|
As you read through the various threads by FTL, you will come to the conclusion that the number of races that he BETs probably represents a very small subset of all races in which he finds contenders/qualifiers. He demands value, never bets favorites and plays multiple tracks, a quality, not quantity formula, involving patience proven over time.
At the summer DMR meet over one third of all races had more than two unknowns, either races with more than 2 first time or foreign starters, or races with two or more entrants changing surface or distance,or any horse that had not run in 90 days or whose ML was 20/1 or greater.
If you choose, you can integrate FTL's contenders with those in the top 4 and ties ML. These add backs can use best of 3 comparable or any other Sartin decision making tool, and many of them will not deserve to be put back.
One of the things FTL has said many times relative to modeling is that you need to break down races by sex, distance, surface and even class levels. I have done some of that.
I used FTL's contenders (some of which are not top 4 ML!!) and integrated (or gave a second look to) any top 4 and ties ML horse. Here are some of the best subset results:
Non maiden females (any age) on dirt were top 3 Seg 1F 78.6%, top 3 Seg 2C 85.7% (Segments screen)
Non maiden males (any age) on dirt were top 3 Seg 1F 74.2%, top 3 Seg 2C 81.8% (Segments screen); HOWEVER, they were top 2 Seg 2C 66.7%
I am sure that there are better subsets of data within these numbers, its just a matter of delving deeper to find them. The problem, as always, with such a small sample is that once you start slicing and dicing, the sample sizes get quite small
|