Second Ranked Horse

[classhandicapper]

Has anyone seen this phenomenon?

1. A method I developed ranks horses in order of preference. I tested the method and the top choice outperformed the track take significantly.

2. Within those top choices, I found a couple of sub groups that did poorly and significantly under performed the take take.

a. The most obvious next move would be to eliminate the bad sub groups and focus on the remaining top choices because the results are better.

b. The next move might be to take a look at the races where there was a bad top choice and focus on the second choice assuming that if the top choice under performs the take badly, the second choice should do much better.

That's what I just tested. To my surprise, the 2nd choices did not do well in scenario b. They also slightly under performed the track take. The sample is not very large (just about 100 races), but it was surprising to me.

I'm throwing out horses that are clearly overbet, but I'm not gaining any value with the next ranked horse. Perhaps the sample is just too small, but I'm wondering if there's some oddball thing going on where the method is powerful at finding most likely winners but not so good at separating the rest of the contenders.

[Robert Fischer]

you have competing unknowns here


you don't know if your subgroups that appear to be underperforming are simply outside your method's circle of competency, or whether they are randomness due to the law of small numbers.
Then you are trying to take those subgroups and try them as a negative or 'elimination' factor.


It sounds like a good mental exercise.

[CincyHorseplayer]

I will only bet from my top 2 choices and I rarely have 1 or 3 choices to bet. Tracking their win % for years they win roughly the same amount of time over large samples. For me this year it's about 28% each. I almost make no distinction between the two when betting. 90% of the time I will take the higher odds on the two. The impact of this is, from 2001 to 2014 my average mutual was 9.40. Since 2015 it has gone up to over 12.00. Sits at 12.62 this year.

The other sub group, marginal contenders will never be key bets. 3rd, 4th, 5th choices collectively are only winning 32% of the races for me this year. It is a futile and fruitless ground for mining anything. They are simply throw ins in other bets.

[classhandicapper]

you have competing unknowns here


you don't know if your subgroups that appear to be underperforming are simply outside your method's circle of competency, or whether they are randomness due to the law of small numbers.
Then you are trying to take those subgroups and try them as a negative or 'elimination' factor.


It sounds like a good mental exercise.

I'm pretty sure the negative sub groups are legitimate because they are long established negatives (certain length layoffs, certain surface switches etc..).

The method is simply ranking horses without much regard to distance, surface or much else. It would comparable to looking at top Beyer without doing any other handicapping. The negatives are adding a small layer of handicapping.

[Dave Schwartz]

Has anyone seen this phenomenon?

That's what I just tested. To my surprise, the 2nd choices did not do well in scenario b. They also slightly under performed the track take. The sample is not very large (just about 100 races), but it was surprising to me.


No disrespect meant here, but a 100 race sample is literally nothing.

When things seem out of whack after 100 races that is usually because they are. At 500 races the whole thing will probably smooth out.

[classhandicapper]

No disrespect meant here, but a 100 race sample is literally nothing.

When things seem out of whack after 100 races that is usually because they are. At 500 races the whole thing will probably smooth out.

I'm hoping that's it. But since these are higher ranked horses they also tend to be shorter priced horses. So the sample is not as small as if it was a sample of random horses given the result is far away from my expectations.