Pattern Recognition

[traynor]

Originally Posted by InControlX

Wow. I would think that if one wishes to exploit handicapping using pattern recognition it might not be wise to dismiss the work of those who are successful in handicapping. Also, realize that some handicapping is forced to be based upon small samplings, and wide-search attempts completely miss infrequent but repeatable outcomes. For example, an annual throroughbred race count in the US and Canada is around 85,000. I know of quite a few successful handicappers who have recognized patterns involving combinations of certain age horses at certain tracks on certain months against certain differing entries. This narrows your annual race count down to about 50. Statisticians will scoff at such a small sample, but these guys have been profiting from this pattern for years!

As for data base pattern mining on multi-year collections, the common outcome is that the better win percentage and ROI patterns themselves are small sample counts, and unless you have some very good proving tests, not very repeatable. For large sample patterns (in the 2000+ annual range) the outcomes approach the morning lines.

Probably the biggest fallacy of pattern recognition in horse racing is the lack of competitive evaluation. In simple terms, your pick has the background pattern you're looking for, but what about the unknown patterns of the other entries?

ICX

I think you may be confusing pattern recognition with spot plays. There is a lot of overlap, but they are not exactly the same thing.

[traynor]

Originally Posted by InControlX

Wow. I would think that if one wishes to exploit handicapping using pattern recognition it might not be wise to dismiss the work of those who are successful in handicapping. Also, realize that some handicapping is forced to be based upon small samplings, and wide-search attempts completely miss infrequent but repeatable outcomes. For example, an annual throroughbred race count in the US and Canada is around 85,000. I know of quite a few successful handicappers who have recognized patterns involving combinations of certain age horses at certain tracks on certain months against certain differing entries. This narrows your annual race count down to about 50. Statisticians will scoff at such a small sample, but these guys have been profiting from this pattern for years!

As for data base pattern mining on multi-year collections, the common outcome is that the better win percentage and ROI patterns themselves are small sample counts, and unless you have some very good proving tests, not very repeatable. For large sample patterns (in the 2000+ annual range) the outcomes approach the morning lines.

Probably the biggest fallacy of pattern recognition in horse racing is the lack of competitive evaluation. In simple terms, your pick has the background pattern you're looking for, but what about the unknown patterns of the other entries?

ICX

I refer you back to Schlesinger's comment that Anderson's (very successful) three years of full-time play were statistically insignificant.

The reason many handicappers prefer short models is because the aberrations that seem to indicate positive results quickly disappear in larger samples. Some get lucky and win a few bucks betting on those short models. Some are not so lucky, and lose a bundle chasing what they believe to be "winning patterns."

Again, I refer you to Schlesinger. Small models are mostly irrelevant, regardless of whether one wins or loses by using them. Think of it like those pretty lights indicating the result of the last 15-20 spins of the roulette wheel in casinos. Joe Schmoe sees a bunch of red, and bets his rent money on black. He wins. Lucky Joe Schmoe. Joe Schmoe's win is irrelevant, just as his loss would be irrelevant (to everyone except Joe Schmoe).

[traynor]

Originally Posted by Capper Al

Ok, I want to know how Quirin got involved with racimg. Would someone post it or PM me if you rather not give out the negative info?

You know, I have had a really nice life, many great experiences, many great memories, lots of fun, excitement, interesting things to do. And all that while, I had no idea that William Quirin was a bad guy. So I ask myself, "Hey, self, do you think our life will be better if we know a little dirt about WQ?"

And my self--as it is given to do occasionally--says to me, "Nah. Pay attention to that seventh race tomorrow, not something that happened long ago and far away to people you don't know, don't care about, and that have no relevance whatsoever to your life. Ad hominem is bad for the digestion."

[traynor]

Originally Posted by Gamblor

Weight doesn't mean nothing. It has a mass, and gravity doesn't cease to exist once the starting gates open. Weight matters. It is ludicrous of every author who suggests it does not.

HOW MUCH it matters, is the important question, not whether it matters. Traditionally, here in Aus as well as elsewhere, it's effect has been over-emphasised. Certainly these days, as horses are doubtless stronger than they once were, weight matters less than it once did. Horses are able to withstand extra weight better than they could, by being stronger than they once were.

So, the impact of weight has lessened. 1kg extra weight on a horse's back has less impact now than it did 100 years ago.

My own research leads me to believe 1kg of extra weight = roughly 0.35 lengths slowing over the course of a race, or about 90 cm, or 3 feet in the outdated system.

That sounds like some seriously detailed research. A question, if you don't mind? How do you handle adjustments for different distances and different tracks in Australia?

[Maximillion]

Originally Posted by traynor

Actually, it is way easier than that. Give the problem to a grad student in MIS or Statistics, and the first thing they will ask is, "Where is the target?" The "target" is the winning of the race. Once the target is known, the next step is to ask, "What indications were available before the event that point to this outcome?" Absolutely no knowledge of horse racing is needed, nor are any preconceived ideas of what should be--only a critical analysis of what is.

The trainable skill is to ask the right questions. Ask the wrong questions, and regardless of how accurately they may represent the available data, they will be less useful that if the right questions are asked.

Maybe another "question" you can ask is..what "target" is consistently outrunning its odds?
Are their any patterns they share?

Maybe these "questions" are more suited for exotic play..Im mainly a win bettor but considering changing my overall approach.

[bisket]

Originally Posted by traynor

Actually, it is way easier than that. Give the problem to a grad student in MIS or Statistics, and the first thing they will ask is, "Where is the target?" The "target" is the winning of the race. Once the target is known, the next step is to ask, "What indications were available before the event that point to this outcome?" Absolutely no knowledge of horse racing is needed, nor are any preconceived ideas of what should be--only a critical analysis of what is.

The trainable skill is to ask the right questions. Ask the wrong questions, and regardless of how accurately they may represent the available data, they will be less useful that if the right questions are asked.

this is absolutely wrong.... if you don't have an understanding of the game. you're wizzing in the wind if you think you win at this game. to win at this game you not only have to have an understanding of what you're computations are revealing. you also need to have an understanding of what you're fellow bettors think is important. yes you can get to a point where you will profit by doing the same things at a higher level than your fellow bettors, but to win, you need to know how the rest of the pool views the race. the best plays are the ones that make no sense to your competition...

[bisket]

Originally Posted by Capper Al

Being able to query a database isn't the be all of handicapping development. One has to experiment with the game and the nature of horses, not the numbers. The numbers will come once the ideas are right. The numbers by themselve won't advance one's handficapping. The wishful thinking is that all there is to handicapping can be accomplished from the database that I know how to use. The trainable skill that is needed is to understand that there are patterns everywhere in our data, and most of them are false positives. And the few that are not false postives need to be understood and place within their proper context to use.

the best advice i've seen in cyber space!!!!!!

[Gamblor]

Originally Posted by traynor

That sounds like some seriously detailed research. A question, if you don't mind? How do you handle adjustments for different distances and different tracks in Australia?

I am quite different to the current mainstream thinking, in that I cannot find any evidence to suggest that weight has a bigger impact the further you go. All my research indicates it's impact is roughly the same at all trips. So, I treat it the same.

[traynor]

Originally Posted by Maximillion

Maybe another "question" you can ask is..what "target" is consistently outrunning its odds?
Are their any patterns they share?

Maybe these "questions" are more suited for exotic play..Im mainly a win bettor but considering changing my overall approach.

Exactly so. Sometimes (especially if you concentrate on exactas, as I do) it is more important to ask, "Which horse is going to place?" The target is a starting place--an end result from which to work backwards to determine the pieces of information (available before the event) that could have been used to predict that specific outcome.

[traynor]

Originally Posted by bisket

this is absolutely wrong.... if you don't have an understanding of the game. you're wizzing in the wind if you think you win at this game. to win at this game you not only have to have an understanding of what you're computations are revealing. you also need to have an understanding of what you're fellow bettors think is important. yes you can get to a point where you will profit by doing the same things at a higher level than your fellow bettors, but to win, you need to know how the rest of the pool views the race. the best plays are the ones that make no sense to your competition...

I didn't say the grad student could instantly become Andy Beyer. If you will look carefully at what I wrote, I suggested what is lacking is a clearly defined target. Once that target is defined, and the parameters of available information are defined, no knowledge of horse racing whatsoever is necessary to do the research. In fact, the less the researcher "knows" about the topic, the fewer misconceptions, front-loading, and personal biases exist to corrupt the research.

[traynor]

Originally Posted by Gamblor

I am quite different to the current mainstream thinking, in that I cannot find any evidence to suggest that weight has a bigger impact the further you go. All my research indicates it's impact is roughly the same at all trips. So, I treat it the same.

I should have made it clearer what I was asking. Apologies. You refer to an adjustment in distance that seems intended to apply to a pace/speed evaluation. My question was in regard to comparisons of performances at different distances and different tracks, rather than specifically to weight.

I did a lot of research on Australian races a couple of years ago and one of the biggest difficulties I encountered was the mind-boggling assortment of distances and tracks. Many attempts at "performance equalization" in the US and Canada are based on the theory of par times. The time(s) of a $10,000 claimer (or whatever) at six furlongs at Track A are compared to the time(s) of a $10,000 claimer (or whatever) at six furlongs at Track B, and the difference used to "adjust" the performance at one track to an "equivalent value" performance at the other track. Nice in theory, not so good in reality.

Is there a better way in Australia?

[Capper Al]

Originally Posted by traynor

I didn't say the grad student could instantly become Andy Beyer. If you will look carefully at what I wrote, I suggested what is lacking is a clearly defined target. Once that target is defined, and the parameters of available information are defined, no knowledge of horse racing whatsoever is necessary to do the research. In fact, the less the researcher "knows" about the topic, the fewer misconceptions, front-loading, and personal biases exist to corrupt the research.

The highlighted may be true to begin. Then the first thing an unexperienced researcher must do is to learn the game and the nature of horses. So we are back at the beginning with this conversation. One needs the fundamentals. The paper and pencil handicapper is still to this day the best pioneer model we have for speculating.

[Capper Al]

There has been only one pattern in this thread. We're going nowhere with this discussion. That's the only pattern recognizable.

[thaskalos]

Originally Posted by Capper Al

There has been only one pattern in this thread. We're going nowhere with this discussion. That's the only pattern recognizable.

I agree!

The reason discussions of this type often go nowhere is because we don't go into any detail about what we really mean. We start threads, we hint about different topics, and we offer vague generalities about things that we think we know about...but we seldom give any concrete examples, or offer any reasonable explanations which might allow those we are communicating with to really understand what we are talking about.

Why don't we put our cards on the table and start saying what we really mean to say...so we can get some sort of constructive dialog going?

I'll be more than happy to start things off.

I don't believe in pattern recognition as it applies to horse race handicapping...and I don't consider patterns like "workouts within a race"...or "maintaining the same position throughout the race"...to have any merit at all when assessing a horse's chances against the field he is asked to face today.

I don't have all the answers when it comes to handicapping and betting, and I am readily willing to admit that I am wrong about practically anything that I think I might "know" about this game...but I do have several years of professional-level play to back me up, and it takes more than a few veiled handicapping opinions to get me to change my mind about things.

My own opinion when it comes to handicapping is this:

Every race is unique, and it's made up of components which are not likely to ever be repeated again. The circumstances that a given horse is asked to face today are UNIQUE...and they do not fit any pre-conceived "patterns" that we think we might have identified...based on any number of races that we happen to have analyzed in the past.

All wire-to-wire efforts are different...as are all "even throughout" efforts. They have occurred under different pace scenarios...and under different track and weather conditions. They have taken place during different stages of the horses' form...and have been governed by differing race "dynamics".

To think that these efforts can somehow be "categorized" into readily identifiable "patterns" is wishful thinking, as far as I am concerned...and I am dying to see someone use such a method to handicap a race in advance.

This is an extremely difficult game to beat...and there is a reason for that. It does not lend itself to be easily "categorized".

I welcome different opinions, of course...and I relish the opportunity to compare my own handicapping methods to any "pattern" handicapping methods out there.

[gm10]

Originally Posted by Gamblor

I am quite different to the current mainstream thinking, in that I cannot find any evidence to suggest that weight has a bigger impact the further you go. All my research indicates it's impact is roughly the same at all trips. So, I treat it the same.

I would say that this is not quite right ... especially outside the US where there is a bigger range of race distances.

The traditional formula is:

15 = distance in furlongs * (pounds/length).

So for example ... 1 length equals 2.5 pounds over 6F.

My own research has shown that, except when the going is truly slow, this formula exaggerates the effect of weight. The real effect is only about half as big.

In my opinion, this is because tracks are in much better shape these days than they were in the old days.