question on odds

[Jack950]

Originally Posted by the little guy

The answer is that without a sophisticated computer model, you can't. Given takeout, there won't actually be a lot of horses that represent "value" and if they do, it will be marginal....i.e. getting 5:2 on a horse that actually should be 2:1 is really very good value. What you might be able to do over time is develop a reasonable feel. Also, understand that if you are playing exotics, vertical or horizontal, the diffusion of takeout through added positions can create better "value" than straight win odds.


The best chance you have at getting "value" is to effectively eliminate favorites. If you can accurately remove even a 4:1 shot, you have effectively eliminated the takeout. If you eliminate a 2:1 from the first two slots, you could dutch the field and make money. This is the best way to give yourself a chance to win.

Spot on. Good post

[formula_2002]

Originally Posted by Parson

Trying to add another weapon in my arsenal. I have been a bit more selective on what races I wager the past few months. Fewer wagers have resulted in more profit at the end of the wagering day.
Now I want to really concentrate on my pick and only wagering if there is value instead of just wagering to pick a winner. My question is this: For those of you that only bet if you think the horse has value, meaning he is 4/1 and should be 2/1, is how do you determine if the horse should be 2/1?

I need a bit of help in that area and any suggestions would be much appreciated.

you are showing a PROFIT.?
you dont need our help, we need yours

[biggestal99]

Originally Posted by the little guy

The answer is that without a sophisticated computer model, you can't. Given takeout, there won't actually be a lot of horses that represent "value" and if they do, it will be marginal....i.e. getting 5:2 on a horse that actually should be 2:1 is really very good value. What you might be able to do over time is develop a reasonable feel. Also, understand that if you are playing exotics, vertical or horizontal, the diffusion of takeout through added positions can create better "value" than straight win odds.

.

Computer Models? dont need em. My takeout is 12%. I lay horses on the exchange to make money. so if I make a mistake valuing horses I pay though the nose. I have been handicapping horses a long time and know my horses.
Once the round reaches 97 or 98 or so thats when I find my underlays and bet them to lose. (or offer lower odds). But you are correct I concentrate on laying between 2.00 (evens) to 6.00 (5-1). I am right about 93% of the time. The other 7% yikes, I hurts when you lay a 4-1 that you think is going to lose; wins for fun.

Allan

[Parson]

Thanks for all the advice everyone, I will take it all into consideration and try and implement some of your suggestions, even those that remain unspoken and subjective

[jahura2]

Originally Posted by the little guy

The answer is that without a sophisticated computer model, you can't. Given takeout, there won't actually be a lot of horses that represent "value" and if they do, it will be marginal....i.e. getting 5:2 on a horse that actually should be 2:1 is really very good value. What you might be able to do over time is develop a reasonable feel. Also, understand that if you are playing exotics, vertical or horizontal, the diffusion of takeout through added positions can create better "value" than straight win odds.


The best chance you have at getting "value" is to effectively eliminate favorites. If you can accurately remove even a 4:1 shot, you have effectively eliminated the takeout. If you eliminate a 2:1 from the first two slots, you could dutch the field and make money. This is the best way to give yourself a chance to win.

Very helpful post. Good to have you back TLG!

[ultracapper]

Eliminating the fav is the key. I don't know how many times I've found a horse in the 8 to 15 range that I knew would run, and then look at the fav and know damn well he would run also. Of course, the exacta is the answer for that, but I'm usually too slow to come up with that answer. The seldom times I play the exacta, I hate using the favs there also.

I know lots of guys make their own line, but I don't think it's as effective a way to find "value", or "price", as just simply eliminating the fav. So you walk away from a 7/5 because you think he should be 2/1. What about the 8/1 that you think should be 10? I'll bet that 8/1 all day if I really believe he's going to run and I'm pretty sure the shorter priced horses are susceptible.

[zerosky]

A 4/1 horse on the tote board can be converted into the public's subjective probability of winning by removing the effects of the takeout.

P(pub)= (1-t)/(odds+1)

so if takeout is 16% then P(pub)= 0.84/5 =16.8%

If you believe its real chance of winning is 2/1 the the calculation of the probability does not involve take out (it will win one in three times) therefore P(your)= 33%

In this particular instance the whole line is probably out of whack due to a false favourite or something, or you have privileged information hidden from the public.

In the long term you would have to compare your odds with the public odds over the same selections to see if you really do have an edge. A good method would be to compute the Brier Score or Brier Skill Score.

http://www.statisticshowto.com/brier-score/

[Fox]

Originally Posted by formula_2002

you are showing a PROFIT.?
you dont need our help, we need yours

Good point

[JeremyJet]

Originally Posted by biggestal99

Computer Models? dont need em. My takeout is 12%. I lay horses on the exchange to make money. so if I make a mistake valuing horses I pay though the nose. I have been handicapping horses a long time and know my horses.
Once the round reaches 97 or 98 or so thats when I find my underlays and bet them to lose. (or offer lower odds). But you are correct I concentrate on laying between 2.00 (evens) to 6.00 (5-1). I am right about 93% of the time. The other 7% yikes, I hurts when you lay a 4-1 that you think is going to lose; wins for fun.

Allan

So you're betting with Betfair? I was wondering how it's going in New Jersey. I never see anyone talking about this subject. What kind of handle is Betfair generating in NJ?

My hope was that Betfair would expand to other states .... eventually. Nothing.

[biggestal99]

Originally Posted by JeremyJet

So you're betting with Betfair? I was wondering how it's going in New Jersey. I never see anyone talking about this subject. What kind of handle is Betfair generating in NJ?

My hope was that Betfair would expand to other states .... eventually. Nothing.

Handle was about 300,000 per week last June. (2017) (30% from old customers, 70% from new ones)

I am sure it continues to grow.

Allan

[JeremyJet]

Originally Posted by zerosky

A 4/1 horse on the tote board can be converted into the public's subjective probability of winning by removing the effects of the takeout.

P(pub)= (1-t)/(odds+1)

so if takeout is 16% then P(pub)= 0.84/5 =16.8%

If you believe its real chance of winning is 2/1 the the calculation of the probability does not involve take out (it will win one in three times) therefore P(your)= 33%

In this particular instance the whole line is probably out of whack due to a false favourite or something, or you have privileged information hidden from the public.

In the long term you would have to compare your odds with the public odds over the same selections to see if you really do have an edge. A good method would be to compute the Brier Score or Brier Skill Score.

http://www.statisticshowto.com/brier-score/

What would be considered good results? What does a good Brier score look like?

[JerryBoyle]

Originally Posted by JeremyJet

What would be considered good results? What does a good Brier score look like?

For something like horse racing, where your selections are multi class instead of binary, the Brier score is going to range from 0 - 2 with 0 being best and 2 being worst. It's best not to think of "good" vs "bad" in absolute sense, but rather, is this model's score better than the previous model? That said, in my personal experience, I had a model which was about break even over ~5k races with a Brier score of .773. One of the downfalls of using the Brier score is that it's not scoring your predictions relative to other bettors, which is what we would want if we plan to bet.

Personally, for scoring my predictions without actually running a betting simulation, I prefer something called Bayesian Information Reward. It has a natural fit with gambling, because it compares your estimates to the public's, and rewards you in cases where you had a prediction greater than the public when the event happens (or when you had a prediction less than the public when the event did not happen), and penalizes you in the opposite cases. The actual formula and a discussion about its properties can be found here: http://users.monash.edu/~korb/shadowfax/pubs/ai02.pdf. The formula is Definition 3 on page 4.

[JeremyJet]

Originally Posted by JerryBoyle

For something like horse racing, where your selections are multi class instead of binary, the Brier score is going to range from 0 - 2 with 0 being best and 2 being worst. It's best not to think of "good" vs "bad" in absolute sense, but rather, is this model's score better than the previous model? That said, in my personal experience, I had a model which was about break even over ~5k races with a Brier score of .773. One of the downfalls of using the Brier score is that it's not scoring your predictions relative to other bettors, which is what we would want if we plan to bet.

Personally, for scoring my predictions without actually running a betting simulation, I prefer something called Bayesian Information Reward. It has a natural fit with gambling, because it compares your estimates to the public's, and rewards you in cases where you had a prediction greater than the public when the event happens (or when you had a prediction less than the public when the event did not happen), and penalizes you in the opposite cases. The actual formula and a discussion about its properties can be found here: http://users.monash.edu/~korb/shadowfax/pubs/ai02.pdf. The formula is Definition 3 on page 4.

Thanks for the reply.

I must be doing something different than you. I followed the link provided by the other poster and followed the instructions. Here are the Brier Scores I came up with:

.099
.116
.078
.065

I used 1 for the winner and 0 for the losers.

[JerryBoyle]

Originally Posted by JeremyJet

Thanks for the reply.

I must be doing something different than you. I followed the link provided by the other poster and followed the instructions. Here are the Brier Scores I came up with:

.099
.116
.078
.065

I used 1 for the winner and 0 for the losers.

If those are the scores for 4 races, that is possible. The score I posted is the mean of scores over 6024 races. Within that sample of 6k races, there are some (105) with scores within the range you've posted. However, if your scores are consistently as you've posted, whatever model you're using is significantly better than mine .

EDIT: over those 6k races, the public's final odds line using the win odds converted to probabilities had a brier score of .7698

[JerryBoyle]

Originally Posted by JeremyJet

Thanks for the reply.

I must be doing something different than you. I followed the link provided by the other poster and followed the instructions. Here are the Brier Scores I came up with:

.099
.116
.078
.065

I used 1 for the winner and 0 for the losers.

Also, note that you need to use the scoring rule for non binary events. E.g., don't divide by the number of runners within a race. For example, if your odds like looks like:

PGM Prob Winner
1 .12 1
2 .23 0
3 .05 0
4 .4 0
5 .2 0

Your brier score for that race is 1.03.