red but then need a real decent recovery to get back to normal. In many races this has hindered me because I seem to have that ability to dig "too" deep to stay on on a climb for example but then pay a bigger price and get dropped over the top. Maybe this is just poor pacing, but as I always race with the goal of giving it all I have to stay on the front bunch I cannot seem to hold back at these crucial moments. Is this to do with a high AWC (I have always been more fast twitch and have had to work hard to develop muscular endurance)

peterwright said:

Andy

As a point of interest - how can I get a feel for or measue my AWC ?

Two ways:

1) see how your 1 min power falls relative to your other abilities on the power profiling chart: http://www.cyclingpeakssoftware.com/profile.html

2) better still, using the critical power paradigm to calculate your AWC in J/kg - for a male road cyclist, ~250 J/kg might be considered about average.

peterwright said:

Andy

As a point of interest - how can I get a feel for or measue my AWC ? I have always seemed to have an ability to push deep into the red but then need a real decent recovery to get back to normal. In many races this has hindered me because I seem to have that ability to dig "too" deep to stay on on a climb for example but then pay a bigger price and get dropped over the top. Maybe this is just poor pacing, but as I always race with the goal of giving it all I have to stay on the front bunch I cannot seem to hold back at these crucial moments. Is this to do with a high AWC (I have always been more fast twitch and have had to work hard to develop muscular endurance)

Peter, you'll be pleased to know that my latest version of my matches application (due out today) has been conformed to the Critical Power Model. Now, if I can just get Andy to help me with the match recovery algorithm, we'll be cooking with nitromethane. Of course, I guess I have to get the courage to send him my current version.😄

acoggan said:

Two ways:

1) see how your 1 min power falls relative to your other abilities on the power profiling chart: http://www.cyclingpeakssoftware.com/profile.html

2) better still, using the critical power paradigm to calculate your AWC in J/kg - for a male road cyclist, ~250 J/kg might be considered about average.

Andy

Please can you give me some instructions on how to do this ? (critical power paradigm calc)

Thanks

Peter

RapDaddyo said:

Peter, you'll be pleased to know that my latest version of my matches application (due out today) has been conformed to the Critical Power Model. Now, if I can just get Andy to help me with the match recovery algorithm, we'll be cooking with nitromethane. Of course, I guess I have to get the courage to send him my current version.😄

Cmon Rap - get it to him - it is a superb piece of work and you know it 🙂

peterwright said:

Andy

Please can you give me some instructions on how to do this ? (critical power paradigm calc)

Thanks

Peter

http://lists.topica.com/lists/wattage/read/message.html?mid=903428277