Saturday, May 5, 2007

Lies, Damn Lies and Statistics

I've been doing a little investigation on race "predictors". There are a bunch of calculators that will give you a predicted time for a race based on your time in a recent race of a different length. I'm trying to determine just how fast a goal I want to set for my Marathon in December, based on my best half marathon time.

The various calculators all give similar sorts of results, and I've been using the VDOT equivalence tables from Jack Daniels' Running Formula. If you don't know about VDOTs or Jack Daniels (the running dude, not the drink) you can go buy the book but he describes VDOT as "A measure of your current running ability". There's a whole bunch of science and measurements of various things involved but basically he created a set of tables that show what race times you would get for a given VDOT. So someone with a VDOT of 51 has a predicted half time of 1:30:02 and a full marathon time of 3:07:39. Google on VDOT Tables and you'll get lots of stuff. Here's a link to the tables....


So assuming (and admittedly this is a pretty big assumption...) I meet my goal of getting to 1:30 for a half marathon, I SHOULD be able to hammer out a 3:07 full marathon based on VDOT equivalent performance. Given that my best marathon time is 4:10 (off of 25MPW, it got very ugly about mile 21), this is a massive improvement and I'm a little suspicious of it all since I'd be surprised as hell if I ran a 3:07 in December.

So I decided to do a completely bogus, scientifcally questionable study!

I wanted to see how people really performed compared to VDOT predictions and also see if there was a significant difference based on how many miles people ran during training.

I posted this on 3 running message boards (Letsrun, Coolrunning, and Runners' World)

I'm doing some analysis on the relationship between half marathon and full marathon times and the effect that weekly mileage has on that ratio.

Need some data from runners (that would would be where you come in.....)

Eligability
- Half marathon run within 6 months of full
- Both run as races (no easy training runs)
- "Similar conditions" for both races
Data Required
- Half time
- Full time
- Average weekly mileage training for full

I got 57 sets of results from the 3 boards. Here's the result.
First I converted the times to VDOTs and calculated the differences people saw in their half VDOT to their full VDOT. If everything was perfect, everyone should have performed at exactly the same VDOT level.
Here's the chart showing number of runners against VDOT difference. Out of 57, only 16 met or beat their predicted times, 41 did worse than "predicted".

On average people were 2.1 "VDOTs" below predicted time for the full marathon, which is about 8 minutes in the 3:10 range, so that means instead of doing a 3:07, I'd do a 3:15.

Next I looked to see if there was any correlation between weekly training mileage and VDOT difference. Here's the chart...
Data is scattered, but a best fit line shows that lower mileage created a higher VDOT difference. So based on a 1:30 half (VDOT=51) if I train 30MPW I could expect to perform at a VDOT of 48 for my full marathon (3:17) whereas if I train 120MPW I could expect a VDOT of 50 (3:10).
I plan to train about 65MPW in the fall so will be setting my training paces for a 3:14 marathon (VDOT 49) assuming I manage to get the 1:30 half done in July. 3:14 is also a good target time since it is comfortably under the Boston Qualifying time for my age (3:20) and also considered "Good For Age" (3:15) in the London Marathon, which I plan to run in April 2008.
Of course, this is all suitably dubious science, based on people who aren't me. I'll end up adjusting my actual race goal based on how the training goes leading up to the race. Ultimately I will still be happy in December if I can
a) finish without crashing at mile 21
b) finish comfortably under 3:45
but unfortunately the scientist in me now realises that 3:14 is not an unreasonable goal.

Updated to add a couple charts that I never included.



1 comment:

Anonymous said...

Good info, thanks very much for sharing it. As someone who is using Daniels marathon plan A to prep for his first marathon in a few weeks, having this data is super interesting and helpful.