TGJB –

Since my practice is to sit on my hands for several races during any given race card, I confess to having followed along with the study to pass the time. In retrospect, I think your decision to focus on the winners of races (and in some cases 2d-place horses) and the ranking system was a mistake.

As I indicated shortly after he posted it, I think P. Eckhart's suggestion of treating every race as a series of match races, matching each runner against each of the others, was a good one. In most matched pairs, TG and Ragozin would agree on which horse is fastest, and that match race would be disregarded. Only match races in which the services disagree on who is the fastest would be scored.

I think one of your reasons for objecting to this approach and focusing on winners was that you wanted to avoid the effect of horses who don't run their race. If you think this issue would alter the outcome of the study, a more direct way to deal with this would be to require that both horses in the match race run reasonably well (perhaps defined as finishing in the top half of the field).

The approach you settled upon does not really address the problem and introduces some anomalous results. It's simplest to talk about this in terms of concrete examples. Let's name the 8 horses in a race A through H. Say TG ranks them ABCDEFGH. Ragozin, however, ranks them ABCDFEGH, identical to TG's ranking except that E and F are inverted. In terms of the study, this produces essentially the match race discussed above, between E and F. But the match race is scored only if E or F runs 1st or 2d. First, this approach eliminates potentially useful data. Second, I would argue that if E or F runs 1st or 2d, it is probably because E or F ran a new top, and not because its past efforts were faster than the other's. Third, your approach does not ameliorate the effects of E or F throwing in a clunker.

Take another more dramatic example. TG ranks the horses ABCDEFGH, but Ragozin dramatically underrates C, ranking him 7th (ABDEFGCH). This produces a windfall for Ragozin. In terms of the study, Ragozin gets a point for the race if any of four runners (D, E, F, or G) wins, while TG gets a point only if C wins.

I'm passing along these thoughts just for what they're worth. Caveats are that I don't read the Ragozin board, so that I have no idea how this relates to what may have been discussed there, and that I'm not in a position to evaluate the rating system itself (although the idea of completely ignoring a horse's most recent race if it happens to be the worst of the last 3 troubles me). Any mechanical rating system is going to have issues. To quote H. L. Mencken: "For every complex problem there is a simple solution. And it is always wrong."

In closing, I want to compliment scavsiu8 on his fine work and congratulate you on your win at Del Mar.