Just produced this report and thought it might be interesting to some here. I've taken all races from 2016-01-01 to 2019-02-18 and broken it up into ~2k chunks of races. I then generated a model for each chunk using just the final public estimate as an input to the model. Attached is the report including the start/stop date for each chunk, the number of races used, and the associated r2. Also included rolling mean of 3 and 6 periods.
Note, I've removed races with entries and races with dead heats for first place.
Edit: This is for all North American tracks (all distances, etc). Only restrictions are those above. For those unfamiliar with this type of model or modeling in general, r2 can sort of be interpreted as how well the independent variable (public's estimate) fit the dependent variable (coming in first). In this case, the increase in r2 indicates that the public's estimate contains more information that's useful for predicting the winner of race (relative to models with smaller r2).
yes public odds is a very important part of the model, that mr. benter uses in his model. cx wong one of mr benter's modelers way back in 2011, mentions in writing that the importance of incorporating public odds in their model was one of the biggest discoveries they had in moving forward with improvement in the model development.
lessman sung and johnson, 2011; discovered the same phenomenon in their studies.
Using public odds as part of line making always struck me as an admission that your own model is not considering some relevant factors and/or not weighting some of them properly.
I'll be the first to admit that sometimes I don't know all the relevant information, misinterpret information, don't weigh things correctly etc.. So adding in public odds will make my overall odds lines more accurate.
But the goal in gambling is to find situations where the public is wrong.
For example, if I started handicapping races at Hastings (a place I almost never play) there's no doubt that adding in public odds would make my overall odds lines way more accurate.
But would that help me win?
If I happen to have a specific trainer pattern there that the public is not aware of yet, the last thing I want to do is add in the public odds. I already know that relevant information is missing. That's why I am considering a wager.
The goal isn't making more accurate lines.
The goal is finding races where the public is wrong.
__________________
"Unlearning is the highest form of learning"
Using public odds as part of line making always struck me as an admission that your own model is not considering some relevant factors and/or not weighting some of them properly.
I'll be the first to admit that sometimes I don't know all the relevant information, misinterpret information, don't weigh things correctly etc.. So adding in public odds will make my overall odds lines more accurate.
But the goal in gambling is to find situations where the public is wrong.
For example, if I started handicapping races at Hastings (a place I almost never play) there's no doubt that adding in public odds would make my overall odds lines way more accurate.
But would that help me win?
If I happen to have a specific trainer pattern there that the public is not aware of yet, the last thing I want to do is add in the public odds. I already know that relevant information is missing. That's why I am considering a wager.
The goal isn't making more accurate lines.
The goal is finding races where the public is wrong.
I would think adding in the public odds as a factor will help eliminate blatant errors in a personal odds line.
I would think adding in the public odds as a factor will help eliminate blatant errors in a personal odds line.
I can agree with that, but I'm not sure how that helps me win in most cases.
I could see using public odds for first time starters, long layoffs, and a few other situations where some clockers or other insiders might have some information about a horse I don't have. That way if I am interested in another horse in the race, I might get a better line on one of the question marks before I decide what price to take.
But usually, when I'm going to bet, I go into the race with some specific insight into a horse or that race that I think the public is likely to miss. Then I play if I get my price.
__________________
"Unlearning is the highest form of learning"
I can agree with that, but I'm not sure how that helps me win in most cases.
I could see using public odds for first time starters, long layoffs, and a few other situations where some clockers or other insiders might have some information about a horse I don't have. That way if I am interested in another horse in the race, I might get a better line on one of the question marks before I decide what price to take.
But usually, when I'm going to bet, I go into the race with some specific insight into a horse or that race that I think the public is likely to miss. Then I play if I get my price.
It could save some bets on some horses, that is all.
With what you are doing, and I do the same many times, I agree. But you really don't need to construct a line for those kinds of bets in my opinion.
We do tend to undervalue stuff that doesn't fit our play.
On the other hand, if we have an understanding of why the odds are going to behave in some irrational way, -- and then the odds do in fact behave in that irrational way, - - we don't have to blend the public odds.
Also depends on the approach. Some players are doing Ziemba/Benter style math, and some players are doing more hands-on handicapping. Depending on your system, 'public odds' could well be a significant factor.
Blending also lends itself as a sloppy/indirect way to incorporate a margin of safety.
Wasn't really posting it as a comment on whether one should/shouldn't incorporate the public estimate into the model. More so posting as a proxy for measuring how the public is improving.
I know the public is right more than any other 'system' but not more profitable. Honestly not sure how it ranks in profitability but I understand that public opinion by itself is NOT profitable.
I like to gage how well a horse 'fits' within a group by looking at the odds it took in similar level past races. It the horse consistently gets 30, 70-1 odds at the level then I would toss it.
But if I 'liked' a horse and it's sitting at 40 to 1 in today's race then I can't help but get a bit more nervous but I'd bet it.
