"The 33"-Road Bike Racing - Power Normalizations- Lies, Damn lies and Stats

A recent thread got me interested in power normalization algorithms so today I decided to compare a couple for myself. I used Coggan’s normalized power (CNP) as well as an exponentially weighted moving average (EWMA). For the EWMA I used a time constant that yields a half-life value of ~30s (alpha =0.0465).

CNP
1) starting at 30 seconds, calculate a 30 second rolling average for power
2) raise the values obtained in step 1 to the 4th power
3) take the average of all the values obtained in step 2
4) take the 4th root of the number obtained in step 3

EWMA
http://upload.wikimedia.org/math/5/8/d/58d7ee77df3201aa5ea95c7a380e2a45.png

The Challenge

Warm-up followed by a work period that includes 2x10min work periods that both average 250W -but differ significantly in the amount of strain. Furthermore, the efforts were bracketed with 5min@250W for calibration purposes. All efforts were preceded with 5min@150W to establish baseline.

http://i168.photobucket.com/albums/u164/Enthalpic/PNworkload.jpg?t=1198473508

The Results – Heart rate

http://i168.photobucket.com/albums/u164/Enthalpic/PNHR.jpg?t=1198473464

Power Normalizations

http://i168.photobucket.com/albums/u164/Enthalpic/PNormalizations.jpg?t=1198473738

Summary Statistics



Workload
Average 198.1
Std Dev 52.3
%RSD 26.4

30s Average
Average 198.1
Std Dev 50.1
%RSD 25.3

EWMA
Average 198.1
Std Dev 44.7
%RSD 22.5


Normalized Power
217.0

Heart Rate
Average 139.8
Std Dev 16.1
%RSD 11.5



Correlation coefficients



Correl HR-30s 0.673678
Correl HR-EWMA 0.872638
Correl HR-Workload 0.721307


Conclusions

Obviously a straight 30s moving average of workload does not correlate as well with physiological stress as does the EWMA. The plot of EWMA is very similar to the HR response plot indicating that this algorithm really does approximate real-time physiological strain well. Interestingly, you can see that HR remains the ultimate normalizer, as the %RSD of that measurement is the smallest. It is almost humorous that the EWMA of power seems to be a backwards calculation of HR; a measure that some power disciples want to completely ignore.

However, Coggan’s normalized power does a very good job at increasing the strain estimate of variable workouts as confirmed by the HR-Power calibration plot show below (done last month). The entire workout yielded an average HR of 139.8bpm, which agrees well with the NP of 217W unlike the average power of 198.1W.

http://i168.photobucket.com/albums/u164/Enthalpic/MAP8Nov07.jpg?t=1198476709

Both algorithms have their advantages. EWMP better matches instantaneous strain, and is more elegant, however, CNP matches strain over the course of the entire workout very well.

ElJ I suggest you use try the EWMA algorithm as a normalization protocol in your most excellent TT optimization program. The syntax is actually easier than NP.

bored last night?

Well.. in order to agree with you I think I'd have to assume that HR vs Power is a constant relationship, which... well.. in my experience it is not, soooo.... therefore, all I can do is nod and look nicely at your graphs.

ElJ I suggest you use try the EWMA algorithm as a normalization protocol in your most excellent TT optimization program. The syntax is actually easier than NP.

Will do, thanks for the analysis. It'll be a few days until I'll be able to throw that in there, though.

You seem to spend a lot of time trying to find fault in coggan's methods and analysis software but for what reason? If you do not like his methods then do not use them. I am sure someone out there is using a philosophy you do agree with. I really don’t think anyone is going to suddenly view you as an exercise physiology messiah and hand you a grant for your tireless efforts on BF.

*written by some one to lazy to look any further into his files than the main page of cyclingpeaks so feel free to ignore

You seem to spend a lot of time trying to find fault in coggan's methods and analysis software but for what reason? If you do not like his methods then do not use them. I am sure someone out there is using a philosophy you do agree with. I really don’t think anyone is going to suddenly view you as an exercise physiology messiah and hand you a grant for your tireless efforts on BF.

*written by some one to lazy to look any further into his files than the main page of cyclingpeaks so feel free to ignore

No, I respect Coggan's work. I seek understanding, not just the ability to use his stuff. Furthermore, the EWMA idea was recently brought up by Dr P.S. in another thread, so I played around with it.

Lastly, Coggan didn't develop a lot of this stuff. It's not like he invented math :rolleyes: He refined and expanded some existing ideas while making them much more mainstream. Good stuff.

Challenging authority makes us all better learners. He points out my mistakes, I learn. ;)

You seem to spend a lot of time trying to find fault in coggan's methods and analysis software but for what reason? If you do not like his methods then do not use them. I am sure someone out there is using a philosophy you do agree with. I really don’t think anyone is going to suddenly view you as an exercise physiology messiah and hand you a grant for your tireless efforts on BF.

*written by some one to lazy to look any further into his files than the main page of cyclingpeaks so feel free to ignore

Read his signature quotes. ;)

Ok, can someone please explain what normalized power means. Preferably without using calculus as i haven't taken that class yet. I understand SD but have no idea where normalized power comes from.

Ok, can someone please explain what normalized power means. Preferably without using calculus as i haven't taken that class yet. I understand SD but have no idea where normalized power comes from.

http://www.cyclingpeakssoftware.com/power411/defined.asp

Normalized power is a wonderful tool that allows people to brag about their numbers, while conveniently leaving out that fact that their power output was normalized rather than average, therefore giving others the false impression that their power output matches that of many pro tour riders.

EDIT: I don't ride with a power meter, so I may very well be wrong. :)

Normalized power is a wonderful tool that allows people to brag about their numbers, while conveniently leaving out that fact that their power output was normalized rather than average, therefore giving others the false impression that their power output matches that of many pro tour riders.

EDIT: I don't ride with a power meter, so I may very well be wrong. :)

Yep... you are.

Yep... you are.

So there is absolutely no truth to what I said, at all? ;) :(

I was being sarcastic, but I know that normalized power yields higher numbers than average power, hence my comment.

... but I know that normalized power yields higher numbers than average power, ...

Even that's not right.

Even that's not right.

It isn't?

It isn't?

Normalized power can be higher, lower, or equal to average power. It is true that normalized power is higher than average power much more often than it is lower.

It can, but just cause it's normalized doesn't mean it's going to. The point of normalization (whether using the oft accepted Coggan method or something you put together yourself) is to more accurately identify the physiological strain of an effort or workout.

If you do a workout that avg's 220 watts, normalized at 225, it will be different than one that is avg 220 and normalized at 300. Anyway... if you're using normalized watts to compare yourself to pro average watts, you're an idiot. Not that there aren't enough of those around here.

Normalized power can be higher, lower, or equal to average power. It is true that normalized power is higher than average power much more often than it is lower.

I know it's possible for it to be lower, but in all of the real world situations I've seen it, it's always higher by a pretty significant amount. I've seen it be close to equal on uphill TT's by riders of the same weight with very close times.

Coming off of Cross season, I know the differences can be huge in a cross race, where there is a constant fluctuation of explosive power/no power. I remember differences of 50 watts (between average and normalized) over the course of an hour long race, which is pretty significant.

In terms of gathering race date (and again, this is only our of curisoty, since I don't have a power meter) which would be more applicable; normalized or average?

It can, but just cause it's normalized doesn't mean it's going to. The point of normalization (whether using the oft accepted Coggan method or something you put together yourself) is to more accurately identify the physiological strain of an effort or workout.

If you do a workout that avg's 220 watts, normalized at 225, it will be different than one that is avg 220 and normalized at 300. Anyway... if you're using normalized watts to compare yourself to pro average watts, you're an idiot. Not that there aren't enough of those around here.

I've never heard anyone do it, but I'm sure people do.

How close are the power readings from a PowerTap and an SRM?

I don't use a powertap, but I've strapped my bike to a Dyno before. Let me show you:
http://www.dragtimes.com/images_dyno/12134-1997-Toyota-Supra-Dyno.jpg

819ft/lbs of torque at my rear wheel. Pretty impressive, no?

In terms of gathering race date (and again, this is only our of curisoty, since I don't have a power meter) which would be more applicable; normalized or average?

Normalized power more accurately reflects the physiological strain/effort required. Average power is actual work done.

So, we expect that two workouts with normalized power of 300 will induce the same amount of stress/fatigue on the body for the same rider. That says nothing about adaptation/muscle recruitment required to do it or gained from it - just the effect of the effort on the body in terms of stress. It will also help display for other people the type of ride done in terms of variability. You know that if I claim a ride average watts 250 and normalized 255 that it was a reasonably steady state effort. If I claim a ride average watts 250 and normalized 325 or 350, that the workout had some pretty large efforts at higher power.

Normalized power more accurately reflects the physiological strain/effort required. Average power is actual work done.

So, we expect that two workouts with normalized power of 300 will induce the same amount of stress/fatigue on the body for the same rider. That says nothing about adaptation/muscle recruitment required to do it or gained from it - just the effect of the effort on the body in terms of stress. It will also help display for other people the type of ride done in terms of variability. You know that if I claim a ride average watts 250 and normalized 255 that it was a reasonably steady state effort. If I claim a ride average watts 250 and normalized 325 or 350, that the workout had some pretty large efforts at higher power.

I think this is the most reasonable explanation for normalized power's application.

In something like a 'cross race (or a mountain bike race) where an overwhelming majority of the time it's a solo effort (unlike a road race or criterium where you can sit in) yet you have huge peaks and valleys in your power due to varying terrain and corners (unlike a solo TT) would average power serve as a good representation of weak spots in your race? For example, someone who manages to hammer on the descents and pedal harder into and out of corners (as opposed to someone who coasts down the descents and doesn't get on the gas early after cornering) would obviously see a closer relation between average and normalized... Would there be any advantage of using average or normalized to analyze these types of terrain-specific races?

I would expect that the hard pedaling in/out of corners and much more varied terrain would actually increase the variability of the power applied during race further separating the two values. (from not having raced cross, myself) Average power (for my riding at least) only really tells me valuable information about my riding in steady state intervals (of any length).

Once you induce terrain, wind, drafting, or anything that would cause you to vary your power output between significant numbers for significant lengths of time over a longer period (20+ minutes), normalized power tends to tell me more.

I'm not a 100% Coggan follower, but I can't find much reason to dispute/disagree with the normalized power methods or algorithm. You'd have to show me more than a HR graph from one ride to say that a different algorithm was more beneficial. On the other hand, I've found that the longer I ride with a power meter the less I depend on the resulting numbers and am back to riding a lot by feel this time of year. I still track the numbers as meticulously from week to week, I just don't get as wrapped up in them. It doesn't take a mathematical formula to tell me that the last two days of 4+ hr rides have made my legs tired and that I rode harder than I probably should have.

819ft/lbs of torque at my rear wheel. Pretty impressive, no?

Hm, I can only do about 200 ft-lbs. I'll have to try for some more.

I would expect that the hard pedaling in/out of corners and much more varied terrain would actually increase the variability of the power applied during race further separating the two values. (from not having raced cross, myself) Average power (for my riding at least) only really tells me valuable information about my riding in steady state intervals (of any length).

Once you induce terrain, wind, drafting, or anything that would cause you to vary your power output between significant numbers for significant lengths of time over a longer period (20+ minutes), normalized power tends to tell me more.

I know from experience that the difference between average and normalized during a cross race is pretty significant.

I guess my question boils down to an unlikely scenario, but this is the best I can explain it:

Imagine two riders of equal weight and very simmilar phsyical abilities on a 'cross course (lets just say the two riders have the same wattage-kilo but not the same bike handling skills) and both of them are using power meters. On the flats, they are probably going to be equal, but through corners/up steep climbs/and down descents the one with better handling skills is probably going to do better (pedal hard down the hill as opposed to coasting, pedal further into corners, brake late, and pedal sooner out of the corner or even pedal through the turn)

After the race, the two riders compare their power files. I know there are so many variables to factor in, but both of them might have a very simmilar normalized power, while the rider with better handling skills probably has much higher average power. I know in a race, the riders finish probably will speak more than the power file and I would imagine the more skilled rider would have won...

So, while analyzing race data (as opposed to training files) is there a possiblity that average power might be a better tool or indication of what type of real-world power you actualy put to the ground, rather than the physiological strain it placed on the individual? I would imagine that during training, especialy while training within certain "zones" that normalized could be a beter indication of your training, but would this still hold true for a race situation? I hope that all made sense, haha.

Again, I don't train or race with power (although I've borrowed a bike during a few cross races with a power meter, it's pretty fun being able to compare files of consecutive races) so this is all out of curiosity.

Hm, I can only do about 200 ft-lbs. I'll have to try for some more.

EPO

Won't help me at all, because I'm not doing any work at my max torque.

Regarding the scenario posted for the two riders of similar power outputs but different handling skills, I would actually expect the rider with the better handling skills to have lower numbers for average power and normalized power (given that they are otherwise pretty equal). There are several reasons for this including less wasted energy from the things you mentioned braking, not pedaling into corners etc. This is because each time the rider with poor handling skills slows down due to any of the listed reasons he must then apply more power to accelerate back up to speed. If two riders that have tested out to have similar Watts/kg have very different race performances and other factors are negligible then it would be reasonable to assume that one of them might be more experienced and be able to get more out of his power production by doing things like handling the bike well, choosing good lines, proper pacing...

Trying to think of an example... Think of it in a crit where our one rider knows his way around a bit and rides with the pack "sitting in" throughout the day not putting in much effort (actually probably coasting around 20% of the time) while our other rider dangles off the front all day with another rider who doesn't want to do much of the work. Then with a couple of laps to go the rider who has been sitting in bridges up and wins a sprint over the other two riders who have been out front. Because of the advantages of having drafted all day when you download the meters our rider who was sucking wheels all day will not have posted numbers anywhere near our guy who was out front (again assuming they are otherwise equal). In fact the only time during the entire race when his numbers will be better than our off the front man are probably during his bridge up and then the sprint neither of which would probably be long enough to raise his average or normalized power up to the level driving hard off of the front would.

Normalized power is very important in terms of understanding the demands that an event takes on the rider. Average power in mass start races would often lead you to believe that you put out very little effort and would have gotten a better workout on your own even though your legs, back and mind are exhausted and you know you road hard. This is because average power cannot accurately represent the demands on your body of the constant changes in pace, accelerations and so on, normalized power provides a means of measuring the actual impact on your body.

For the differences between an SRM and a Powertap you are talking around about 2.5%, enough that it would matter if you had one on a TT bike and the other on a road bike or trying to predict race performance for someone. But that would be what 7.5 watts at 300 watts? That just represents the efficiency of the drivetrain. For the individual user, using either one or the other all the time it makes no difference, its the repeatability that matters the most.

If you want to read a ton of material on using powermeters for all types of racing check out the wattage group on google groups. The amount of information available there on racing, and training with powermeters is impressive and free.

Oh and NomadVW, Semper Fi Mac.

^^^
Semper Fi!

VW

This is the stuff that makes the eyes roll back in my head. I'd throw the Power Tap off my cliff at this point, but the Schwag Factor won't allow me to do so.

Please Lord, give me the strength to plow through this incredibly dense bullsiht. Pcad was an art major in college boys. I got a BS degree from NYU, and in my case we all know what BS stands for. I'm dumber than all of you. Thank God for my good looks, wry wit and excellent dog.

Thanks for the insight :)

I think this is the most reasonable explanation for normalized power's application.
Actually, I think this is

Normalized power more accurately reflects the physiological strain/effort required.
Normalized power lets me track the physiological strain of my workouts better than time, distance, or any of the other common metrics. With this data, I can more accurately track the volume of training and control volume to allow the highest volume possible that doesn't lead to overreaching, injury, or illness.

Please Lord, give me the strength to plow through this incredibly dense bullsiht.

Don't bother; this won't help your training. This is for entertainment purposes only.

Don't bother; this won't help your training. This is for entertainment purposes only.

Which explains why I've gone from missing training 1-2 times a year due to illness before I started tracking TSS, to not missing a day since, while increasing my overall training load. But if you choose not to use the data, then of course it won't help your training.

I should have mentioned my best character trait: Humility.

Coggan didn't develop a lot of this stuff. It's not like he invented math :rolleyes: He refined and expanded some existing ideas while making them much more mainstream. Good stuff.

Challenging authority makes us all better learners. He points out my mistakes, I learn. ;)

Then here's another chance for you to learn ;): while TSS and the Performanc Manager are both derived from Banister's work, the normalized power algorithm is, AFAIK, a completely novel idea.

BTW, how well does an exponentially-weighted moving average of power track with heart rate during extended efforts requiring >100% of VO2max (answer: it doesn't, since heart rate is 'capped' at maximal heart rate).

Downloaded my first file into the Saris PowerAgent software on my Mac. My ride was busted up into 7-8 sub-rides. I 'merged' them (or so I thought) and wound up with two files instead of 8. I cleared the data off the PowerTap and I'll try again next time. I was able to set myself up as a user (I have no idea where to set the power ranges, HR zones, etc) and successfully download an actual ride file. So that's a start.

I don't know how bad the Saris software sucks compared to PowerAgent, but it does appear to work after a fashion. Made some pretty cool looking graphs. Don't even ask me to post them here, they're far too embarrassing.

This is like building an Iranian A-Bomb.

http://www.stopahmadinejad.com/wp-content/uploads/2007/01/mullah-bomb.jpg