combing factors

[jasperson]

Does anybody think that a program can or is developed that will predict winners that have a positive roi for all races? It is my opinion that it can't be done.
I have two programs oddline and combined factors. Both get about 31%
winners but a negative roi. If I throw out maiden races with lots of first time
starters,most turf races and 7 f races I get a small positive roi ,but nothing
great . Something like 2.03 roi I would like to know what you think about this.Remember it must use the top pick in every race and only data from the drf.

[JustRalph]

More Brylcream ?

[bobphilo]

Originally Posted by jasperson

Does anybody think that a program can or is developed that will predict winners that have a positive roi for all races? It is my opinion that it can't be done.
I have two programs oddline and combined factors. Both get about 31%
winners but a negative roi. If I throw out maiden races with lots of first time
starters,most turf races and 7 f races I get a small positive roi ,but nothing
great . Something like 2.03 roi I would like to know what you think about this.Remember it must use the top pick in every race and only data from the drf.

I thought this was going to be about combing factors. I prefer soft bristles and teeth myself.

[jasperson]

Originally Posted by bobphilo

I thought this was going to be about combing factors. I prefer soft bristles and teeth myself.

This is what I use to combine factors
a(f1)+b(f@)+c(f3)+d(f4)
That way you can add as many factors a you want and combine them anyway you want. If don't want a factor just set the coefficient to zero. In the spread sheet you change the coefficients to different values to see what works better.
I use class ,speed, early speed, jockey,trainer,form,and key trainer stat.

[Robert Fischer]

Originally Posted by jasperson

Does anybody think that a program can or is developed that will predict winners that have a positive roi for all races?

If 85% of the races that I run through the program outputs a 'pass' recommendation, does that count as 'for all races'? Does that meet your rule, or is that considered 'not for all races'?

[jasperson]

Originally Posted by Robert Fischer

If 85% of the races that I run through the program outputs a 'pass' recommendation, does that count as 'for all races'? Does that meet your rule, or is that considered 'not for all races'?

No it doesn't I have 2 programs that would make your 85% rule I know if rule out some of those almost impossible races like the derby and the large fields in the breeders cup the average will go up
.

[AltonKelsey]

With enough data mining , you could probably come up with something .


The chance of it working going forward, not so great .

[bobphilo]

Your best bet there is to do a multiple regression analysis that weighs all the factors you chose to include. Quirin does this in his book "Winning at the Races" using DRF data. Their data has improved since then so your results would be even better if you did that now.
You also have to be careful using data like trainer and jockey since you have already taken their contribution into account with speed and class or you would be counting their contribution twice, Iunless there has been a change with them. I always cringe when I hear a commentator say the horse has a poor record but he's trained by so and so or ridden by so and so. Meaningless unless these have changed for the better.

[JJMartin]

Originally Posted by jasperson

Does anybody think that a program can or is developed that will predict winners that have a positive roi for all races? It is my opinion that it can't be done.
I have two programs oddline and combined factors. Both get about 31%
winners but a negative roi. If I throw out maiden races with lots of first time
starters,most turf races and 7 f races I get a small positive roi ,but nothing
great . Something like 2.03 roi I would like to know what you think about this.Remember it must use the top pick in every race and only data from the drf.

It probably cannot be done with a strictly black box approach. It would have to account for too many possible scenarios. Sometimes a horse looks good in every category, figures etc but all of the best Jky's are on other horses, why is that? The trainer is one of the best too. To me it means the trainer is not going for a win and is just exercising the horse or some other reason. Sometimes you will see the opposite which could be a spot play. The best Jky is riding a horse of a trainer with lousy stats, why would he do that? You have to handicap the people as well as the horse. They know things the public doesn't. I think if someone were to get information from every jockey and trainer in every race along with all of the DRF information it could be done (in theory). There is no possibility of that in reality. You just have to do your best to read between the lines.

[zerosky]

There's plenty of 'Home brew' advice out there, this is a good a place as any to start ...

http://bote.wikidot.com/probability-estimation

[Robert Fischer]

Originally Posted by jasperson

No it doesn't I have 2 programs that would make your 85% rule I know if rule out some of those almost impossible races like the derby and the large fields in the breeders cup the average will go up
.

Then, relative to my perspective, it looks doomed to failure.

The game is not in such a market where every race can be played. Most of the races offered are negative-expectation races. It's a parimutuel market with a 'fixed' takeout rate. So the cause of most of these negative-expectation races tends to be a combination of insufficient field-size and/or strong favorites(specific cases of efficient market).

Although relative to my own perspective, I see a problem in your general approach, it is not to say that there are no possibilities of exceptions. If your rebate is high enough there are some possiblities. I haven't worked on this problem to have complete mastery, but obvious instant-recall brainstorm ideas for application of large rebates over all races include; Occassionally simply/obviously betting the strong favorites (that un-rebated would be negative-expections), occasionally dutching multiple win horses, and occasionally betting the pick-6 or pick-5, super-high-5 and superfecta wagers. If the rebate is strong enough, these tactics can hypothetically leverage the esoteric understanding deeply enough to come out to a positive expectation even when the market being offered is normally a sucker-bet.


Although hypothetical, it also unlikely to be the preferred approach.

[green80]

Originally Posted by jasperson

Does anybody think that a program can or is developed that will predict winners that have a positive roi for all races? It is my opinion that it can't be done.
I have two programs oddline and combined factors. Both get about 31%
winners but a negative roi. If I throw out maiden races with lots of first time
starters,most turf races and 7 f races I get a small positive roi ,but nothing
great . Something like 2.03 roi I would like to know what you think about this.Remember it must use the top pick in every race and only data from the drf.

It will never be profitable betting every race. Picking winners is only half of the game. You only bet when the price that you are getting is greater than the horse's chance to win. The same horse that is not a good bet at 2/1 is a must bet at 8/1.

[Robert Fischer]

Even if I had brylcreem, the only way I could bet a card like today's Belmont entries, would be to only bet turf, or to bet into one of their Pick-5s.

The field sizes for their dirt races would all be an auto-pass for me.

Sometimes, I'll still have a bet in a race that fails my field-size requirements. Almost always, those exceptions are heavy favorites that have a huge negative model(s) against them. The price is generally very disappointing in those half-fields. However, using the Pick-5, and the NYRA-BETS LATE PICK-5, does at times offset the paltry half-fields on dirt enough to be a profitable expectations.

[jasperson]

Originally Posted by green80

It will never be profitable betting every race. Picking winners is only half of the game. You only bet when the price that you are getting is greater than the horse's chance to win. The same horse that is not a good bet at 2/1 is a must bet at 8/1.

I am sure that winning betting a top pick every race is impossible,but I for one will continue to improve both of my programs That is what makes handicapping fun

[pandy]

Originally Posted by jasperson

I am sure that winning betting a top pick every race is impossible,but I for one will continue to improve both of my programs That is what makes handicapping fun

Top pick, short time possible, long term, near impossible. But something like, betting one of the top-ranked contenders that is off at or above a certain odds range, possible.