Handicap The Late Money

[lansdale]

Originally Posted by JerryBoyle

I have a model that WOULD win if I knew the final odds when I made a bet. Unfortunately, as we all know, the odds swings between at the gate and after start are so large that it's hard to know if it'll actually win .

Hi Jerry,

I may be a little confused here, but not sure why you wouldn't know if you have an edge vs. the final odds, aka 'reality.' I well understand the feeling of frustration with the sudden odds drop, but if you step back, isn't this problem (at least for a player using a model, like yourself) ultimately irrelevant? If you recall Benter's well-known article, he was comparing his own model with both the public model (pre-race known odds) and the final odds. Obviously, to have a winning model, you must be have an edge over the final odds. You talk about the final odds as being a 'god' model, but it's not. But to the extent that it's a model that it includes relevant information about the odds not contained in your model, it may be a stronger model than yours. And regardless, it should be easy enough, given your skills, to determine, as Benter did, whether you have an edge over the final odds -- the only odds that matter.

FWIW, Magister Ludi, although he rarely posts here any longer (though he appears to sometimes lurk) was using a math model as part of a syndicate and mentioned an algorithm he used to estimate (accurately enough, as he put it) final odds, to deal with this problem. Tend to doubt he would respond to a PM, but if he spots this thread, he might have some relevant comments.

[Nitro]

Originally Posted by jay68802

How would you go about doing this? You need the pool totals for a certain time, lets say when the first horse loads into the gate. You need the final pool total. Track the horses that had been bet at 3-1 or lower in the last dump.

What else?

Hi there Jay!
I’m afraid the premise of what you’re suggesting is of no real value. As others have already suggested; from a practical standpoint even if you knew what the final odds might be you couldn’t take advantage of that information. The Time factor for placing a wager is obviously over. More importantly there’s absolutely no guarantee that any of the entries getting very late action (in terms of changing odds) will win the race.

When you’re actually following the money from when the betting starts with a sophisticated tote analysis (which is monitoring much more than just the Win pool (the odds)), you’re able to get a much clearer picture of which entries are getting played in earnest. These tell-tale signs are provided by understanding specific betting patterns that allows you compare the betting on each entry in real time with the overall betting on all the entries combined.

Unfortunately, these betting patterns are not the same from track to track or even from race to race. However once they become familiar there’s no need to wait for the last minute to decide on which entries are live. By viewing these betting patterns at specific time intervals prior to post time it sometimes becomes very obvious which entries are playable. The tote analysis doesn’t discriminate by the odds. In other words a 10/1 shot could look just as promising as a 2/1 shot. The question then becomes what is the best approach to taking advantage of those entries that show promise. Fortunately, there are many types of bets that can be made.

For those who might have any doubts about what I’ve described and the method that I use, then I would suggest visiting the Selection forum once and a while. I have no problem occasionally sharing some worthwhile information.

[JerryBoyle]

Originally Posted by lansdale

Hi Jerry,

I may be a little confused here, but not sure why you wouldn't know if you have an edge vs. the final odds, aka 'reality.' I well understand the feeling of frustration with the sudden odds drop, but if you step back, isn't this problem (at least for a player using a model, like yourself) ultimately irrelevant? If you recall Benter's well-known article, he was comparing his own model with both the public model (pre-race known odds) and the final odds. Obviously, to have a winning model, you must be have an edge over the final odds. You talk about the final odds as being a 'god' model, but it's not. But to the extent that it's a model that it includes relevant information about the odds not contained in your model, it may be a stronger model than yours. And regardless, it should be easy enough, given your skills, to determine, as Benter did, whether you have an edge over the final odds -- the only odds that matter.

FWIW, Magister Ludi, although he rarely posts here any longer (though he appears to sometimes lurk) was using a math model as part of a syndicate and mentioned an algorithm he used to estimate (accurately enough, as he put it) final odds, to deal with this problem. Tend to doubt he would respond to a PM, but if he spots this thread, he might have some relevant comments.

Hey lansdale, my answer wasn't super clear before. I'm confident that I DO have an edge over the final odds, when using the final odds as an input to my model. I'm using the standard 2 step model that Benter advocates in his paper - I generate a fundamental model as step 1 that only considers things relating to the horse, track, trainer, etc. I then merge this into an ensemble model with 2 factors - the fundamental model estimate and the market estimate. The output of this is my final estimate.

The problem is that, as we know, the market estimate varies wildly up until all money is totaled after the race has started. This leads to a modeling problem and a testing problem. First, when I train that ensemble model, do I use the final market estimate or the market estimate as it'd be known when I place my bet, say 30 seconds prior to close of betting? Second, when I actually test how the model would have done via simulating betting, do I use the final price, or do I use the price I would have seen at 30 seconds to close? Let's say final estimate is option A and 30 seconds to close estimate is option B.

A: leads to the strongest model. I know FOR SURE that the ensemble prediction I generate, which combines the fundamental estimate and the final market estimate has an edge over the final market price. As you've said, this is easy to test in any number of ways. However, when I actually go to use the model to bet in real time, I can't combine the final odds with my fundamental estimate, because I don't know the final odds. My options are to either use odds at 30 seconds to post or some estimate of the final odds. Both lead to uncertainty with regards to knowing how the system will perform compared to my backtest. I can test how this set up would perform by using the odds at 30 seconds to post historically, but the data I have doesn't go back very far, so I can't be very confident.

B: a weaker model, but a better representation of how the system would work in real time. I train the ensemble model using the fundamental estimate and the market estimate at 30 seconds to close. Simulating betting using the 30 seconds to close price and this model generally best simulates how the model would bet in real time. It also generally results in over betting, as the final price is almost always more efficient for the bets the model would make.

In the past, I've been using option B. However, I'm starting to think the best route is option A followed by developing a model which, as you've said, estimates final price. I'm not finished researching that yet, but I'm starting to believe that it's possible to build a model which is accurate enough for estimating final price, that it can be used as a substitute for final price when betting.

[lansdale]

Originally Posted by JerryBoyle

Hey lansdale, my answer wasn't super clear before. I'm confident that I DO have an edge over the final odds, when using the final odds as an input to my model. I'm using the standard 2 step model that Benter advocates in his paper - I generate a fundamental model as step 1 that only considers things relating to the horse, track, trainer, etc. I then merge this into an ensemble model with 2 factors - the fundamental model estimate and the market estimate. The output of this is my final estimate.

The problem is that, as we know, the market estimate varies wildly up until all money is totaled after the race has started. This leads to a modeling problem and a testing problem. First, when I train that ensemble model, do I use the final market estimate or the market estimate as it'd be known when I place my bet, say 30 seconds prior to close of betting? Second, when I actually test how the model would have done via simulating betting, do I use the final price, or do I use the price I would have seen at 30 seconds to close? Let's say final estimate is option A and 30 seconds to close estimate is option B.

A: leads to the strongest model. I know FOR SURE that the ensemble prediction I generate, which combines the fundamental estimate and the final market estimate has an edge over the final market price. As you've said, this is easy to test in any number of ways. However, when I actually go to use the model to bet in real time, I can't combine the final odds with my fundamental estimate, because I don't know the final odds. My options are to either use odds at 30 seconds to post or some estimate of the final odds. Both lead to uncertainty with regards to knowing how the system will perform compared to my backtest. I can test how this set up would perform by using the odds at 30 seconds to post historically, but the data I have doesn't go back very far, so I can't be very confident.


B: a weaker model, but a better representation of how the system would work in real time. I train the ensemble model using the fundamental estimate and the market estimate at 30 seconds to close. Simulating betting using the 30 seconds to close price and this model generally best simulates how the model would bet in real time. It also generally results in over betting, as the final price is almost always more efficient for the bets the model would make.

In the past, I've been using option B. However, I'm starting to think the best route is option A followed by developing a model which, as you've said, estimates final price. I'm not finished researching that yet, but I'm starting to believe that it's possible to build a model which is accurate enough for estimating final price, that it can be used as a substitute for final price when betting.

Okay -- this is a more complex version of the same basic problem that every handicapper has to deal with. Since your terms can get a bit confusing, I'm going to translate them into Benterese, since his article is a common reference --you can just point out where I misunderstand. By 'final market estimate' I'm assuming you mean Benter's 'combined estimate' -- your 'market odds' I understand as Benter's 'public model,' which, given the subject of this post, I'll call 'last-seen odds', and your 'final odds' I'll call, per Benter, 'actual odds.'

So, you're saying your 'combined estimate' has no edge vs. 'last-seen odds' at the 30-second threshold -- the magnitude of the odds-drops plus the number of drops erases any edge. However, had you known the 'actual odds' before betting (handicapping status quo before the previous decade or so) the 'combined estimate' would have produced an edge.

As you say, this would all seem to rule out option B and an esimated edge for option A sounds like the way to go -- I would guess that the whales must be doing the same. One speculative idea, which you might well have already discarded, is to model the 'profitable' vs. 'non-profitable' bets among horses that that the model would have bet at the 30-second threshold -- of course the results might be random, but possibly some discernible patterns that some slight shifts in parameter-weighting could exploit. Regardless, best of luck -- imho, this game has never been more difficult to beat.

[lansdale]

BTW, someone other than myself put the thumbs-down icon on my previous post. Sorry. Your posts are always among the best.

[cj]

Originally Posted by lansdale

BTW, someone other than myself put the thumbs-down icon on my previous post. Sorry. Your posts are always among the best.

Only you could have done that, obviously by accident. You post hasn't been edited by anyone but you.

I now edited it to remove it.

[lansdale]

Originally Posted by cj

Only you could have done that, obviously by accident. You post hasn't been edited by anyone but you.

I now edited it to remove it.

Thanks for cleaning this up, cj. This must have happened as you say -- since I never use emojis or icons, I must have hit it by accident. When I tried to re-post with 'no icon', the time for editing must have elapsed -- therefore the short explanatory post. I think anyone who read the text would have realized it contradicted the icon.

[Lemon Drop Husker]

I feel like I just went through an infomercial.


Do I get an abcruncher for free if I order?

[Lemon Drop Husker]

Originally Posted by Nitro

Hi there Jay!
I’m afraid the premise of what you’re suggesting is of no real value. As others have already suggested; from a practical standpoint even if you knew what the final odds might be you couldn’t take advantage of that information. The Time factor for placing a wager is obviously over. More importantly there’s absolutely no guarantee that any of the entries getting very late action (in terms of changing odds) will win the race.

When you’re actually following the money from when the betting starts with a sophisticated tote analysis (which is monitoring much more than just the Win pool (the odds)), you’re able to get a much clearer picture of which entries are getting played in earnest. These tell-tale signs are provided by understanding specific betting patterns that allows you compare the betting on each entry in real time with the overall betting on all the entries combined.

Unfortunately, these betting patterns are not the same from track to track or even from race to race. However once they become familiar there’s no need to wait for the last minute to decide on which entries are live. By viewing these betting patterns at specific time intervals prior to post time it sometimes becomes very obvious which entries are playable. The tote analysis doesn’t discriminate by the odds. In other words a 10/1 shot could look just as promising as a 2/1 shot. The question then becomes what is the best approach to taking advantage of those entries that show promise. Fortunately, there are many types of bets that can be made.

For those who might have any doubts about what I’ve described and the method that I use, then I would suggest visiting the Selection forum once and a while. I have no problem occasionally sharing some worthwhile information.

Hi Nitro.

Can you post your actual win/loss results of actual wagers for us all to see?

I mean, if they are as good as you say, there should be something other than 5 to 7 horse non-descriptive wagers of who knows what as actual results, amirite?

I can box entire fields and show amazing $2 Super and $2 Trifectas, but I do need to see receipts of said results as proof if you don't mind the trouble.

[Nitro]

Originally Posted by Lemon Drop Husker

Hi Nitro.

Can you post your actual win/loss results of actual wagers for us all to see?

I mean, if they are as good as you say, there should be something other than 5 to 7 horse non-descriptive wagers of who knows what as actual results, amirite?

I can box entire fields and show amazing $2 Super and $2 Trifectas, but I do need to see receipts of said results as proof if you don't mind the trouble.

Hey friend since when does anyone around here post that sort of information? I don’t even let my wife see it and I’m going to make it public? So don’t hold your breath!

Why not just take my word for it and perhaps examine the Hong Kong selections for this morning’s card.
Posts 364 thru 390
http://www.paceadvantage.com/forum/s...154102&page=25
http://www.paceadvantage.com/forum/s...154102&page=26
If you think I stay up all night for my health you’re greatly mistaken.
While you're at it you can also view my O.A. profit margin on the Dutch Win betting for the entire HK meet on the same thread.

I know it may hard for the so-called handicappers to swallow the fact that I don’t handicap at all. But as I’ve mentioned before I have no problem sharing my information in the Selection forum occasionally.

Let’s call an ace an ace. What I don’t see is same thing being offered by many of those who claim to know more about the game than everyone else. You may not agree but I've always believed that Actions speak louder Words.

[FakeNameChanged]

Originally Posted by Lemon Drop Husker

I feel like I just went through an infomercial.


Do I get an abcruncher for free if I order?

Nah, just start wearing bigger shirts.

[CheckMark]

Originally Posted by Lemon Drop Husker

I feel like I just went through an infomercial.


Do I get an abcruncher for free if I order?

Just pay the extra shipping and handling, and we will give you 2!

[Frost king]

Originally Posted by Nitro

Hey friend since when does anyone around here post that sort of information? I don’t even let my wife see it and I’m going to make it public? So don’t hold your breath!

Why not just take my word for it and perhaps examine the Hong Kong selections for this morning’s card.
Posts 364 thru 390
http://www.paceadvantage.com/forum/s...154102&page=25
http://www.paceadvantage.com/forum/s...154102&page=26
If you think I stay up all night for my health you’re greatly mistaken.
While you're at it you can also view my O.A. profit margin on the Dutch Win betting for the entire HK meet on the same thread.

I know it may hard for the so-called handicappers to swallow the fact that I don’t handicap at all. But as I’ve mentioned before I have no problem sharing my information in the Selection forum occasionally.

Let’s call an ace an ace. What I don’t see is same thing being offered by many of those who claim to know more about the game than everyone else. You may not agree but I've always believed that Actions speak louder Words.

Hey Nitro, I know if I won as often as you do, my wife would be by my side, with her hand out, like the Hong Kong product would be my personal ATM machine. Who would have to work? With all the money you are making with your Dutching method, and all the exotics you are hitting, you would be able to make a living doing this. So staying up once a week on a Fri/ Sat, and getting up early on a Wed morning, would be the perfect job. I think, I could handle this twice a week, and have the rest of the week to spend with my wife. If this is not the case, the we know what the real truth is.

[Nitro]

Originally Posted by Frost king

Hey Nitro, I know if I won as often as you do, my wife would be by my side, with her hand out, like the Hong Kong product would be my personal ATM machine. Who would have to work? With all the money you are making with your Dutching method, and all the exotics you are hitting, you would be able to make a living doing this. So staying up once a week on a Fri/ Sat, and getting up early on a Wed morning, would be the perfect job. I think, I could handle this twice a week, and have the rest of the week to spend with my wife. If this is not the case, the we know what the real truth is.

Where do come up with this stuff? Just because I don’t give my wife details of my betting doesn’t mean I don’t share the positive results. She’s very supportive and knows that I take the game seriously and is very considerate when it comes to not presenting me with any distractions. BTW aside from SSI and a retirement account my only other source of income is playing this game particularly at HK twice a week.

My free time is also very important to me. Not only is it related to the days I’m not playing but to the way I play. My method is very efficient and the time it takes for making final selections for any race only requires the length of time for a typical betting cycle. If I had to handicap the way I used to I would have quit this game a long time ago. As far as I’m concerned, the time involved to thoroughly examine the data and prepare for successful betting is just too tedious and not worth the effort.

[AltonKelsey]

Interesting that Nitro posts such gory detail before and after the race, but never an actual bet structure or ticket. How hard is that ? He seems to be good at cut & paste.



Maybe we know why