Thursday, February 16, 2012

Analysis of pacing variation in the 2011 Eugene marathon

UPDATE: a better version of this analysis, and covering 2008 to 2011 is  in this posting


The Eugene marathon makes it easy to scrape all the results data off their site.

I wrote a python program to parse this data and calculate the first half vs second half splits.

I was curious to see if there was a difference between the men and the women..I had a theory that due to being less 'macho' the ladies might be a bit better at's the stats:

numberave % fade max % fade med % fade max % speedup
TOT(2340) 5.76 31.53 4.98-9.47
M(1222) 6.2031.534.97-9.47

The median fade was about the same for both men and women.


  1. This is very interesting...of course I am hoping to have very little fade, but it's hard to know what will happen. Maybe overconfidence is universal between sexes. I would be interested in the data for first time marathoners Vs. people racing 2-4 times or 5+ times.

    1. Yes, this is hard to figure out on your first.

      The running calculators that can predict marathon times from half marathon times are usually optimistic for noobs.

      In my case my half time predicted a 3:5x time, but I ran a conservative 4:25 and did not fade at all (but could tell I was right on the cusp of it).

      Eventually I figured out how not to fade...but it took while to get the pacing recipe figured out.

  2. Are you trying a triathon later? Is that what you meant by first?

    1. Yes, I am..if you look in my sporting goals you'll see that someday I want to do a HIM.

      Hooowwwever..this summary I'm going to try a sprint tri, like you.

  3. This is awesome info! You are great at calculating things like this Paul. To calculate your percentage fade, do you just divide your first half marathon time by your 2nd half marathon time? I think I had about a 2% fade then if I did the math right (1:57 first half, 1:59 2nd half). 2% fade is the largest bar above, so I am just like everyone else, lol

    1. No, I calculate it as a percentage of the marathon total time..i.e.

      My equations is:

      fade_pct=100.*(full_secs - 2.*first_half_secs)/full_secs

      In your case that quotient works out to about 120/14160 which is only about .8% fade. Not bad at all!

  4. Wow, you really are a techy geek!! I love it! :)

  5. Very interesting. I've seen a similar men/women comparison for the Comrades marathon in SA. The women were far more consistent.

  6. That's pretty cool! Would you mind sharing the source code?


What do you think about all this? Please leave me a comment! 8)