Interesting article on the (in)accuracy of race time predictors (Read 519 times)

Not my usual look

LedLincoln and flatfooter,

It should work now.  The culprits were tab characters from the notes which were not properly escaped (I'll have to let eric know).  So it will now strip any tab characters before parsing.

And in the future it should still plot the races that it can parse, it will just put any un-parsable entries as an warning message below the graph.

Yes it does work, and it's way cool.  Hovering over the race dots and the trendline are great features!  Thanks, Doug.

I took out my first year of running, and races that I didn't really race, and got this.   I'm a little confused, though.  Using this 1.0975 coefficient in Eric's calculator for my next marathon predicts that I should be about 900 seconds faster than the trendline.  Shouldn't it put me right on the trendline?

Well at least someone here is making relevance to the subject.

It depends on which race you put in Eric's calculator.

The typical formula is presented as: T2 = T1 * (D2/D1)^C

But a more general form would be: T = a * D^C     (where T is T2, D is D2,  and a is  T2 * (1/D1)^C for the two point case)

So my calculator attempts to find the best fitting 'a' and C (although it currently only prints the C). But the 'a' may not actually correspond with any of your actual races.

If you used the C and a point from the curve from my calculator and used that in Eric's calculator instead of a actual race, it will spit back only points on the curve.

Not my usual look

I tried it with both of my 10 mile races, both of which are quite close to the curve.  Then I tried it with a point on the curve, and yes, it's a lot closer.  It's very sensitive.  I think it's telling me I'm going to need to run my little heart out to hit my marathon goal.

MTA:  Depending on which races I choose to include, even among the ones that I deem to be pretty valid recent efforts, my coefficient varies from 1.06 to 1.11, and marathon prediction varies from 3:35 to 3:45.

Well at least someone here is making relevance to the subject.

I tried it with both of my 10 mile races, both of which are quite close to the curve.  Then I tried it with a point on the curve, and yes, it's a lot closer.  It's very sensitive.  I think it's telling me I'm going to need to run my little heart out to hit my marathon goal.

MTA:  Depending on which races I choose to include, even among the ones that I deem to be pretty valid recent efforts, my coefficient varies from 1.06 to 1.11, and marathon prediction varies from 3:35 to 3:45.

Well, I feel like if you don't run your heart out, you are doing it wrong. :-)

But yeah, it does seem to be pretty sensitive.  And in the end it does come down to the specific race environment (terrain, weather), and exactly how you are prepared (both physically and mentally).  And models are always wrong.  I feel like a Stats teacher told me that sometime.  All models are wrong, but some are useful.

Reran my data after taking out my two sucky marathons (only left in my BQ) and it basically stayed the same at 1.11.  So it must be the races at the lower end of the scale that are influencing the line.  In fact, my BQ is well below the line.  This AM I ran a 15K that was spot on with what McMilian predicted based on a 10K in Jan and a 5K in November and the line fell to 1.09.

This is really cool, thanks for building and sharing it.

I don't have many races yet -- 10 total, including two 10K and one HM. Everything hits the line:

If I remove my HM time, it goes up to 1.067.

Ran this in Eric's pace calc, and I get a marathon time of 3:05:00. Was planning to aim for 3:15:00 at my first marathon next month. Hmmm.

Not my usual look

Ran this in Eric's pace calc, and I get a marathon time of 3:05:00. Was planning to aim for 3:15:00 at my first marathon next month. Hmmm.

Hmmm indeed.  I'm obsessing over what this means for my upcoming marathon too.  In your case, since it will be your first, I would suggest that you err on the conservative side.

Well at least someone here is making relevance to the subject.