Getting a Grip on Training Stress Balance (TSB)

31 10 2010

Some background in the previous post.

OK. Fine. I needed to learn more so I dove into Dr. Banister’s (et al.) Training Impulse and Response Model (TRIMPS). Which basically looks like this:

Dr. Banister's TRIMPS model

Eq.1

I learned about the gain factors ka and kf,, the timing constants, etc. And ground myself silly getting this thing to work in MS Excel. OK great.

I then learned that Dr. Andrew Coggan gave us TSB = CTL – ATL. This exponentially weighted moving average (EWMA) model looked something like this:

Equation for EWMA

Eq.2

Note: here’s the proper source for the equation for λ=2/n+1.

Equation #2 was much easier to slam through MS Excel. Both CTL and ATL use equation #2 with the difference that CTL uses a period of 42 days, and ATL a period of 7 days. Lambda equals 0.046 and 0.25 respectively.

In my previous post I rambled on about graphing some of this stuff in effort to quantify my training effort for racing. So I updated my data and ended up with a graphical method to guesstimate where my performance would be. I’ve only two months worth of PowerTap data, but it looks like this:

Screen picture of the performance estimate chart

The yellow line (CTL) tracks my historical training score over the default 42 day period (y-axis left side). The purple line (ATL) tracks my training balance (TSB) over time (y-axis right side). The purple splats (asterisks) are daily TSS averages displayed by week. What the graph tells me is that my balance is (currently) slightly negative with my CTL steadily increasing over time.

So what does all this mean?

Accordingly, (and knowing that implementing this stuff is half science, half art), I should be ready for a really good effort this Monday.

BTW, if you can afford the ~$129 cost, get a license of WKO+ software. It will track all this for you and you won’t have to pull your hair out trying to make it work.

So the challenge is twofold:

  1. To figure-out what ATL period setting best represents my physiology. That is to say, the recent trend of the TSB should correlate with how I’m feeling performance wise. This means I get to plug-in different period values and observe how close the graph tracks recent performance. (Doesn’t sound that easy hmm?). That’s the half-art portion.
  2. To figure out what training impulse (read intensity and duration) I readily respond to without burying myself in fatigue.

Anyway, I’ll throw-up (jeez, I hope that wasn’t foreshadowing) another post on training results in the near future.

You don’t know until you measure…right?

Advertisements

Actions

Information

8 responses

4 11 2010
Jeff

This idea of measuring, tracking and planning has really got me thinking about investing in a power meter and it’s benefits of training progression. Thanks the the article and food for thought.

5 11 2010
Eric Abbott

Jeff,
Purchasing my PowerTap was the best decision I’ve made towards improving as a road racer! I don’t have to guess as the numbers are right in front of me.

“You don’t know until you measure” is one of my favorite quotes and guidelines.

Thank you for commenting and taking the time to read my article!
-Eric

20 11 2010
Diandio
20 11 2010
raz-siood.net

Stress causes the body to release certain chemicals. Cortisol and adrenaline, which are normally issued in extreme situations.

20 11 2010
Eric Abbott

Yes it does. Thanks for your comment!

1 06 2011
Kai

Eric,

thanks for this blog.

The hint with the EWMA did the trick for me. I wanted to spare the WKO+ software and so far I used to have a moving average over my TRIMP values, but with the EWMA the recovery looks much better now.

But on the calculation on lambda you lost me a bit.

First I notice an equation error you refer to 7 days for ALT with a value of ,25 which is actually 9 days according to your formula.
But than digging deeper I had troubles with the formula as such, so
I finially decided to use do an simulation in excel to see for which Lambda the sum of all EWMA (integral of EWMA) till the day (CTL 42 and ATL 14) for one training load with 100 units cross the 95%.
I got as a result lambda_ctl42 = 0,06885 and lambda_atl_14 = 0,1927.

Thanks again, for your help and maybe you hint me where you got the formular for lambda. So maybe my approach is wrong here even though my graphs are looking reasonable 😉

1 06 2011
Eric Abbott

Kai,
Thank you for your response. My source for the equation is found here:
http://en.wikipedia.org/wiki/Moving_average
Refer to the section on exponential moving average, where lambda is expressed in terms of “n” time periods.

I’m also correcting my blog article on the topics…thanks again.

Some background:
http://home.trainingpeaks.com/articles/cycling/the-science-of-the-performance-manager.aspx#http://home.trainingpeaks.com/articles/cycling/the-science-of-the-performance-manager.aspx

13 05 2013
My Website

I absolutely love your blog and find a lot of your post’s to be exactly what I’m looking for.
Do you offer guest writers to write content available for you?
I wouldn’t mind publishing a post or elaborating on a number of the subjects you write with regards to here. Again, awesome weblog!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s




%d bloggers like this: