The realities of Betting to Win

[Capper Al]

I'm spinning this conversation off the thread:

Betting To Win- Impossible Mathematically To Profit Long Term ? (http://showthread.php?t=77147)

The sentiments in the above thread should ring true for most of us. If I have a 25% win ratio why can't I just play 3/1 horses or better and make money? Many books point out the simplicity of this logic. The mistake is that they are using simple probability to establish these truths. What I call one dimensional reasoning.

Let's look at the premise for one dimensional reasoning used by almost all authors: With a fair dice, there is a 1 out 6 probability that any number will come up. If the payoff is greater than the chance of 1 to 6, one must make money. The numbers will just work in your favor over the long run. If one could get 8/1 for every hit then on average for every 6 rolls of the dice they should make $2.00-- $8 in verse $6 out. True.

Why doesn't this apply to para-mutual wagering? Because our choices are second picks, not first picks. We get the leftovers when it comes to the betting window. How can that be? This is because we are using the same information which is what is at the heart of the matter. When I place my 4/1 bet, I am saying based on my interpretation of the same information, I believe this horse should win. It isn't the horse's true changes. It is our interpretation of the information. This moves the conflict from the horse's real changes to our abilities against the crowd. This is the second dimension.

If one is a 25% handicapper and notices the odds that these horses went off at, most of them would be under 3/1. If there really is fair and independent reasoning going on with my choices from the Public and if the capper is right 1/4 and the favorite wins 1/3 then only 1/12 (1/3 * 1/4) will they have the same bet that wins. It isn't so. Usually 2/3 of the 1/4 capper's horses are favorites or second favorites.

How should the capper figure out their wagers? This is close enough for handicapping. Since your bet must be against the crowd, the Public's chance of being wrong is 2 out of 3 times. If you are a 25% handicapper then you are hitting 1 in 4. The computation would be 2/3 times 1/4 or 1/6 which makes your minimum odds 5/1.

[Overlay]

Yes, the player gets the "leftovers", in that the public has already reflected its assessment of available information in the odds that it establishes. But those odds are still the product of opinion, rather than of immutable mathematical truth (as in games of pure chance). And opinions (even collective ones) can be wrong at all odds levels -- including favorites. Detecting the opportunities/overlays created by that fact is the player's challenge. Excluding any horse or category of horses (such as favorites) from that detection process lessens the player's chance of long-term success.

As I believe you were expressing, many people approach the game as if the only object were to pick "the winner", when there are actually two questions to be answered for each horse in a race: 1) What is this horse's chance of winning?; and 2) Will its odds/payoff compensate me for the risk (i.e., the probability that the horse will not win) that I would be taking if I bet it?

[Capper Al]

Yes, the player gets the "leftovers", in that the public has already reflected its assessment of available information in the odds that it establishes. But those odds are still the product of opinion, rather than of immutable mathematical truth (as in games of pure chance). And opinions (even collective ones) can be wrong at all odds levels -- including favorites. Detecting the opportunities/overlays created by that fact is the player's challenge. Excluding any horse or category of horses (such as favorites) from that detection process lessens the player's chance of long-term success.

As I believe you were expressing, many people approach the game as if the only object were to pick "the winner", when there are actually two questions to be answered for each horse in a race: 1) What is this horse's chance of winning?; and 2) Will its odds/payoff compensate me for the risk (i.e., the probability that the horse will not win) that I would be taking if I bet it?

I set my minimum odds at 5/1 for ease and so I can have a little action. It is actually a more complicated formula. If your horse is the third favorite then you have to take into account the odds of the second favorite failing also. This works out to 7/1. For the fourth favorite now we need to include the third favorite failing also. This goes on and on. Also one needs to multiple the field size/ 7 (average field size). You'll never make a bet. What we don't say to ourselves is that everyone picks them around a 25% rate and with a takeout whats our advantage over the long run?

[Dave Schwartz]

The sentiments in the above thread should ring true for most of us. If I have a 25% win ratio why can't I just play 3/1 horses or better and make money? Many books point out the simplicity of this logic. The mistake is that they are using simple probability to establish these truths. What I call one dimensional reasoning.

This is such a simple question to answer. Your logic is all wrong.

As an example, suppose you took favorites. As you said they win at a hit rate of 33%. (Actually, they are, on average, a little higher.)

So, what if you bet any favorite to be over 2/1?

The answer becomes obvious: The average win rate on all favorites may be 33% but that does not tell us the win rate on this favorite.


Dave

[mrhorseplayer]

exotics is where its at. it is hard to make it playing win bets

[Overlay]

This is such a simple question to answer. Your logic is all wrong.

As an example, suppose you took favorites. As you said they win at a hit rate of 33%. (Actually, they are, on average, a little higher.)

So, what if you bet any favorite to be over 2/1?

The answer becomes obvious: The average win rate on all favorites may be 33% but that does not tell us the win rate on this favorite.
Dave,

I don't think that CapperAl was presenting that as his own logic or viewpoint, or claiming the statement to be true. He was talking about the thinking of a confused horseplayer with an overall 25% hit rate who would not be able to understand why he couldn't just bet his selections that go off at odds higher than 3-1, and be assured of making a profit. (That thought process is incorrect, for the reasons that you pointed out.)

[Capper Al]

This is such a simple question to answer. Your logic is all wrong.

As an example, suppose you took favorites. As you said they win at a hit rate of 33%. (Actually, they are, on average, a little higher.)

So, what if you bet any favorite to be over 2/1?

The answer becomes obvious: The average win rate on all favorites may be 33% but that does not tell us the win rate on this favorite.


Dave

I might have mixed up the 25% hit ratio of the player in the thread with the favorites. You're right that you would need 2/1 odds for 33% favorites to break even. My point is still valid. The higher odds favorite is less likely to win. We are not betting on horses but betting on or against opinions about horses. In the story about the die, we know the probability is 1 out of 6. We never know that about a horse. We do know that favorites win so much, and second favorites win so much,and we win so much. So if we are a 25% winner then we still need our probability of 2/3 (favorite loses) times 1/4 (that we are right) or 1/6 for odds of 5/1. And this is if our horse is the second favorite. If it is the third favorite then we need to include the fact that the second favorite fails also. It goes on and on the longer we go.

[Capper Al]

Dave,

I don't think that CapperAl was presenting that as his own logic or viewpoint, or claiming the statement to be true. He was talking about the thinking of a confused horseplayer with an overall 25% hit rate who would not be able to understand why he couldn't just bet his selections that go off at odds higher than 3-1, and be assured of making a profit. (That thought process is incorrect, for the reasons that you pointed out.)

You're right. I was trying to explain why a 25% capper can't make money betting on his picks that go off at 3/1 or more. I stand by my thought process, until someone can really convince me that they know the true value of a horse in a race with the same certainty as we know the value of any number coming up on a die. They can't.

[raybo]

This is such a simple question to answer. Your logic is all wrong.

As an example, suppose you took favorites. As you said they win at a hit rate of 33%. (Actually, they are, on average, a little higher.)

So, what if you bet any favorite to be over 2/1?

The answer becomes obvious: The average win rate on all favorites may be 33% but that does not tell us the win rate on this favorite.


Dave

This is correct. When you take 2/1 or higher favorites out of the classic favorites statistic of 33% win rate, you have destroyed that stat. You have just lowered the stat on the favorites you will bet and raised the stat on the favorites you won't play, the ones less than 2/1.

Dave is right, in order to hit 33% on favorites you must bet all favorites, not just the ones 2/1 or higher. Once you separate an odds category from the total group, you are at the mercy of your own handicapping abilities, not the public's.

[Dave Schwartz]

Oh, sing it, Ray and 'Lay. 'Splain it to him.

See, Al, it just isn't that easy. I used the favorite example because it is so obvious. The example would work equally well with (say) #1 ranked horse for Beyer in the last race.

I believe such a horse wins at around a 25% clip. So, why can't you automatically make money at above 3/1? The answer is because when you break all those #1 ranked horses by odds you find that as the odds go up, the hit rate goes down.

In other words, the "average" odds on such a horse might be 5/2 (hit rate 25%), but as the odds rise to 4/1 the hit rate probably drops to like 17% or so, maybe less.


Dave

[raybo]

I've made "pat" statements about hit rate, payouts, and profitability, but, I have always made sure that I was talking about a player's average hit rate vs his average payout.

When your average hit rate, compared to your average payout, calculates to a profit, then you have very little to worry about, as long as you remain consistent in your handicapping, selections, and wagering methods, and the dynamics of the game haven't changed (which means you must keep records, religiously!).

[Overlay]

I stand by my thought process, until someone can really convince me that they know the true value of a horse in a race with the same certainty as we know the value of any number coming up on a die. They can't.

I would agree, if that level of certainty were required. But it isn't. You "only" need to be better at estimating the true value than the public collectively. And, despite the public's accuracy in the aggregate, I don't think that they'll ever attain that degree of precision (i.e., comparable to the probability of a given number coming up on a roll of the dice) for every individual wager type on every individual horse in every individual race. There will always be opportunities where a positive expectation is available to the player, but value has to be taken into consideration in order to exploit them.

[Capper Al]

Oh, sing it, Ray and 'Lay. 'Splain it to him.

See, Al, it just isn't that easy. I used the favorite example because it is so obvious. The example would work equally well with (say) #1 ranked horse for Beyer in the last race.

I believe such a horse wins at around a 25% clip. So, why can't you automatically make money at above 3/1? The answer is because when you break all those #1 ranked horses by odds you find that as the odds go up, the hit rate goes down.

In other words, the "average" odds on such a horse might be 5/2 (hit rate 25%), but as the odds rise to 4/1 the hit rate probably drops to like 17% or so, maybe less.


Dave

See Dave, read post #1 the second paragraph from the bottom. I believe we may be saying the same thing.

[Dave Schwartz]

I am confused, then. Why did you ask the question if you knew the answer?

[TrifectaMike]

The probability of being a long term winner is nearly zero, if one relies too heavily on averages or groupings.

The tote board is an excellent average estimator. A potential winner has to develop a methodology that makes him/her ( for lack of a better term) an outlier.

There is a very good reason why odds follow a Pareto (Power Law) Distribution.

Mike

[thaskalos]

See Dave, read post #1 the second paragraph from the bottom. I believe we may be saying the same thing. I think that what you are saying is something we all agree with...and which could be made a lot simpler to read.

"Our shorter-priced selections win more often than our longer-priced ones."

[raybo]

I think that what you are saying is something we all agree with...and which could be made a lot simpler to read.

"Our shorter-priced selections win more often than our longer-priced ones."

Not necessarily. If you are betting value, and not your top probability horse, then yes, your longer priced horses will hit less often, but, if you're betting your horses on the basis of their probability of winning (according to your own handicapping skill) then the post time odds make no difference in your hit rate.

[Cratos]

The probability of being a long term winner is nearly zero, if one relies too heavily on averages or groupings.

The tote board is an excellent average estimator. A potential winner has to develop a methodology that makes him/her ( for lack of a better term) an outlier.

There is a very good reason why odds follow a Pareto (Power Law) Distribution.

Mike

If I understand you, you are stating as an outlier in the pari-mutuel wagering scheme you set your own wagering odds.

Simply stated is that Horse “A” might be the horse to win by your handicapping, but it might not be the horse to bet based on your betting threshold; you pass this race

Therefore you only bet when your handicapping and betting threshold is in sync with each other.

However in betting we get into the “Arrow Paradox” which generally states that when it comes to risk (wagering) we are either risk-preferred (preferences for longshots), risk-against(preference for favorites), or risk-neutral(indifferent about favorites and longshots).

Thus it is incumbent for any bettor to do a self-evaluation and understand where they are on the risk continuum.

[Dave Schwartz]

Not necessarily. If you are betting value, and not your top probability horse, then yes, your longer priced horses will hit less often, but, if you're betting your horses on the basis of their probability of winning (according to your own handicapping skill) then the post time odds make no difference in your hit rate.

So, are you saying that when you make a horse 30% and he goes off at 10/1 he wins like a 2/1 horse?

If so, you are a wealthy man, my friend.

My experience is that my perception of longshot plays' hit rate is always over-rated.

In other words, I make two different horses 30%. One goes off at 2/1 and the other at 30/1. Only in some kind of dream world do I think those two horses have the same chance of winning when the gate opens. Granted, the 30/1 horse is probably a far better bet, but his hit rate will not be 30%.

[thaskalos]

Not necessarily. If you are betting value, and not your top probability horse, then yes, your longer priced horses will hit less often, but, if you're betting your horses on the basis of their probability of winning (according to your own handicapping skill) then the post time odds make no difference in your hit rate.
Raybo, I make an odds line on every horse...and I find that my 2/1 shots win more often than my 4-1 shots...who in turn win more often than my 8-1 shots.

If YOU find that YOUR 8-1 shots win with the same frequency as your 2-1 shots...then more power to you!

It might mean that your odds line is flawed though...:)

[TrifectaMike]

If I understand you, you are stating as an outlier in the pari-mutuel wagering scheme you set your own wagering odds.

Simply stated is that Horse “A” might be the horse to win by your handicapping, but it might not be the horse to bet based on your betting threshold; you pass this race

Therefore you only bet when your handicapping and betting threshold is in sync with each other.

However in betting we get into the “Arrow Paradox” which generally states that when it comes to risk (wagering) we are either risk-preferred (preferences for longshots), risk-against(preference for favorites), or risk-neutral(indifferent about favorites and longshots).

Thus it is incumbent for any bettor to do a self-evaluation and understand where they are on the risk continuum.

You are correct. If I don't have at least one overlay in my top four contenders, I pass.

And by the way, I am definitely risk-neutral.

Mike

[TrifectaMike]

So, are you saying that when you make a horse 30% and he goes off at 10/1 he wins like a 2/1 horse?

If so, you are a wealthy man, my friend.

My experience is that my perception of longshot plays' hit rate is always over-rated.

In other words, I make two different horses 30%. One goes off at 2/1 and the other at 30/1. Only in some kind of dream world do I think those two horses have the same chance of winning when the gate opens. Granted, the 30/1 horse is probably a far better bet, but his hit rate will not be 30%.

Dave, your statement is basically true, but not entirely. I have a metric that I use to determine a longshot play, and my hit rate is way, way higher than their fair odds.

The answer one needs is not tote board odds groupings, but your own subjective odds, and hit rate.

Mike

[Robert Fischer]

using the same information which is what is at the heart of the matter. When I place my 4/1 bet, I am saying based on my interpretation of the same information, I believe this horse should win. It isn't the horse's true changes. It is our interpretation of the information. This moves the conflict from the horse's real changes to our abilities against the crowd. This is the second dimension.

good stuff :ThmbUp:


using the same info ... using the same strategy ... using the same activities
don't do it! strive to be "competitively unique"

there is also a little bit of room for using the exact same information, strategy, and activities, and doing them better - but this is generally rare. Try to do this with frequency and you will really feel the effects of takeout.


are there advantages at different ranges of odds??

sure


a gambler may appreciate high strike rates even at minimal odds
higher odds ranges are less susceptible to late odds-drops
lower odds ranges are less susceptible to losing streaks
playing all or most odds ranges offers more churn
playing a select odds range allows for specialization activities
blah

[Dave Schwartz]

Higher that normal, of course. But not without some loss of projected hit rate.

[raybo]

So, are you saying that when you make a horse 30% and he goes off at 10/1 he wins like a 2/1 horse?

If so, you are a wealthy man, my friend.

My experience is that my perception of longshot plays' hit rate is always over-rated.

In other words, I make two different horses 30%. One goes off at 2/1 and the other at 30/1. Only in some kind of dream world do I think those two horses have the same chance of winning when the gate opens. Granted, the 30/1 horse is probably a far better bet, but his hit rate will not be 30%.

Come on Dave, was my post that screwed up? I usually explain myself pretty well.

If I wager and keep records, I know that no matter what the post time odds were on my top pick, I still have an average ROI. When I handicap I have no idea what the odds are, I am handicapping for the winner, period.

Now if I have a horse graded higher than another one, but, his odds are going to be too low to bet, then I pass. But, if instead, I abandon my own wagering rules and bet on the lower graded horse, because of it's higher odds, I have a lower probability of winning the race. I know this because I've kept records, remember.

We are not talking about ROI, but hit rate.

Hit rate depends on your handicapping ability and whether you bet value or winning probability, in your own mind. If you are a very good handicapper and you bet winning probability, your hit rate will be better than if you bet your 2nd, 3rd, 4th, etc., ranked horse, because you are betting value. I'm not talking about 2 horses graded even or really close, I'm talking about 1 horse being graded high, no matter the odds, and 1 horse graded lower at good odds.

They are 2 different games, 1 for hit rate and the other for value, for the sake of value.

Originally Posted by thaskalos
I think that what you are saying is something we all agree with...and which could be made a lot simpler to read.

"Our shorter-priced selections win more often than our longer-priced ones."

The above quote is what thaskalos stated. It simply isn't true. It depends on your ability to pick winning horses, not the odds that the public makes them. It's actually possible for a really good handicapper to hit just as many longer shots as shorter ones. After all the public is wrong 66% of the time.

[Capper Al]

I am confused, then. Why did you ask the question if you knew the answer?

What question? This has always been a topic that I liked, so I jumped in the conversation.

[raybo]

Raybo, I make an odds line on every horse...and I find that my 2/1 shots win more often than my 4-1 shots...who in turn win more often than my 8-1 shots.

If YOU find that YOUR 8-1 shots win with the same frequency as your 2-1 shots...then more power to you!

It might mean that your odds line is flawed though...:)

I didn't say anything like that, did I? If you got that understanding then I'm sorry. I wasn't talking about a personal odds line. I was talking about post time odds, developed by the public's betting. It doesn't matter what the horse pays, your hit rate remains intact, if you are betting the horses you believe have the best chance of winning.

Now, I didn't say you would have a positive ROI under those conditions, either, but, the discussion was about hit rate and low odds vs higher odds, public odds, not personal odds lines.

[Capper Al]

I think that what you are saying is something we all agree with...and which could be made a lot simpler to read.

"Our shorter-priced selections win more often than our longer-priced ones."

Well put. Now consider a 25% strike ratio. He'll win 25 races out of 100, but probably 20 of the 25 will be under 3/1. This isn't like the die. We have another factor here, the Public.

[raybo]

Dave, your statement is basically true, but not entirely. I have a metric that I use to determine a longshot play, and my hit rate is way, way higher than their fair odds.

The answer one needs is not tote board odds groupings, but your own subjective odds, and hit rate.

Mike

Correct! I thought that was what I was saying.

It all depends on how good you are at picking winners. You don't have to bet if the odds aren't good enough. Say your minimum acceptable odds are 3/1, then your hit rate is based on odds of 3/1 and up. If you're applying the same handicapping method on all horses and you wager on only the top graded animal every time, then the public's odds mean nothing. The public only hits 1/3 of the time, they lose twice as many as they hit. There's plenty of room there for the good handicapper to get some of the longer priced horses that the public misses.

[Dave Schwartz]

Raybo,

I'm not trying to struggle with you.

I am sure you are an excellent player and likely very profitable. I just see it a little differently.


Dave

[thaskalos]

The above quote is what thaskalos stated. It simply isn't true. It depends on your ability to pick winning horses, not the odds that the public makes them. It's actually possible for a really good handicapper to hit just as many longer shots as shorter ones. After all the public is wrong 66% of the time.What isn't true Raybo? That our shorter-priced horses win more often than our longer-priced ones? My comment was made casually, without "heavy" statistical proof or detailed expanations.

IMO...this conversation has to do with the following question:

"Can a horseplayer with a 25% winning percentage become a winning player by just confining his bets to horses going off at odds of higher than 3-1?"

The obvious answer is NO!

He cannot be assured that this betting method will make him a winning player because his 25% hit-rate stats no doubt include horses 3-1 and lower. Eliminate these "short-priced" horses from your hit-rate stats...and you will find that your winning percentage no longer stands at 25%. It will now be a lot lower.

If we confine our bets to even money horses, our winning % will skyrocket...eventhough we may still lose money. If we restrict our plays to horses going off at 8-1 or higher, we may very well show a profit...but our winning % will be MUCH LOWER than usual.

That's what I meant by my comment, whose accuracy you have questioned. I dare say that it applies to almost all the players playing this game.

If it doesn't apply to you, consider yourself to be in the "tiny" - and most likely much wealthier - minority.

[raybo]

Raybo,

I'm not trying to struggle with you.

I am sure you are an excellent player and likely very profitable. I just see it a little differently.


Dave

I'm glad!

I was only commenting on a statement within a post that, in my opinion (seems like common sense to me), was not completely true, in all cases.

For instance, my hit rate on winners is only 29%, favorites and all, but, I have a 40+% hit rate on a particular "rule" horse that is always the 4th favorite at 1 minute to post. I think you'll agree that the 4th favorite will always be higher odds than the favorite, 2nd favorite and 3rd favorite.

If I have an example like that, what makes anyone think that there aren't players out here that have higher hit rates on higher priced horses than lower priced ones?

[Dave Schwartz]

Ray,

That is really impressive. What is your sample size on that play?

Dave

[raybo]

What isn't true Raybo? That our shorter-priced horses win more often than our longer-priced ones? My comment was made casually, without "heavy" statistical proof or detailed expanations.

IMO...this conversation has to do with the following question:

"Can a horseplayer with a 25% winning percentage become a winning player by just confining his bets to horses going off at odds of higher than 3-1?"

The obvious answer is NO!

He cannot be assured that this betting method will make him a winning player because his 25% hit-rate stats no doubt include horses 3-1 and lower. Eliminate these "short-priced" horses from your hit-rate stats...and you will find that your winning percentage no longer stands at 25%. It will now be a lot lower.

If we confine our bets to even money horses, our winning % will skyrocket...eventhough we may still lose money. If we restrict our plays to horses going off at 8-1 or higher, we may very well show a profit...but our winning % will be MUCH LOWER than usual.

That's what I meant by my comment, whose accuracy you have questioned. I dare say that it applies to almost all the players playing this game.

If it doesn't apply to you, consider yourself to be in the "tiny" - and most likely much wealthier - minority.

I didn't know that players with a "winning" percentage of 25% included horses that were not bet. Maybe I'm the only one that doesn't include races that I do not bet in my hit rate. To me, the races that I bet are the only ones that I'm interested in, and, the only ones I keep records on.

My mistake (I guess).

"Can a horseplayer with a 25% winning percentage become a winning player by just confining his bets to horses going off at odds of higher than 3-1?"

Yes! Obviously, because his 25% winning percentage only includes horses 3/1 and up, making his ROI over 3/1, maybe over 4/1, or 5/1. It depends on how many of each odds range he is actually betting.

That's my take on the statement you quoted. If your hit rate includes races that you don't bet then obviously it would be impossible for us to agree.

[raybo]

Ray,

That is really impressive. What is your sample size on that play?

Dave

536 plays over the last 6 years. (actually 5 of the last 6 years, I didn't play for a year)

[TrifectaMike]

I have a doctorate in theoretical math. I'm well known in my field. My methodologies are Bayesian based, and differ greatly from normal handicapping.

That said, let me outline my approach to the game.

I play this game as a game within a game against myself.

The tote odds are NEVER observed in my game. In place of the tote odds I have a model of expected tote odds. These odds are well calibrated and have fidelity to the rank of 5. This is the game I play against. In essence, I have modeled what to expect from the public at large.

My handicapping model attempts to exploit my expected tote odds model.

I do most of my modeling parametrically. I always use distributions ...parametric distributions, whenever possible and sample boosting when I don't have a good handle on a distribution.

I avoid regression ( I know Benter used regression, but he was in a closed environment) like the plaque with one exception and that is in my expected tote odds model. Averaging is NOT your friend.

Some of the metrics I have developed are unique. My models are constantly and automatically updated ( For you engineers, you can view it as closed loop adaptive control system).

Mike

[JustRalph]

Mike, Thanks for ruining a perfectly easy to follow thread............ :lol: ;)

[TrifectaMike]

While I'm on a roll, let me also add that since I have a well calibrated tote estimator, I can easily determine where the "whales" money is headed. How?

The odds follow a power law distribution that feeds upon itself.

Mike

P.S. I hope this makes it simpler. OK Ralph?

[JustRalph]

While I'm on a roll, let me also add that since I have a well calibrated tote estimator, I can easily determine where the "whales" money is headed. How?

The odds follow a power law distribution that feeds upon itself.

Mike

P.S. I hope this makes it simpler. OK Ralph?

Sure, whatever you say buddy! :lol:

[gm10]

I have a doctorate in theoretical math. I'm well known in my field. My methodologies are Bayesian based, and differ greatly from normal handicapping.

That said, let me outline my approach to the game.

I play this game as a game within a game against myself.

The tote odds are NEVER observed in my game. In place of the tote odds I have a model of expected tote odds. These odds are well calibrated and have fidelity to the rank of 5. This is the game I play against. In essence, I have modeled what to expect from the public at large.

My handicapping model attempts to exploit my expected tote odds model.

I do most of my modeling parametrically. I always use distributions ...parametric distributions, whenever possible and sample boosting when I don't have a good handle on a distribution.

I avoid regression ( I know Benter used regression, but he was in a closed environment) like the plaque with one exception and that is in my expected tote odds model. Averaging is NOT your friend.

Some of the metrics I have developed are unique. My models are constantly and automatically updated ( For you engineers, you can view it as closed loop adaptive control system).

Mike

Can you tell me how you use your expected tote odds model?

[Overlay]

The answer one needs is not tote board odds groupings, but your own subjective odds, and hit rate.
Wouldn't one's own odds based on objective data be more reliable and replicable from race to race?

[TrifectaMike]

Wouldn't one's own odds based on objective data be more reliable and replicable from race to race?

Subjective odds are based on all forms of data and beliefs about the data. I believe there is a misconception about being subjective in terms of probability on your part.

Mike

[TrifectaMike]

Can you tell me how you use your expected tote odds model?

I use it in the same manner as attempting to prove a coin is biased.

Mike

[gm10]

I use it in the same manner as attempting to prove a coin is biased.

Mike

And in terms of horse racing/betting? To find which horses will be overbet/underbet?

[TrifectaMike]

And in terms of horse racing/betting? To find which horses will be overbet/underbet?

Correct. To find which horses should be overbet/underbet.

Mike

[lamboguy]

536 plays over the last 6 years. (actually 5 of the last 6 years, I didn't play for a year)just for arguement sake lets use the number of 10% for a positive roi. and use the constant bet of $200 per race on all 536 races. you would have made about $17,000 in the 5 years. and that is before your cost of any racing data or other tools that one might use to create and make these bets. unless you are are playing on days like breeders cups, kentucky derby's and special days i doubt that you you could bet more than $200 a race without effecting odds in some manor. i just felt like pointing that out because i think your way of doing things is great and what the real problem is to show a worthwhile profit with your skills.

[Overlay]

Subjective odds are based on all forms of data and beliefs about the data. I believe there is a misconception about being subjective in terms of probability on your part.

Mike

Thank you for clarifying the sense in which you were using the word "subjective". At best, our probability determinations are a subjective estimate (even if we are using quantitative data to arrive at that estimate) of an objective number that exists (i.e., the actual chance of winning for each horse in a race), but that perhaps cannot be established (before-the-fact) with absolute certainty. Nevertheless, it is possible to make those determinations by manipulating objective data using statistical and mathematical principles in a manner that excludes personal opinion/judgment from the process. (That's the point that I was trying to make.)

[jfdinneen]

...In place of the tote odds I have a model of expected tote odds. These odds are well calibrated and have fidelity to the rank of 5. This is the game I play against. In essence, I have modeled what to expect from the public at large...

Mike,

Two questions, if I may?

First, in terms of the Bayesian model of expected public odds that you exploit, from what family of distributions are your prior and posterior distributions drawn?

Second, Benter (1994) emphasizes the creation of unbiased probability estimates (on assumption that any statistical model is incomplete) by adjusting his model's estimates to incorporate whatever additional information exists in the public's estimates. Can you elaborate on why you do not include this additional correction to your model?

John

[TrifectaMike]

Mike,

Two questions, if I may?

First, in terms of the Bayesian model of expected public odds that you exploit, from what family of distributions are your prior and posterior distributions drawn?

Second, Benter (1994) emphasizes the creation of unbiased probability estimates (on assumption that any statistical model is incomplete) by adjusting his model's estimates to incorporate whatever additional information exists in the public's estimates. Can you elaborate on why you do not include this additional correction to your model?

John

I use a power law distribution.

Benter attempts to gain an advantage over the public. My tote board model models expected behavior of the betting public at large.

Mike

[TrifectaMike]

On a thread, I can't remember when or who posted it, but it asked the question; Is the Twinspires conditional betting for odds lower than a predefined number? I believe it confused most, but it was a great question.

Mike

[jfdinneen]

Mike,

On what theoretical basis (references?) have you determined that a power law distribution is a "best-fit"?

John

[TrifectaMike]

Let's look at this in a simple form.

For example a coin flip. I have an expected coin flip model. Odds should be even money according to my model. That is what the betting public should aggregate to. My handicapping model observes a distortion. The "assumed" public says 40-60. My handicapping model attempts to find a reason for the distortion. If none is found, we have a biased coin and there is reason to expect there are players with the knowledge that the coin is biased.

Mike

[TrifectaMike]

Mike,

On what theoretical basis (references?) have you determined that a power law distribution is a "best-fit"?

John

John, it's based on my own studies.

Mike

[gm10]

Correct. To find which horses should be overbet/underbet.

Mike

Why not just grab a live feed of odds? Am I missing something?

[jfdinneen]

Mike,

Using the coin flip example, are you using a fundamental (ex. form-based) or a technical (ex. price movements) approach to finding the distortion and, also, how do you baseline the model (how do you know that you are participating in coin toss)?

John

[TrifectaMike]

Mike,

Using the coin flip example, are you using a fundamental (ex. form-based) or a technical (ex. price movements) approach to finding the distortion and, also, how do you baseline the model (how do you know that you are participating in coin toss)?

John

Fundamental approach to determine the baseline for the expected tote odds. The handicapping model seeks out distortions, and after digesting the information comes to a conclusion.

Mike

[jfdinneen]

Mike,

This brings us full circle to Benter's fundamental question:

How do you know that your fundamental analysis of baseline is better than public odds estimate?
If you are using publicly available information (ex. form), then according to Page (2007) Diversity Prediction Model (Being different is as important as being good), the crowd will outperform even the best individual over time due the diversity of their prediction models.

John

[Overlay]

If you are using publicly available information (ex. form), then according to Page (2007) Diversity Prediction Model (Being different is as important as being good), the crowd will outperform even the best individual over time due the diversity of their prediction models.

Page says "over time", but does that imply that it is impossible to selectively find favorable wagering situations (i.e., where the crowd's short-term judgment is faulty) on individual horses in individual races, and consistently bet them profitably, even if using publicly available information?

[TrifectaMike]

Mike,

This brings us full circle to Benter's fundamental question:


How do you know that your fundamental analysis of baseline is better than public odds estimate?
If you are using publicly available information (ex. form), then according to Page (2007) Diversity Prediction Model (Being different is as important as being good), the crowd will outperform even the best individual over time due the diversity of their prediction models.

John


John, I don't want my tote-odds model to be better. I have no interest in the model to outperform the crowd. I want it to reflect a well behaved betting public that will generate odds according to a well defined power law distribution.



Mike

[jfdinneen]

Mike,

Understood. However, this brings us back to the "best-fit" question below!
Incidentally, thanks for your willingness to provide additional details on your mathematical model - much appreciated.

John

[TrifectaMike]

Mike,

Understood. However, this brings us back to the "best-fit" question below!
Incidentally, thanks for your willingness to provide additional details on your mathematical model - much appreciated.

John

Enjoyable discussion. You have a great day.

Mike

[PaceAdvantage]

And thaskalos was complaining to me recently about the dearth of meaty subjects in the General Handicapping section...see buddy...all is not lost over in this area of the board! :lol:

Great thread guys...

[thaskalos]

And thaskalos was complaining to me recently about the dearth of meaty subjects in the General Handicapping section...see buddy...all is not lost over in this area of the board! :lol:

Great thread guys...If only I could understand some of the advanced mathematical "stuff"...:)

[Overlay]

If only I could understand some of the advanced mathematical "stuff"...:)

I have found that the real mark of an expert in a subject (no matter how many formal credentials one can claim, or how great the degree of one's technical knowledge) is being able to convey that information in a manner that someone without the same background or depth of study can comprehend. To me, this thread is falling short in that respect (as your post illustrates), regardless of how much "meat" there may be in it for some.

[thaskalos]

I have found that the real mark of an expert in a subject (regardless of how many formal credentials one can claim, or how great the degree of one's technical knowledge) is being able to convey that information in a manner that someone without the same background or depth of study can comprehend. To me, this thread is falling short in that respect (as your post illustrates), regardless of how much "meat" there may be in it for some.I could not agree more. :ThmbUp:

We don't REALLY understand something, until we are able to explain it in simple language.

I am reminded of Ernest Hemingway...who wrote a long, complicated letter to a friend, and apologized at the end of it...for not having the necessary time to write a simple one.

[Robert Fischer]

I have found that the real mark of an expert in a subject (no matter how many formal credentials one can claim, or how great the degree of one's technical knowledge) is being able to convey that information in a manner that someone without the same background or depth of study can comprehend. To me, this thread is falling short in that respect (as your post illustrates), regardless of how much "meat" there may be in it for some.

Even Magic Johnson will tell you that "a good pass is simply one that's caught".

However, I'm not sure Trifecta Mike really wants to convey every technical detail of his hard earned process into layman's terms (or even those of a professional) in so much as share a bit of perspective and possibly discuss general ideas.

[Robert Fischer]

on the concept of whether or not to "do a Benter" and normalize your opinion(estimate) to some degree by using the actual tote odds(public) to arrive at Actual Odds estimate -

Ultimately I don't do it.

I find that the amount of "weight" to give the Public's odds depends upon my level of insight into THE REASON FOR THE DIFFERENCE between my estimate(of the public's odds) and the public's tote odds.

There are times when I'm waiting for a horse to run back and(because) I know that the public will probably make a very poor estimate,
and I know THE REASON it is going to happen.

In other words my insight or understanding is strong enough that I expect the odds to be wrong in a certain direction, by a certain amount.

When this happens, the result of normalizing a small amount, against what is already a big edge, is not significant, and I find that I never? end up changing my decision to wager in these cases.
I suppose if I was large-scale player, I would study the value of making small adjustments to wager amounts to strive for best efficiency.

What about when I don't know why the public's odds are different??
Well in that case of course, I am so good that they must have inside information... Just kidding... more like a mix of some races with "smart money" along with others where I am ignorant. However, when I get to this point in the process, I take the easy way out, and abstain from wagering on this event.

Note there is a "loop" to this process. When I see significant unexpected money on a horse, in certain pools, and even certain "types" of horses(call 'em indicators), I will anticipate that there may be inside information. Should the scenario continue to unfold into additional pools that I require to meet these spot plays, I will then search for still other pools(different wagerTypes) where the inside information didn't happen to greatly depress that horse's odds.

[Overlay]

However, I'm not sure Trifecta Mike really wants to convey every technical detail of his hard earned process into layman's terms (or even those of a professional) in so much as share a bit of perspective and possibly discuss general ideas.

Posters (speaking in general terms) are entitled to discuss as much or as little of their approaches as they wish, and no one would begrudge them that. But if they are going to use specialized terms in what they do choose to post, they should not make the assumption that everyone has a level of comprehension similar to their own. They should either explain terms used, or express the same concepts in a non-technical fashion that can be more readily understood. Otherwise, I would think that, even if the posters might genuinely wish to (as you put it) share a bit or perspective and discuss general ideas, they will achieve the opposite of that objective, since people will perceive them as being unclear (for whatever reason, deliberate or not), and will tune them out.

[Gallop58]

Am I following, in that the leading rationale for modelling the expected and looking for the unexpected, is this "loop" exercise and finding ways to bet on or against the "mystery" money. Is that essentially the end game for the modelling and decisions being discussed? (Or have I fumbled Magic's pass out of bounds?)

"When I see significant unexpected money on a horse....I will then search for still other pools(different wagerTypes) where the inside information didn't happen to greatly depress that horse's odds."

[Enigma]

This is an interesting thread. However, if I may I would like to make a constructive criticism.There is way too much technical jargon. My experience is to be smart and keep it simple. 33.0% of races are won be the favorite and the first three in the betting win approximately 65.0% Why go anywhere else?

Using conventional capping, speed, class,recency, plus a speed/distance/surface comparison chart and an odds percentage table. Wait until your selection has a 30.0% edge. Wager where you can obtain a 5.0% rebate and you have got it made.

Will you make a fortune? Not likely but you will be a winning player.
I keep hearing references to Joe Public, we are Joe Public.

[Dave Schwartz]

Overlay,

By "doing a Benter," he is referring to what I call "tote smoothing." That is, changing the probabilities in the direction of the public's betting.

Just so you know, Benter advocated that 70% of his handicapping output was the tote board.

My experience indicates that some form of smoothing is an absolute necessity if one expects to make a realistic line.

Regards,
Dave Schwartz

[Dave Schwartz]

Enigma,

30% edge over what?
Do you mean $2.60 $net (where $2.00 is break even)? That is pretty rare to see.

I recently ran a 200-race sample playing "optimum bet." The average bet size was $273 in the sample. Those plays that called for a $500 wager or more produced 43% winners and a $2.30 $Net. I was very surprised!


30% edge in what?
Do you mean 30% edge over the next horse in win probability?


Dave
Who is always ready to
learn something new.

[Robert Fischer]

Am I following, in that the leading rationale for modelling the expected and looking for the unexpected, is this "loop" exercise and finding ways to bet on or against the "mystery" money. Is that essentially the end game for the modelling and decisions being discussed?

(oops see Dave's post as well^^ i type slow)
one of Benter's popular ideas is that he advocates adjusting your estimate of what the "Actual Odds" should be for a horse by using the Tote Odds.

"We found that if we gave a horse, say, a 20 per cent chance of winning, and the public odds reflected a 10 per cent chance, then the true odds were nearer 16 per cent. Those who bet in the pools have information which is not available to us - track gossip, stable money, that kind of thing - and we learnt to take account of it."

I think if you bet a lot of races, and your goals include skimming a small-moderate edge for the best that you can, this is something that makes a bunch of sense.

Above I'm saying that for my bet-back spot plays, I don't adjust.

The "loop" about noticing smart-money, and exploiting it isn't as popular although a lot of players probably have a pet play like that or even do it instinctively when they see the pools.
I had one the other day at a mid-level track when I saw the daily-double pool for the next race. A 1st-time-starter who should have been 5-1 was being unexpectedly hammered under the 3 likely winners. - Next race comes up and he is hovering at about his morningLine odds - until a minute to go when his win price crashed. There were a couple other pools that they didn't depress his odds in.

[Overlay]

Overlay,

By "doing a Benter," he is referring to what I call "tote smoothing." That is, changing the probabilities in the direction of the public's betting.

Just so you know, Benter advocated that 70% of his handicapping output was the tote board.

My experience indicates that some form of smoothing is an absolute necessity if one expects to make a realistic line.

Regards,
Dave Schwartz

Thank you for the clarification, Dave, but I was not posting for my own benefit, nor was that the only word or term to which I was referring. A sampling from this thread of terms for which less-mathematically-inclined individuals might appreciate some additional definition or explanation (if the posters intend to provide it):

Pareto (Power Law) Distribution
Bayesian probabilities
fidelity to the rank of 5
parametric modeling
regression
Benter regression
closed environment
closed loop adaptive control system
family of distributions
prior distribution
posterior distribution
best fit
well-behaved betting public

[Enigma]

Dave,

My wager has to be in the first three in the betting and have a 30.0% edge over the second choice in my final ratings. This of course leads to many favorites which are more likely to be valid than vulnerable which is a distinct advantage when trying to take advantage of a 5.0% rebate.

[Capper Al]

Gees, I don't recognize this thread anymore. But we did get to the bottom of it. A 25% can't make a profit betting 3/1 or better horses because most of his hits will be under 3/1.

[raybo]

just for arguement sake lets use the number of 10% for a positive roi. and use the constant bet of $200 per race on all 536 races. you would have made about $17,000 in the 5 years. and that is before your cost of any racing data or other tools that one might use to create and make these bets. unless you are are playing on days like breeders cups, kentucky derby's and special days i doubt that you you could bet more than $200 a race without effecting odds in some manor. i just felt like pointing that out because i think your way of doing things is great and what the real problem is to show a worthwhile profit with your skills.

lambo,

Not real sure what you're getting at here. You proposed a 10% positive ROI, and $200 flat bet. I agree you probably will affect the post time odds on 4th favorites with that bet amount.

Were you referring to me having a 10% ROI, with this particular wager? If so, I stated in the other post that this wager has a 40+% hit rate. I neglected to say that the odds must fall between 4/1 and 8/1, as the 4th favorite, at 1 minute to post. I think 10% ROI is a little low, for this play, don't you?

Further, as you probably know, I don't bet WPS, only superfectas. This "rule" play is an automatic "move", to the win position, no matter where I have the horse graded It could have graded on the 2nd, 3rd, or 4th ticket positions, but because of the "rule" it automatically becomes my win horse, on that particular superfecta ticket.

The reason I got into this discussion is because, although I don't bet win bets, I do have to have the winner to cash my superfectas. This particular "rule" play attracted my attention several years ago and I started tracking the hit rate, for it. The fact that it hits 40+% of the time, not only gives me a very good chance of hitting the top of my ticket, it also makes for large to very large payouts, when other odds spread criteria are in place, which they often are because most of the races where this anomaly takes place turn out to be chaos races with the vast majority of superfecta players not hitting the super.

[raybo]

Gees, I don't recognize this thread anymore. But we did get to the bottom of it. A 25% can't make a profit betting 3/1 or better horses because most of his hits will be under 3/1.

Really?

See post 34.

[Capper Al]

Really?

See post 34.

Interesting.

[Dave Schwartz]

My wager has to be in the first three in the betting and have a 30.0% edge over the second choice in my final ratings. This of course leads to many favorites which are more likely to be valid than vulnerable which is a distinct advantage when trying to take advantage of a 5.0% rebate.

Enigma,

So, the horses have a score... 30% up or down?

In other words, 2 horses:

1: 500
2: 650 makes #2 a play? (500 x 1.30 = 650)

1: 500
2: 350 makes #1 a play? (500 x .70 = 350)


Dave

[Cratos]

I have a doctorate in theoretical math. I'm well known in my field. My methodologies are Bayesian based, and differ greatly from normal handicapping.

That said, let me outline my approach to the game.

I play this game as a game within a game against myself.

The tote odds are NEVER observed in my game. In place of the tote odds I have a model of expected tote odds. These odds are well calibrated and have fidelity to the rank of 5. This is the game I play against. In essence, I have modeled what to expect from the public at large.

My handicapping model attempts to exploit my expected tote odds model.

I do most of my modeling parametrically. I always use distributions ...parametric distributions, whenever possible and sample boosting when I don't have a good handle on a distribution.

I avoid regression ( I know Benter used regression, but he was in a closed environment) like the plaque with one exception and that is in my expected tote odds model. Averaging is NOT your friend.

Some of the metrics I have developed are unique. My models are constantly and automatically updated ( For you engineers, you can view it as closed loop adaptive control system).

Mike


A very good post and your statement: “The tote odds are NEVER observed in my game. In place of the tote odds I have a model of expected tote odds” should be remembered by a bettor who is looking to exploit the general betting public (and that is the secret to this game).

However the possible pitfall with the use of parametric methods in your handicapping of the expected tote odds is possible faulty assumptions. It is my understanding that parametric methods are often not considered robust (am I correct?). What I am getting at is that your method probably emulates the popular statistic methods, but is not immoderately affected by small differences in the assumptions between the two.

[gm10]

Let's look at this in a simple form.

For example a coin flip. I have an expected coin flip model. Odds should be even money according to my model. That is what the betting public should aggregate to. My handicapping model observes a distortion. The "assumed" public says 40-60. My handicapping model attempts to find a reason for the distortion. If none is found, we have a biased coin and there is reason to expect there are players with the knowledge that the coin is biased.

Mike

I am fascinated by this post, maybe because I don't understand it :blush: . Who is the "assumed" public above? Are they the 'people' that drive your tote model?

I am also a bit confused wrt your use of the term 'handicapping model'. You say that it observes a distortion, it attempts to find a reason for the distortion. What is it actually modelling if it's doing all of this?

[raybo]

I am fascinated by this post, maybe because I don't understand it :blush: . Who is the "assumed" public above? Are they the 'people' that drive your tote model?

I am also a bit confused wrt your use of the term 'handicapping model'. You say that it observes a distortion, it attempts to find a reason for the distortion. What is it actually modelling if it's doing all of this?

If there is money going on a horse that shouldn't be, then you need to find out why. If you can't, then something fishy is going on. Now, do you jump on the bandwagon, or pass the race.

I think that when this happens, he assumes that the excess money on the horse is "smart" money, "insider" money, or whatever you want to call it.

My "rule" play is about exactly this.

[Enigma]

Dave,

My only wager is on my top rated horse provided it has a 30.0% or more edge over the rest of the field. If the top rated horse does not meet this criteria I pass.

[TrifectaMike]

Thank you for the clarification, Dave, but I was not posting for my own benefit, nor was that the only word or term to which I was referring. A sampling from this thread of terms for which less-mathematically-inclined individuals might appreciate some additional definition or explanation (if the posters intend to provide it):

Pareto (Power Law) Distribution
Bayesian probabilities
fidelity to the rank of 5
parametric modeling
regression
Benter regression
closed environment
closed loop adaptive control system
family of distributions
prior distribution
posterior distribution
best fit
well-behaved betting public

Point taken.


Pareto (Power Law) Distribution
The distribution of the final odds in a horse race shows a power law
P(Odds) is directly proportional to 1/Odds

The interesting aspect of this is that it agrees with empirical data (real life odds). And it fits the data only if you assume that bettors are irrational (bettors do not try to maximize their expected rewards (winnings)). If the assumption is made that bettors are rational (bettors try to maximize their expected rewards), we would obtain a different power law,
P(Odds) is directly proportional to 1/(Odds times Odds),
which is different from empirical data.

Looking at the distribution of the winners odds it also follows a power law distribution,
P(Odds) is directly proportional to 1/(Odds times Odds).

Bayesian Probability
Instead of defining as it exist in the math world (you can google it), I'll simply describe why I use Bayesian by describing a Bayesian belief network.

A Bayesian belief network uses probability to decide the most likely outcome. This is accomplished by combining past events and current data dependent and independent variables. A belief network can reach reasonable conclusions about potential outcomes.

Fidelity to rank of 5
Used in reference to expected tote odds model, simply means I want to be accurate with the top 5 in odds rank.

Parametric Modeling
Parametric models are built from a set of mathematical equations. In order for parametric models to have any validity they must be proven on actual data. it is the sophistication of the data analysis methods and the extensiveness of the underlying data which determines the efectiveness of the modeling solution.

Overlay, if this is helpful I'll continue.

[Overlay]

Overlay, if this is helpful I'll continue.

Please do.

[gm10]

If there is money going on a horse that shouldn't be, then you need to find out why. If you can't, then something fishy is going on. Now, do you jump on the bandwagon, or pass the race.

I think that when this happens, he assumes that the excess money on the horse is "smart" money, "insider" money, or whatever you want to call it.

My "rule" play is about exactly this.

Cheers.

Any ideas on how this 'handicapping model' tries to explain the distortion?

[teddy]

OK
Sat the 13th
Hollywood park Race 2 number 6


This was and Enigmas pick.. I cant for the life of me see how hes 30% better than the other two ml pics. Anyone want to chime in. The 3 and 8 both ran faster than him at DMR.

[Enigma]

Teddy,

The horse does not have be 30.0% better than the other two horses in the morning line betting...it has to be 30.0% better than the other two horses in my top three contenders.

[TrifectaMike]

Thank you for the clarification, Dave, but I was not posting for my own benefit, nor was that the only word or term to which I was referring. A sampling from this thread of terms for which less-mathematically-inclined individuals might appreciate some additional definition or explanation (if the posters intend to provide it):

Pareto (Power Law) Distribution
Bayesian probabilities
fidelity to the rank of 5
parametric modeling
regression
Benter regression
closed environment
closed loop adaptive control system
family of distributions
prior distribution
posterior distribution
best fit
well-behaved betting public

Regression
A regression is a statistical analysis assessing the association between variables. Simple regression is used to find the relationship two variables:
Regression Equation Y = a + bx
where x and y are variables
b is the slope of the regression line
a is the intercept point of the regression line and the y axis

Multiple regression is a regression equation with multiple independent variables.
Regression Equation y = c0 + c1x1 + c2x2 + .... + cnxn

Benter Regression
I didn't use the term, but I assume it refers to Benter's use of Logistic regression to derive a probability curve and then using a pooling scheme to incorporate the tote odds. Logistic regression is a technique employed when the dependent variable is a dichotomy ( two states, yes/no, win/lost etc.)

Closed Environment
I used this term in referencing Benter's method. I stated that he operating in a closed environment. The horses he was observing was a closed set(same horses same track)

Closed loop adaptive control system
I only used this term to give those technically astute in the area of control systems an idea of the over mechanism. An adaptive control systems corrects problems of variability by making changes to "something" in the system based on what it has "learned" about the system.

Mike

[Robert Fischer]

my posterior distribution has gotten worse over the years :liar:

[teddy]

Teddy,

The horse does not have be 30.0% better than the other two horses in the morning line betting...it has to be 30.0% better than the other two horses in my top three contenders.


Much thanks for the explanation. Actually much more interesting. Didnt get to match picks with you today.. got on too late. None matched my picks anyway. GOOD LUCK

[Cratos]

Regression
A regression is a statistical analysis assessing the association between variables. Simple regression is used to find the relationship two variables:
Regression Equation Y = a + bx
where x and y are variables
b is the slope of the regression line
a is the intercept point of the regression line and the y axis

Multiple regression is a regression equation with multiple independent variables.
Regression Equation y = c0 + c1x1 + c2x2 + .... + cnxn

Benter Regression
I didn't use the term, but I assume it refers to Benter's use of Logistic regression to derive a probability curve and then using a pooling scheme to incorporate the tote odds. Logistic regression is a technique employed when the dependent variable is a dichotomy ( two states, yes/no, win/lost etc.)

Closed Environment
I used this term in referencing Benter's method. I stated that he operating in a closed environment. The horses he was observing was a closed set(same horses same track)

Closed loop adaptive control system
I only used this term to give those technically astute in the area of control systems an idea of the over mechanism. An adaptive control systems corrects problems of variability by making changes to "something" in the system based on what it has "learned" about the system.

Mike

Mike, the term definitions given by you are well taken, but am I thinking correctly in post #81 of this thread?

It would seem to me that the accuracy of the assumptions that you have in your model is the heart and soul of its success.

I don’t expect you to give out your assumptions, but it would be good to know if I am thinking correctly about your model.

[Cratos]

..................I am also a bit confused wrt your use of the term 'handicapping model'. You say that it observes a distortion, it attempts to find a reason for the distortion. What is it actually modelling if it's doing all of this?

Behavior of the wagering bettors and to me this is a fascinating endeavor because this is where profitability lies in horseracing; exploiting that behavior.

Also in general this is what speed figure handicappers, pace figure handicappers, etc have missed in their assessment of handicapping the race and that is the understanding of the risk profile of the bettors in the wagering pool.

It has been clearly defined by many studies that there are three major risk profiles of bettors, but the interaction between them become complicated and not easily understood without behavioral modeling.

[TrifectaMike]

A very good post and your statement: “The tote odds are NEVER observed in my game. In place of the tote odds I have a model of expected tote odds” should be remembered by a bettor who is looking to exploit the general betting public (and that is the secret to this game).

However the possible pitfall with the use of parametric methods in your handicapping of the expected tote odds is possible faulty assumptions. It is my understanding that parametric methods are often not considered robust (am I correct?). What I am getting at is that your method probably emulates the popular statistic methods, but is not immoderately affected by small differences in the assumptions between the two.

Your understanding is correct.
Mike

[raybo]

Cheers.

Any ideas on how this 'handicapping model' tries to explain the distortion?

I'm sure if he told us exactly, he would be giving away too much information.

My guess is that the modeling looks for things that some players will bet, like track's best jockey, best trainer, early or late speed in last race, change in surfaces or distances, etc., anything that might bring money into the pools that, according to the model, the horse doesn't deserve. In other words, it's unexplainable.

I doubt the model tries to explain the distortion, rather it accepts that a distortion exists.

That's what my "rules" play is about. A horse drops into the top 4 horses, on the tote, from outside the top 4, at some point during the wagering (I start tracking the horse at about 10 or 12 minutes to post). Dropping into the top 4 is not unusual, it happens all the time. What is unusual about this play is that, after the drop into the top 4, there is a methodical fluctuation in odds, dropping then rising, then dropping, then rising, but never below 4/1. And, at 1 minute to post the horse is in the top 4 and it's odds are between 4/1 and 8/1, with no reason for being there, the horse looks bad, outclassed, no speed, no endurance, no top jock, no top trainer, no class drop, no bullet work, no nothing.

Why would I believe that there is "smart" money on this horse. Not only is the money "smart", the way it has been bet was also smart.

Let's say that you have inside information, not available to the public, and you want to wager a large sum of money on the horse, but you don't want to cut your own throat by driving it's odds down too much. How would you do it?

The smart way would be to wait until the pools have a significant amount of money in them, so your bets don't have as much impact. Then you drop a portion on the horse, the odds drop, then you wait for the public to see it and put more money in the pool, a click or 2 later the odds start going back up, you drop some more on it, the odds drop again, but not as much because there is more money in the pool, then you wait for the public, a click or 2 later the odds start rising again, then you drop some more, etc., etc..

My observations about this anomaly is that a "true" representation of this type horse (I actually call the horse a "recipe horse") have been that the horse never gets bet below 4/1 by 1 minute to post, and it never is allowed to rise above 8/1, by 1 minute to post. What happens after 1 minute to post? It usually drops slightly, (but usually they go off at between 4/1 and 8/1), as the wagerer gets the remainder of his money on the horse, after most players have stopped betting. This additional money isn't noticed by the public, therefore, the only drop in odds that occurs after 1 minute to post, occurs due to that wagerer getting the rest of his money on the horse.

I've posted about my "recipe horses" several times in the past, and as I said, I've been playing these horses for 5 of the last 6 years, and the hit rate is over 40%.

I don't try to explain why it happens, I just accept that it fits the pattern and I play the horse on top of my super, if I decide to play the race at all, which I normally do because of the potential for a large payout.

[MPRanger]

Interesting but it seems like you are handicapping yourself vs the tote board instead of the horses vs each other. And I'm sure that's important like if you are using the kelly criterion to determine your bet size or something.

However...

If you are truly able to quantify the value of the horses relative to each other, you should be able to capitalize when the toteboard offers value. The stochastic nature of the game assures there will always be value to be had. The public/toteboard can win it's subset and other subsets which yield bankroll growth are available to sharp handicappers who understand how to bet in relation to the true odds, the payout and their bankroll size.

Respectfully,
The Ranger

[gm10]

I'm sure if he told us exactly, he would be giving away too much information.

My guess is that the modeling looks for things that some players will bet, like track's best jockey, best trainer, early or late speed in last race, change in surfaces or distances, etc., anything that might bring money into the pools that, according to the model, the horse doesn't deserve. In other words, it's unexplainable.

I doubt the model tries to explain the distortion, rather it accepts that a distortion exists.

That's what my "rules" play is about. A horse drops into the top 4 horses, on the tote, from outside the top 4, at some point during the wagering (I start tracking the horse at about 10 or 12 minutes to post). Dropping into the top 4 is not unusual, it happens all the time. What is unusual about this play is that, after the drop into the top 4, there is a methodical fluctuation in odds, dropping then rising, then dropping, then rising, but never below 4/1. And, at 1 minute to post the horse is in the top 4 and it's odds are between 4/1 and 8/1, with no reason for being there, the horse looks bad, outclassed, no speed, no endurance, no top jock, no top trainer, no class drop, no bullet work, no nothing.

Why would I believe that there is "smart" money on this horse. Not only is the money "smart", the way it has been bet was also smart.

Let's say that you have inside information, not available to the public, and you want to wager a large sum of money on the horse, but you don't want to cut your own throat by driving it's odds down too much. How would you do it?

The smart way would be to wait until the pools have a significant amount of money in them, so your bets don't have as much impact. Then you drop a portion on the horse, the odds drop, then you wait for the public to see it and put more money in the pool, a click or 2 later the odds start going back up, you drop some more on it, the odds drop again, but not as much because there is more money in the pool, then you wait for the public, a click or 2 later the odds start rising again, then you drop some more, etc., etc..

My observations about this anomaly is that a "true" representation of this type horse (I actually call the horse a "recipe horse") have been that the horse never gets bet below 4/1 by 1 minute to post, and it never is allowed to rise above 8/1, by 1 minute to post. What happens after 1 minute to post? It usually drops slightly, (but usually they go off at between 4/1 and 8/1), as the wagerer gets the remainder of his money on the horse, after most players have stopped betting. This additional money isn't noticed by the public, therefore, the only drop in odds that occurs after 1 minute to post, occurs due to that wagerer getting the rest of his money on the horse.

I've posted about my "recipe horses" several times in the past, and as I said, I've been playing these horses for 5 of the last 6 years, and the hit rate is over 40%.

I don't try to explain why it happens, I just accept that it fits the pattern and I play the horse on top of my super, if I decide to play the race at all, which I normally do because of the potential for a large payout.

How do you monitor these horses, if I may ask?

[Dave Schwartz]

Raybo,

That's brilliant!


Dave

[raybo]

How do you monitor these horses, if I may ask?

All my handicapping is done well before the wagering begins. All I do, after the wagering starts, is make sure no changes have been made, scratches, jockey changes, etc.. If a horse scratches, I scratch the horse in AllData and re-rank the horses.

After that, all I do is watch the tote. 1 of my priorities, while watching the tote, is to look for changes in the top 4 morning line horses. If a horse drops into the top 4, at any time, I start watching his/her win odds. Sometimes more than 1 horse moves into the top 4, but usually only 1 is left by 1 minute to post. If more than 1 are still in the top 4, then my "recipe horse" play is not valid.

I look for the recipe betting pattern to begin and continue until just before post time. If a recipe horse exists, at 1 minute to post, I put him/her on the top of my superfecta ticket, moving all horses down, in rank, 1 position (usually I have already spotted a recipe before 1 mtp, which gives me plenty of time to restructure my ticket).

Then I just place the wager, or if something else disqualifies the race, I pass.

[raybo]

Raybo,

That's brilliant!


Dave

Thanks Dave, but, it's not really brilliant, to me it just makes sense. Like I said, the pattern just emerged enough times that it drew my attention and I started tracking it.

I don't worry about sharing the knowledge because I don't bet to win anyway. Sharing it is not likely to affect the payouts on my superfectas.

Keep in mind that I have only tracked these horses in races that have superfecta wagering. It seems logical that it would apply to all races, however. The betting pattern just makes sense to me.

[Dave Schwartz]

Ray,

There aren't 3 people in this forum - including me - that would work that hard at a single race. Your non-secret is safe.


Dave

[JustRalph]

When my Dad was in his last days and too ill to handicap, just watch races, he would watch the tote and every once in a while he would describe to me what he was viewing on the tote. He would ask me to make a small bet on a horse every once in a while. He was right more often than not. He attributed it to years of experience watching the way money comes in. I was never very good at it. A real art form I guess.

If it works for you Raybo........ good on you! It's not my game. But sounds like you have a solid view of it. And having seen it, I don't doubt you much at all.

[Pell Mell]

Ray,

There aren't 3 people in this forum - including me - that would work that hard at a single race. Your non-secret is safe.


Dave

Unless sometime in the near future Dave has another Eureka moment and decides to offer a new program dealing with tote watching. :lol:

[raybo]

Ray,

There aren't 3 people in this forum - including me - that would work that hard at a single race. Your non-secret is safe.


Dave

Oh I think that there are countless members here who work much harder on single races. It's not that much work anyway. You gotta look at something while you're waiting, might as well be the tote.

[thaskalos]

Ray,

There aren't 3 people in this forum - including me - that would work that hard at a single race. Your non-secret is safe.


DaveDave...

Wouldn't you agree that the current full-card simulcasting landscape greatly complicates the tote board analysis process?

[lsosa54]

Ray: You're only looking at 1 or 2 tracks on a daily basis for your plays, correct?

[raybo]

Ray: You're only looking at 1 or 2 tracks on a daily basis for your plays, correct?

That's correct. Max of 2 tracks, 3 days per week.

[Dave Schwartz]

Wouldn't you agree that the current full-card simulcasting landscape greatly complicates the tote board analysis process?

Sure. Except for this guy:

http://ultimateselector.com/bulboard/viewtopic.php?f=48&t=18

[bigmack]

Sure. Except for this guy:

http://ultimateselector.com/bulboard/viewtopic.php?f=48&t=18
:lol:

Oh, yeah! That's the good stuff.

[raybo]

Dave...

Wouldn't you agree that the current full-card simulcasting landscape greatly complicates the tote board analysis process?

Yes, one of the reasons I wager online.

[gm10]

All my handicapping is done well before the wagering begins. All I do, after the wagering starts, is make sure no changes have been made, scratches, jockey changes, etc.. If a horse scratches, I scratch the horse in AllData and re-rank the horses.

After that, all I do is watch the tote. 1 of my priorities, while watching the tote, is to look for changes in the top 4 morning line horses. If a horse drops into the top 4, at any time, I start watching his/her win odds. Sometimes more than 1 horse moves into the top 4, but usually only 1 is left by 1 minute to post. If more than 1 are still in the top 4, then my "recipe horse" play is not valid.

I look for the recipe betting pattern to begin and continue until just before post time. If a recipe horse exists, at 1 minute to post, I put him/her on the top of my superfecta ticket, moving all horses down, in rank, 1 position (usually I have already spotted a recipe before 1 mtp, which gives me plenty of time to restructure my ticket).

Then I just place the wager, or if something else disqualifies the race, I pass.

Sounds like you have a winning strategy there. Thanks for sharing.