How would you go about comparing two sets of figures?

[andicap]

In your estimation, what is the best method for trying to compare the efficacy of two sets of figures? (Not just which one wins the most races, but ROI as well.)

I've seen a bunch of models

1. Best of last 3 (same distance structure/surface as today)
2. Best 2 of last 3
3. Best last race
4. Best last race on today's distance structure/surface
5. Etc.

And maybe you're better off figuring which method is the best at determining who are the top 2-3 contenders instead of just getting the winner?

One method might be better at picking winners (ROI here) the other at identifying the best 2-4 horses in a race.

I've wanted to compare some figures but haven't been able to settle on the most reliable method.

Or does it NOT matter as long as you're consistent -- that is, applying the same yardstick to both sets of figures?

[Overlay]

A metric that has worked well for me, and for which I have seen satisfactory statistical support, is the average of those races out of the horse's three most recent starts (irrespective of dates or distances) that were run on the same surface as today's race, or (if none of the horse's three latest races were run on today's surface) the average of the figures from all three of the horse's most recent starts. If the horse has only started one or two times, calculate the average based on the race or races of the one or two that was/were run on today's surface. If none of the available races were run on today's surface, use all of the available one or two races to arrive at the average.

Averages calculated in this way produced a smooth progression of impact values in ranking a field from top to bottom (1/2/3/Front Half (but not in Top 3)/Rear Half), and were applicable to multiple figure types.

(This is geared more toward winning percentage than ROI, but has been effective for me when considered as one component of an odds-generation model.)

[proximity]

In your estimation, what is the best method for trying to compare the efficacy of two sets of figures? (Not just which one wins the most races, but ROI as well.)


i would actually go in a completely different direction here: make up a batch of your own figures for a track and see which one of the two sets of figures you're comparing is closer to your homemade figures.

[fmolf]

i think you will find that different figures work at different times at different tracks.Some figs are better than others at tracks with an early speed bias,others work better when track profile is favoring pressers/closers.Figures need to be tested separately at each trqack and each surface and distance.

[JohnGalt1]

I agree with michael Pizzola who says to always use the most recent pace figure UNLESS there is/are a good reason(s) not to.

With some, you may have to go back all ten races.

While some always use the most recent line(s) I unearth some live long shots.

I also tried averaging recent races, it didn't work often enough for me.

Say the lines you average are 45 1:10 and 46.1 and 1:12. Their average time is 45.3 and 1:11.