what are your Top 3 books?

[DeltaLover]

Good question, I will give a pretty simple example from horse racing that with small changes can be applied to more sophisticated cases.

I will not use exact numbers though, since at this moment I do not have access to my database, but the purpose of this example is just to demonstrate so we can go with approximations.

Let's specify a handicapping factor that is widely used by the public: A horse whose last three performances all look very bad. We 'very bad' as a race where the horse finished worse than 5th by at least 10 lengths and at no fractional call he was better than fourth.

This is the main profile of the horse we will do our research about.

We go to our data base and find out that we have 367 such horses out of a total 1,826 starters that exist in our sample of 260 races. Out of these 367 horses only 11 managed to win while the rest 356 lost.

We can now assume that the probability of such a horse to win is 11 / 367 ~ 0.03 = 3%

Note that this is not necessary the best approximation since the number of starters in the race and the matching this profile horses are not distributed equally, but to simplify our example lets assume that this is not the case.


At this point we need another condition that we will use for our probability estimate. It can be any imaginable handicapping factor. For simplicity I am picking as an additional factor one that looks like this:

The race's favorite is coming from a long layoff, drops in claiming price, is of E/P type and the race has two other E type horses.

Let' assume that out of our original sample of 260 races we have 60 of them matching this profile and out of these 60 races we had 6 winners matching our original 'bad' profile while we had 81 total matches of it.

Based in this description is easy to see that the win probability of a match to the bad profile in this type of race is now:

6 / 81 ~ 0.074 = 7.4 %

which compared to the original 3% is more than double so it serves as a good indicator to farther investigate this race scenario.

Of course this is a simplified explanation that I put together just to demonstrate the concept rather to serve as a methodology.

[JJMartin]

OK, so it seems the premise of the formulation is based on arbitrarily inputting factors to compute (the conditions for estimate), observing the computational result and using that percentage (in your example 7.4%) as an opinion of the chances of success. Now the obvious question that comes to mind is: How do you know what combination of factors will give you the highest % probability of a horse winning without computing an enormous amount of conditions (scenarios) based on countless combinations of variables? What if I chose for example, the M/L fav or the track best jockey etc. as a condition? You would have to subjectively ascribe varying degrees of value to factors based on your own way of thinking and then through trial and error, arrive at the most promising predictive conclusion, I would imagine?

[DeltaLover]

Originally Posted by JJMartin

How do you know what combination of factors will give you the highest % probability of a horse winning without computing an enormous amount of conditions (scenarios) based on countless combinations of variables?

This is exactly what we are looking for. This process although complex is not so complicated as it sounds.
In essence what we need is to understand what are the handicapping factor mostly used by the crowd and then to cluster them based in not so obvious factors or those that tend to be overlooked.

Discovering these factors does not have to be so laborious as a trial and error process implies, there exist several algorithmic paths we can follow to automated it. By this I do not want to say that following a more manual approach will be completely fruitless, this depends on our betting competitors; in other words our betting results are dependent to the sophistication of our co-bettors.

Another way to visualize this approach is to view the universe of handicapping factors as a fractal having infinite components. Following this abstraction we can view all possible handicapping strategies as a set of fractals and our job as handicappers is to find the one that presents the best value. Of course we can take this approach to an even higher level where the unknown will not be the fractal anymore (meaning the concrete strategy as a composite of simpler decisions) but the mechanism that actually creates this strategy; in other words a meta - handicapping level where we are handicapping the handicapping process itself.

Please note how relative is the topic of the conversation with the recent thread 'Handicapping with Similarities and dissimilarities'.

[traynor]

No distaste for pace or speed analysis at all. I use both, but not in the simplistic ways advocated by some. That is, I use both pace analysis and speed analysis, but not in the same forms used by anyone else (to the best of my knowledge).

Match-Up is a little more complex. I could say I use it, but that would imply I follow Jim Bradshaw's interpretation, which would be incorrect. I use a set of pace comparison algorithms that do much the same thing.

Originally Posted by lansdale

Hi Traynor,

I assume this is a rhetorical question, since you've already made clear your distaste for pace handicapping, at least as a stand-alone technique. However, I think it's worth separating the Match-Up, which is subjective enough for eight different people to come up with different Match-Up horses, from, say Quirin's Speed Points, or Randy Giles Pace Pressure, which have been objectively measured to demonstrable effect. Since you are a database guy, I assume you might have tested the latter two, and have some of idea of their current value. Purely as conjecture, I would guess that Speed Points, after decades in existence, and the spread of computer handicapping has little or no wager value. And I'm willing to take Giles' comments about Pressure Point overlays occurring in about 1 race out of 7 to be relatively accurate, but I would be interested in seeing it tested.

Thus, I wouldn't discount the existence of the influence of race dynamics in a limited number of races, but I believe that influence to be exaggerated because it fits into the conventional notion of the race as a contest rather than primarily a demonstration of skill and ability, which I believe database players have demonstrated it is. As DL pointed out Benter ignored it, and another successful database player who posts here, sjk, has described the effect of race dynamics as real, but minimal, in the overall context of racing. I would be very interested, though, in hearing from any database players, with compelling evidence either way.

Cheers,

lansdale

[traynor]

Originally Posted by DeltaLover

... If you happened to know that your player is very tight plus he knows you are tight as well what do you think is the probability of him holding AA?
...

For example if we thing this player will play like this only when he holds KK or AA then the odds are clearly against us as he can have AA using 6 combos and KK with only 1 (since we already have the other two Kings). So the odds are 1 - 6 that he has AA. The condition we used for this calculation (in our example that he can only have KK or AA) is what we need to be aware of.

Assuming, of course, that you have not been led to believe things about what the player will or will not do under specific circumstances that have set you up for a thorough drubbing.

"Mind reading" is as deceptive in poker as it is in combat (or business, or diplomacy, or negotiation). In activities in which lies, deception, and trickery are the tools of the trade, the belief in one's ability to "read minds" can be costly. Several poker-playing friends did well with Mike Caro's "tells" by mimicking what others expected/wanted to see. In NLP, its called "pacing and leading."

The same thing is true in horse racing, especially in the study of "trainer strategies."

[traynor]

Originally Posted by JJMartin

OK, so it seems the premise of the formulation is based on arbitrarily inputting factors to compute (the conditions for estimate), observing the computational result and using that percentage (in your example 7.4%) as an opinion of the chances of success. Now the obvious question that comes to mind is: How do you know what combination of factors will give you the highest % probability of a horse winning without computing an enormous amount of conditions (scenarios) based on countless combinations of variables? What if I chose for example, the M/L fav or the track best jockey etc. as a condition? You would have to subjectively ascribe varying degrees of value to factors based on your own way of thinking and then through trial and error, arrive at the most promising predictive conclusion, I would imagine?

And don't forget that each combination of factors creates a new, discrete factor that must be considered as such. In your example, M/L favorite and best jockey are separate factors. So is M/L favorite WITH best jockey, and M/L favorite NOT WITH best jockey. It is something that many overlook.

[DeltaLover]

Originally Posted by traynor

Assuming, of course, that you have not been led to believe things about what the player will or will not do under specific circumstances that have set you up for a thorough drubbing.

"Mind reading" is as deceptive in poker as it is in combat (or business, or diplomacy, or negotiation). In activities in which lies, deception, and trickery are the tools of the trade, the belief in one's ability to "read minds" can be costly. Several poker-playing friends did well with Mike Caro's "tells" by mimicking what others expected/wanted to see. In NLP, its called "pacing and leading."

The same thing is true in horse racing, especially in the study of "trainer strategies."

That's why the higher levels of poker are about game theory and beysian probabilities. What I find interesting is that short handed games especially hands on seem to be close to be fully analyzed and solved. I remember some time when I found a heads up holdem robot called Sonia which was a very hard opponent.

[DeltaLover]

Originally Posted by traynor

And don't forget that each combination of factors creates a new, discrete factor that must be considered as such. In your example, M/L favorite and best jockey are separate factors. So is M/L favorite WITH best jockey, and M/L favorite NOT WITH best jockey. It is something that many overlook.

Absolutely..

[Capper Al]

Did anyone mention 'How to Pick up Girls?'.

[DeltaLover]

Originally Posted by Capper Al

Did anyone mention 'How to Pick up Girls?'.

Al,
this task is the goodlooking guys , not for race addicts

[thaskalos]

Originally Posted by DeltaLover

Bayesian probabilities express a probability based in a condition.

Let me use an analogy from poker to better explain it.

You are playing heads up with and you are in the beginning of a hand.

You look down you see a pair of Kings.

What is the probability that your opponent is holding a pair of aces?

Easy question. Since there are 50 unseen cards (excluding your Cowboys) there are 1,275 combinations of two and only 6 of them consist of two aces.

So the probability that your opponent is holding AA is going to be: 6 / 1,275 or 0.47%. In other words less than the half of 1%

Of course you go ahead and you raise, your opponent 3bets and you cap. The flop is all rugs and the same betting scenario repeats.

At this point is time to start using your Bayesian calculations...

If you happened to know that your player is very tight plus he knows that you are tight as well what do you think is the probability of him holding AA?

Do you think that after the furious second betting round the probabilities stay the same as they were before the action was initiated?

I am sure you understand that no, the probability of our opponent holding AA is now much larger than it was in the beginning.

The thought process that leads to this conclusion is a Bayesian approach.

For example if we thing this player will play like this only when he holds KK or AA then the odds are clearly against us as he can have AA using 6 combos and KK with only 1 (since we already have the other two Kings). So the odds are 1 - 6 that he has AA. The condition we used for this calculation (in our example that he can only have KK or AA) is what we need to be aware of.

You can read more here:

http://en.wikipedia.org/wiki/Bayes'_theorem

And of course searching PA you will find a lot of excellent posts (most of them by Trifecta Mike) for this topic.

There is no such thing as tight play when you are playing heads-up. The tight players belong in the full games...

[traynor]

Originally Posted by DeltaLover

That's why the higher levels of poker are about game theory and beysian probabilities. What I find interesting is that short handed games especially hands on seem to be close to be fully analyzed and solved. I remember some time when I found a heads up holdem robot called Sonia which was a very hard opponent.

I think I may have played her once. Short blonde from the Ukraine?

[Mike A]

Originally Posted by thaskalos

There is no such thing as tight play when you are playing heads-up. The tight players belong in the full games...

Thask, I really respect you as a poster here, and as a fellow strong player in both racing and poker. Maybe I'm missing some context/ ribbing among friends etc, but as you know tightness/looseness in poker occurs along a continuum, and that includes hu play, even huhu, and both limit and NLHE. And the relativity of tightness, and just where along that continuum to the best of your knowledge you can put villain, do coherently fit into what Delta is explaining, even though he's using the extreme as an example.

Don't get me wrong, I'd guess you're aware of this, but I just wanted to point it out just in case, and since I'm really curious about what Delta is getting into, since it's an area I have a lot to learn in.

[thaskalos]

Originally Posted by Mike A

Thask, I really respect you as a poster here, and as a fellow strong player in both racing and poker. Maybe I'm missing some context/ ribbing among friends etc, but as you know tightness/looseness in poker occurs along a continuum, and that includes hu play, even huhu, and both limit and NLHE. And the relativity of tightness, and just where along that continuum to the best of your knowledge you can put villain, do coherently fit into what Delta is explaining, even though he's using the extreme as an example.

Don't get me wrong, I'd guess you're aware of this, but I just wanted to point it out just in case, and since I'm really curious about what Delta is getting into, since it's an area I have a lot to learn in.

I appreciate the kind words...and in no way am I trying to dissuade you from further pursuing what DeltaLover is getting into here.

DeltaLover is a friend of mine...and my comment was made strictly in jest.

[Mike A]

Originally Posted by thaskalos

I appreciate the kind words...and in no way am I trying to dissuade you from further pursuing what DeltaLover is getting into here.

DeltaLover is a friend of mine...and my comment was made strictly in jest.

I had a feeling. And I couldn't help but notice in some posts that you guys do seem to be friends.

BTW, thaskalos (and I have to head out to town in a few minutes, so I'll be quick here), I want to thank you for being someone who has really been an inspiration to me here, despite the fact I haven't posted much so far, or told you that before.

(And sorry guys, to take this off-topic!) I remember reading about how you lost your wife, and you are raising a young son on your own, and it really touched me...all the more, since I don't have a family of my own yet, but it's all I really want, and I guess I see you as someone with a lot of the same qualities I seem to have..i.e., I guess I seem to be "built" to "gamble" for a living. And my point there is that, we both know what happened in the online poker world on "Black Friday" (and I actually lost my "career", playing professionally at Pokerstars), and now, we both seem to agree about the present precarious state of horse racing in this country.

One thing I'm getting at is, for people like us..who knows how many options we'll have here, and for how long?? SO-- what I'm going to throw at you is (and I have to say I'm actually pretty excited in a way)..well, do you know what our "beloved" (hehheh) Dick Schmidt has been making a living at for the last few years? --Forex trading. I had a PM exchange with Dick a few weeks ago re this. OK...more later, but let me tell you, I'm basically a newb when it comes to forex, but I've been doing a ton of research, and (among other stuff like *yes, there are absolutely plenty of scams out there, especially with guys looking to sell beginners anything ---all the info you need IS free out there*) The bottom line is, forex does NOT look at all like anything even approaching the difficulty of going from newb to pro in racing, nor anything like going from newb to 20 tabling midstakes pro at Stars!!!

I would really love to hear your thoughts, thaskalos!