You can't touch this, Part III

NP Busters again today.

This is third part in a chat about Normalized Power and NP Busters.

In Part I, I reviewed the concepts of Average and Normalized Power and how and when they are useful tools for assessing your ride, the differences between them and how NP accounts for the highly variable nature of our power output as well as the non-linear relationship between the strain we experience and our power output, both things that are masked by inspection of Average Power alone.

In Part II, I expanded on how Normalized Power, well, normalizes rides by providing a reliable indicator (i.e. NP) of the metabolic strain experienced from rides of quite different types, and demonstrated this with an example by comparing a time trial with a criterium race by the same rider.

Part II also provided a definition for an "NP Buster", a term coined many years ago to indicate a ride with an NP somewhat higher than is typical or higher than what would normally be associated with the level of strain experienced. IOW, it significantly over estimates the rider's steady state power output capability. And by significantly, I don't mean double, but by more than 5%.

So, how common are NP busters, and what can we learn from them?

Just to quickly recap, an NP Buster is a ride (or part of a ride) of about an hour's duration where the NP is > 105% of a rider's well established Functional Threshold Power (FTP). There are a few caveats about accurate power measurement, correct application of the NP algorithm, and a valid FTP setting before we declare an NP Buster, but assuming those requirements are satisfied, then we can declare a buster.

Like cyclones, NP busters do happen, and there are places and times when they are more likely to happen. They are also relatively rare, especially when compared to the vast "weather system" of all rides performed. But unlike cyclones, they don't represent some vast destructive force to be feared, but rather provide an opportunity to learn something.

Over the years, of all the files from riders I've coached (and myself of course), I have seen maybe one or two true NP Busters. That represents less than ~0.01% of all rides. Of course I have seen multiple ride files with an NP > 1.05 but they don't qualify for one or more reasons as true busters.

Now it's entirely possible that I have a lower than normal representative sample of NP busters in all of my client's data, but even if NP Busters are under sampled by a factor of 10, that still means they only occur less than 0.1% of the time. For a large proportion of riders, they will just never happen. And that's because for a large proportion of riders, they are just not capable of generating a buster. More on that later.

Over the years the creator of the NP concept, Andy Coggan, did actively seek and collect files from those who had genuinely generated an NP Buster, and I have data on those 20 examples to share (thanks Andy). Yep, only 20 examples out of many tens, perhaps hundreds, of thousands of power meter files. OK, perhaps not a brilliant statistical indicator of their true frequency, but you get the idea that these are not exactly an every day occurrence.

As an exercise in testing his Normalized Power algorithm, Andy also issued a challenge in 2009 at the Google Groups wattage forum for riders to attempt to complete one of a series of suggested workouts, any of which would have resulted in an NP buster. There were few, if any, takers.

Below is a chart showing information about each of the NP Busters Andy had collected files for. Click on the chart to see a larger version.

Let me take you through what is shown.

Each horizontal green and blue bar on the chart represents an NP Buster. The green bars on the left show how far below FTP the Average Power of the NP Buster was. The blue bar on the right shows how far above FTP the Normalized Power of the NP Buster was. Hence, only rides with an NP more than 5% over FTP are shown.

On the right side of the chart I have also included the Intensity Factor (IF) and the Variability Index (VI) for each ride. IF is simply the ratio of NP to FTP (IF = NP/FTP). VI is the ratio of the NP to the AP (VI = NP/AP). So looking at the first bar at the top of the chart, the Average Power for the ride was 22% less than the rider's FTP, the Normalized Power was 5% higher than the rider's FTP, the IF was 1.05 and the VI was 1.34.

Now we can see something interesting when you look at all of the examples of true NP Busters. Even though the Normalized Power over estimates the rider's well defined capability of sustaining an equivalent steady state power output for about an hour, on most occasions NP is significantly closer to the rider's FTP than is Average Power, and by quite some margin. On only four occasions out of the twenty was the rider's AP closer to FTP than NP, and not by much. These four are highlighted by the red translucent boxes.

So even though all of these rides are NP busters (and hence by definition are extreme examples of stretching the NP algorithm) and are clearly nothing like steady state efforts (see the VIs), 80% of the time the NP was still somewhat closer to a rider's FTP than was AP.

So if you have a ride or part of a ride of about an hour with an NP > FTP (and the data and calculation of NP is valid) then it's very likely your FTP will be closer to NP than AP, even if it's an extreme case of an NP Buster.

Indeed, if you are still trying to work out your FTP, and you haven't really performed any testing or settled on a reliable means to establish it just yet, but do have some very hard one-hour ride/race data, then you can peg a reasonable initial estimate of your FTP to be somewhere in the region of 95-100% of the NP.

The other thing to note is if you are in the majority of people who can't or don't produce NP Busters, then charting your 60-minute mean maximal NP, for example with a periodic chart in WKO+ software, is a reliable means to track longer term aerobic fitness changes. Here's an example chart of quarterly progress in 60-minute mean maximal Normalized Power over two years:

Plotting progress with 60-minute mean maximal NP
is a handy way to track longer term fitness changes.

Can you generate an NP Buster?

Of course there are going to be people who see instances of NP Busters a little more frequently, and it comes down to two things:
- the physiological profile of the rider
- the type of rides/races they do

To generate an NP Buster you need to execute a ride which includes a lot of very hard efforts of 30+ seconds duration which are substantially higher than your FTP. Many riders simply do not have the physiological profile to do that, as it requires a rider to posses both high neuromuscular power and a high anaerobic work capacity, especially relative to their aerobic capabilities.

NP Busters often involve out of the saddle efforts that engage the upper body musculature to enable the high power outputs necessary to generate them. An example of the sort of ride where this is likely to occur is in a criterium, and in particular one where there is a 20+ second long hill and/or a U-turn to negotiate each lap. Of course you could design a training session with such efforts or try your hand at one of the challenge sessions posted by Andy Coggan.

The ride file shown below is an example of an NP buster candidate course - a U-turn at the bottom of a hill in a criterium. In this example the rider has an FTP of ~315W (shown by the horizontal dashed line) and was able to continually punch out of the corner hard enough and long enough to contest the race finale. Yellow line = power.

Races such a crits with U-Turns and hills are likely NP Buster candidates
for riders with the right physiological capabilities

To better illustrate the file, this is what it looks like when you apply a 30-second rolling average to the data. This makes it pretty clear how frequently and how hard the rider had to push themselves. And this was with a small break away group, not a large group of riders.

  Using a rolling 30-second average power trace
helps to see why a ride might have been an NP buster

In this case AP was 276W (13% less than FTP) NP was 354W (12% more than FTP).

Now whether we should include rides which involve out of the saddle efforts as examples of NP busters or not is a consideration, but since it's not an uncommon thing in bike racing (especially criteriums like the one above), then I figure we may as well since that's how bikes are raced.


OK, so if you are a rider that can generate an NP buster, then that tells you something about your unique capabilities as a rider, that is you likely posses both a good sprint and a high anaerobic work capacity. You are also in the minority that possess very potent race winning weaponry, provided your aerobic fitness is good enough to use it. But it also means that you'll need to take a little more care in how you choose to interpret the NP from such rides, and don't go immediately assigning yourself a high FTP on the basis of such rides.

For most everyone else, NP provides a robust and reliable means to assess the metabolic strain for a wide variety of rides, and if you see an NP > 105% of your FTP, then there is a very strong likelihood that you've under estimated your FTP and maybe it's time to validate though reliable test methods.

Read More......

Aero for slower riders

A quick chart today for future reference whenever that classic online nonsense argument about aero benefits only being for faster riders, or that aero only matters above a certain speed....

Let's set people straight now: Aerodynamic improvements benefit riders of all speeds and power outputs. But who gains the most benefit?

Whether a slower rider should be putting time/energy/effort/resources into gaining or buying an aero improvement when they might perhaps be better focussed on losing weight and training more (or harder or smarter) is a moot point. Really, though, such an argument is a false dichotomy. Why not do both?

The other consideration of course is if you are going to chase an aero improvement, then there are two main ways to achieve that:

• improved aerodynamic positioning, or
• improved aerodynamic equipment

But again, this is not a case of one or the other. It's quite OK to do both and train better. You know, one could train to improve fitness, work on gaining a better aerodynamic position, and treat themselves to some nice aero wheels, or move from using a road bike with clip on bars extensions to a time trial bike. This is not an either/or scenario.

If you are a back/middle of pack rider, then some bling wheels are not going to make you the next world champion, so some perspective here is warranted but the rationale for why you are looking to improve your performance is a matter of personal choice. If you want to be faster, then you do all the things you can given the constraints you have (time, money, knowledge, rest of life factors etc). And we are talking about people riding in competition-like events, not your cruiser to pick up some milk at the local shops (let's be sensible here).

If you are just happy with participating rather than competing, then sure, what does it matter? If you just like having nice equipment and have the money to spend, heck, go for it. Enjoy yourself.

But let's get the physics out of the way with a chart to quickly summarise the situation with an example.

The chart below plots the time taken to complete 10km on a flat road with no wind at various power outputs, from a modest 150 watts, through to a solid 350 watts. Other assumptions are shown on the chart, but changing the parameters really doesn't change the basic principles here. Click on the chart to see a larger version (right click to view in a new tab/window).

There are two lines, showing the reduction in time to complete the 10km as a rider's power output increases. No surprises there, more power with all else the same, you go faster.

The two lines also show the difference between a rider with a coefficient of drag-area (CdA) of 0.30m^2 and 0.27m^2 (a 10% reduction). That's roughly the sort of reduction in CdA you might expect going from standard low profile spoked road bike wheels to specialist aerodynamic wheelset, or riding on the tops to riding on the drops.

Under that are the blue columns, which represent the time saving over that 10km by reducing CdA from 0.30m^2 to 0.27m^2. As you can see, the slower less powerful rider saves more time in absolute terms than the faster more power rider. However, when expressed as a percentage of time saved, they are nearly equivalent savings, with the faster more powerful rider making very slightly better gains in percentage terms.

Now of course some parameters do change under some conditions, e.g. cross winds can affect the apparent CdA to differing degrees at different speeds, so in those situations, a faster rider may benefit a little more in percentage terms, but in general, there really is no physical reality to the old myth that aero only benefits the faster rider, or that óne needs to ride at X km/h to see benefit.

Pithy Power Proverb:

The largest absolute time savings from a given aerodynamic improvement are made by the least powerful/slowest riders.

Read More......

l'Alpe d'Huez. Again. Top 200

Some charts for fun. Pub chat, no more, no less.

Published on this Finnish forum are the top 200 ascent times for the Alpe d'Huez climb, often used in the Tour de France. I haven't been able to clarify how the times were established, nor if the same start and finish timing points were used. Official timing of the ascent started sometime in the 1990s but different timing points were used from 2001 onwards, so I cannot say with any confidence if we are comparing apples with apples. There are plenty on forums that definitely dispute Pantani's quoted climbing times (and that they should be somewhat slower than shown here, although still fast).

It's possible they reviewed footage to normalise these things, but in any case, nice work on collating the data, I'd say thanks personally but I don't speak or read/write Finnish. If any reader does, perhaps they can pass on my thanks.

Indeed racing context is also needed, e.g. Was it a long stage? Many previous cols? Attacking for the win or defending the maillot jaune? Conditions/weather/wind? Of course in 2004 the ascent was used for an individual time trial, not as the final climb of an alpine road race stage.

And yes, it's full of dopers and naturally covers the glory days when EPO, blood transfusions and other supporting cocktails were, sadly, the norm. Not that these things still don't happen, just seemingly not with the same outrageous impact on raw performance as before, at least not for the top of Pro Tour, but who knows about lower level riders trying to make the grade? Doping is still prevalent, and while the outrageous performance days may be suspended for now, the consequences are still just as insidious - shady characters and corruption, legitimate riders missing out on contracts and racing opportunities, people losing jobs, sponsors leaving, races and race results skewed and screwed, disillusioned fans. Long term health impacts. The list goes on.

As for conversion of ascent times to power to weight ratios, something that's been gaining in popularity lately with talk of mutants and the like as well as regular guesstimates published in online forums, well I cover that in this July 2010 post.

For reference though, using the methodology outlined in that post, the fastest time quoted in this list (whether or not accurate), would equate to a power to weight ratio of ~6.5W/kg +/- 0.4W/kg. Certainly not the nonsense level 7.2+ W/kg (or even 8W/kg) quoted by some.

In summary, just plot the ascent times and map the trends for individual climbs. Converting to a W/kg guesstimate may provide a way to make comparisons between climbs, but such estimates should be plotted with error bars, because there are too many unknowns in key assumptions and estimations are subject to methodological error. Converting actual performances to plot a wattage for a standard rider of 70kg makes even less sense, since W/kg is already normalising such estimates.

So here are two charts I whipped up from that Finnish forum data.

The first plots the fastest five riders each year. Click on it to see a larger version:

The next is a frequency distribution of the top 200 times for the years l'Alpe was raced.

What this means is that of the fastest 200 times recorded, 16 of them were in 1994, and so on. Quite clearly something changed between 1989 and 1991. One could speculate about other changes since then. Use of EPO, introduction of tests and of course doping detection avoidance measures, better tests, move back to blood transfusions as dominant in competition doping method and so on.

And a new chart, this one plotting the combined average speed of the top 5 riders (except for the first three years of data where there were fewer than 5 riders in the list for each of those years).

So, there you have it. enjoy the pub chat. And if you want to know how you compare, perhaps this other post about mere mortals might give you some beer for thought.

Read More......