Wednesday, March 25, 2009

Matchfinder

No, this isn’t a chat about online dating websites! It’s about a method to quickly identify when, during a race, you “burnt a match”.

The concept of a “burning a match” isn’t a new one in cycling – basically it’s a metaphor for saying you did a hard effort, hard enough that it might impact on your ability to do other hard efforts later on in the race, since a match can only be used once. While not a perfect analogy, it’s not a bad one.

How many matches we can burn and how brightly and for how long those matches can shine for might be thought of as the size and quality of our personal matchbook. It’s one measure of our race specific fitness.

Performing well is as much about knowing when to "light one of your matches" as it is about doing the training to build up the size and quality of your matches and the number of matches in your matchbook.

Burning a match is also relative to the race in question. In a 3-week Grand Tour, a match might be akin to a long solo/small group break away or an attack on an alpine ascent, but in a 20 or 30-km points race on the track, it would be an attempt to gain a lap, or go for lots of sprints.

Typically, burned matches are attempts at race winning moves
or, for some, race survival moves.
In any case, often riders like to analyse their power meter files to see when, how many and how big were the matches they set off while racing. It's one way to help assess, post-hoc, the tactical decisions taken while racing (did you really burn an unnecessary match going on a fruitless escapade?) or simply as a way to assess how many and how often matches were lit before you cracked, or even an indication that your matchbook is getting bigger with training.

In the book, Training and Racing with a Power Meter, it shows one way to locate such efforts by using the fast find feature in the WKO+ software. I’m going to show another method, one, that with just a little bit of spreadsheet help, is pretty easy to do and which shows up matches quite clearly.

Again, as with much of what I write, it’s not an original thought. It is based on the Normalised Power concept developed by Andy Coggan and just such a chart can be seen on Slide 15 of his PowerPoint presentation hosted on Google docs:
“Making sense out of apparent chaos: analyzing data from on bike power meters”


OK, so let’s look at an example of what I’m talking about.

Let's take my points race at the recent State Masters track championships. Here’s what the power meter trace looks like for the race:


It shows my power output for the race, as well as horizontal lines showing my Functional Threshold Power (FTP) and my Maximal Aerobic Power (MAP). As is typical with these sorts of races, the power output is highly variable and while you can see some spikes, it is difficult to make all that much sense out of the information presented like this.

But with a little bit of maths applied to the power data, here is a plot of exactly the same race:


Now this really shows up where I burned my matches. In this instance it clearly shows the 6 sprints in which I either contested (the first and the last) or simply needed to put the power down to stay in the race (the other 4). It also shows the periods where I didn’t sprint at all (sprints are every 10 laps in these races about every 2.5 - 3 minutes depending on the race speed), which is when I was in a chasing group as laps were being taken/lost by various riders.

By showing the data in this way, it is really clear when matches were lit. My first one was a pretty big flare, as not only was there a strong sprint but it was clearly an extended effort, perhaps covering an attempted attack after the sprint. But it also shows that after a couple of sprints, I simply didn’t respond when the inevitable attack came. I needed another match but my book was getting a bit thin at that stage.

Here are a couple of other similar plots:

This one is from the State Masters Criterium Championships in 2006:


As is evident from this plot, the first 10-minutes were pretty brutal with some very hard efforts necessary. In this period the field was whittled down to a break of just 6 riders. Then the break settled somewhat, before some more attacks started in an effort to ‘break the break”. It is also clear that on this course, if you couldn’t repeatedly make such hard efforts, you would be toast, limping back to the hotel for an early shower. It was a "repeatedly go hard and recover" kind of course.

Here’s another example from a different type of crit race:


This time I made a solo break very early, then was joined by another rider after about 15-minutes or so and we stayed together up to the finish. You can see the large match early on, and then the smaller efforts while solo, which diminished somewhat when I was joined by my break companion and we established and consolidated our lead. This enabled me to save a big match for when it really counted - the final sprint.

So how are the above plots made?

Well it’s not hard and if you know about Normalised Power, then you’re well on your way.

1. Just take a normal power meter file and open it in Excel (or your preferred spreadsheet software).
2. Then calculate a rolling 30-second average of the second by second power data.
3. Then raise that rolling 30-second power value to the 4th power (watts^4).
4. Then chart that 30-second power raised to the 4th power by time.

That’s it.

I also added lines to show both FTP and MAP raised to the 4th power.

Hint: The chart is a x-y scatter plot, with horizontal (x) values being time and corresponding vertical axis (y) values the power^4 values.


Why 30-second averaging and why raise to the 4th power?

Well, in essence the 30-second averaging and the raising to the 4th power is because (and I quote from Andy's own item on Normalised Power):

  1. the physiological responses to rapid changes in exercise intensity are not instantaneous, but follow a predictable time course, and
  2. many critical physiological responses (e.g., glycogen utilization, lactate production, stress hormone levels) are curvilinearly, rather than linearly, related to exercise intensity

As for #1, as Andy has shown us in the Google docs presentation referenced above, the half lives of many physiological responses to the intensity we are riding at (i.e. our power output) are indeed not instantaneous. The time period for such responses to show up is typically around 30-seconds or so. Some, and I quote, include: PCr kinetics, heart rate/cardiac output and sweating all having half life response times of around 25 seconds. VO2 ~ 30-seconds, VCO2 ~ 45 seconds, ventilation ~ 50-seconds and core temperature changes ~ 70-seconds.

So from the point of view of assessing our body’s responses to intensity (power output), it makes sense to view power meter data as a rolling average power over a 30-second window*. It doesn’t actually have to be 30-seconds but changing the duration of rolling average (to say 25-seconds or to 40-seconds) doesn’t have a large impact on the outcome of the plots.

* except perhaps when assessing maximal neuromuscular sprint type efforts, since the energy systems in use fatigue over a handful of seconds (although our "recoverability" for sprints is still linked to our aerobic or "matchbook" fitness).

This is also why we sometimes refer to things like heart rate as being a "lag indicator" of effort and is one reason why HR is a poor guide to managing shorter harder efforts while training.

As for the #2, the research Andy shows suggests an exponential relationship exists between blood lactate concentration and power expressed as a ratio of 1-hour power (power:FTP, often referred to as the Intensity Factor). In the same presentation one can see (on slide 13) that the best fit for the data shows a relationship very close to the 4th power (3.91).

Again, the use of a nice even number of 4 rather than say 3.9 is simply more convenient and choosing numbers either side really doesn’t affect the nature of the plot all that much.


So if you cracked in a race, or couldn't go with the winning break, then what does your match analysis look like? Did you not have a match when it counted, or did you not use them wisely enough?

Just remember, playing with matches is dangerous, so take care out there!

8 comments:

Anonymous said...

A masterful post Alex! If only my poor excel skills were up to really interpreting your 4 step instructions. I should have been been looking at my maths books at school instead of having my head in boys own adventure. Then I might have clue about 4th power values. cheer peter

Steve Palladino said...

Brilliant! This is an analytical keeper.

APBIORoswell said...

Hey Alex, Can you write a generic formula for rolling average and a comment for us?

Thanks for a great post!

Bob

Axel said...

Awesome. Now if this doesn't show up in the next version of WKO (which it should), I'll have to find out how to produce those funky graphs in Excel. It would get messy.

Andrew Coggan said...

I check with Hunter and, unfortunately, it sounds as if this way of plotting the data didn't make it into the next version. :-(

Axel said...

Well, there I go (I feel a headache coming on)

Rich Gift Of Lins said...

Alex,
I have now used this analysis for my first two road races of the season and it certainly helps by painting a picture of what went on. As I get more data, I am sure that it will help make me more circumspect about my "matches" in the future. Thank you for making the spreadsheet available, Colin

Alex Simmons said...

For anyone interested, I have placed a sample spreadsheet for both PT and SRM file formats on the Google Wattage Groups Files page.

It's called Matchfinder.xls