Mark Cramer: Art vs. Science....(line making)

[Overlay]

Originally Posted by LottaKash

But, MC's book "THE ODDS ON YOUR SIDE" was the deal clincher for me....It had filled the most important void in my personal handicapping process..."THE BETTING LINE PHILOSOPHY"...

Was Cramer's later book, Value Handicapping, essentially an expansion of The Odds on Your Side, or was it a significantly different approach in some way?

[Fastracehorse]

Originally Posted by markgoldie

Well Kash...

Nice post and a particular area in which I have mounted the soapbox in the past.

My answer is yes- all these feelings, shadings, nuances, leanings, suspicions, trepidations, confidences, if-this-then-that dependencies, etc. can be quantified. They can also be programmed into a computer which will then faithfully and quickly reproduce them without the minute-to-minute normal oversights and forgetfulness of the human brain.

Since I'm not a computer guy, you may wonder how in the world I can be so confident this is true. The reason is that the deterministic measurements of the mind are just that- measurements. And all measurements are capturable by numbers.

As I wrote on a simlar thread, the mind craves these numbers when it comes to any sort of quantitative measurement. So, for example, when two people are discussing even a non-standard, highly judgemental, feeling-based measurement, the need for numbers becomes evident. To wit:

Person 1: How angry was she?
Person 2: She was pretty angry. Sort of angry anyway. Not furious. But kinda angry nonetheless.
Person1: Help me out here. On a scale of 1 to 10, how angry was she?

The overwhelming dependency on numbers by handicappers is a simple result of the mind's need for accurate quantification. In fact, when you read these numbers day in and day out for so many years, you begin to understand in a strange way the verbal stories behind them. Like: "Once upon a time there was a three year old filly. She showed some promise as a two year old, but the stress of racing got to her and she needed to be put to bed. When she came back for her sophmore campaign, she initially didn't know how she felt about the racing game in general. She was tentative. Hesitent. And then one day her trainer thought: why not try her with blinkers? Maybe she just gets a little distracted by all the stuff going on around her. And so they did. And the trainer was right. Our little girl put her mind on business and she raced much better. She won two of three races and then her trainer and owner decided to try her against some better-class fillies in a graded stake. This would be the test of her life. She had the ability, but could she dig down and find her competitive spirit against proven rivals?

Stories. And yet, numbers are the only way we have to begin to predict how these intersecting stories will play out in today's race.

Power numbers, as those produced by "black-box" programs are popular because they integrate a conglomeration of numbers into a single digit. Personally, I am fascinated by them, because they attempt to accurately quantify the numerlogical 'soup." I have looked at the most famous of these- the Brisnet Prime Power number- for many years. I know this number like I would a handicapping buddy that I had spent years with going to the races and comparing handicapping notes. I have a deep respect for my buddy and I generally know how he thinks. So we agree most of the time. But the fascinating times are when we are in total disagreement. I sit there and stare at his out-of-character opinion. I go over the target horse with a fine-toothed comb. What am I missing? What does he see that I don't? What does he know that I don't? Can I bet against him here? Should I? Sure I will. But still, I'm always a bit scared because I know that unlike a human handicapping buddy, this one doesn't have brain cramps... doesn't have bad days when the wife is nagging him and his judgement is impaired. No. This handicapping buddy is like the Rock of Gibralter. And if I'm going to beat him, I know it will test the depths of my experience and knowledge of the game. How do I fare in these situations? I usually win. But not always. And when he wins, I tell myself he just got lucky this time.

At any rate, I would love to be able to discuss his handicapping methodologies with him. But I can't. There's one thing I'm totally sure of though. If he can be this good, there is a possibility that a different black-box buddy could easily beat both of us day in and day out. And so, the numbers can and do tell the stories better than humans can. Clearly there is an art into getting the numbers right- of translating all the shadings into the right number. But that's a one-time struggle with possibly periodic updates. On the other hand, once this struggle is complete, it will handicap all North American races on a given day in a matter of seconds. Show me the human artist who can do that.

..........but you don't say how. You go on to a narrative of racing subtleties that can lead to winners but you don't say how to numerically quantify it.

I disagree with you; you cannot program a computer to deal with the myriad of handicapping subtleties.

Yah, I numerically quantify a horse based on prior or projected ability - it's an important part of the game.

And then I make my selections based on a host of factors: like pace scenario and form evaluation.

It takes time to do a race. It takes time to do a card. I write my selections in order; betting on a few selections with conviction.

I know if the public will like my horse or not - when do I have the time to make a M/L? Why would I want 2? Why would I want a computer to do it? Why would I try and convert a feeling into a number via a PC?

fffastt

[markgoldie]

Originally Posted by Fastracehorse@DRF

..........but you don't say how. You go on to a narrative of racing subtleties that can lead to winners but you don't say how to numerically quantify it.

I disagree with you; you cannot program a computer to deal with the myriad of handicapping subtleties.

Yah, I numerically quantify a horse based on prior or projected ability - it's an important part of the game.

And then I make my selections based on a host of factors: like pace scenario and form evaluation.

It takes time to do a race. It takes time to do a card. I write my selections in order; betting on a few selections with conviction.

I know if the public will like my horse or not - when do I have the time to make a M/L? Why would I want 2? Why would I want a computer to do it? Why would I try and convert a feeling into a number via a PC?

fffastt

Maybe my post wasn't quite clear enough. Okay. Let me expand a bit.

You say you make your determination based on a host of different factors. Fine. Presumably you know what these factors are and they are not simply absorbed into your consciousness by an unknown cosmic force of some sort.

So you engage in the process of handicapping, which (correct me if I'm wrong) is sort of a non-verbal assimilation of different facts regarding the past performances of the horses in question. During this process, plusses and minuses strike you about each horse. As you continue the process, the race as a whole begins to take shape; that is, the interactions of these horses with their attributes and shortcomings, fall into a picture in your mind as to how the race will unfold. As this process becomes more clear, you consult the odds' board to see if and how your handicapping opinion can be exploited through some sort of overlay- a flaw in the crowd's thinking that offers you a porthole to profit.

This, you (rightfully) feel is art. Reliant on science, of course, but pure art nonetheless. How in the world could anyone in their right mind think that this elegant intellectual process of the human mind might be broken down into numbers, such that a cold, insensitive machine could possibly reproduce what's going on here?

Well? I do... and here's why. If you slow down your handicapping process, I mean way down, you will see something interesting. At every juncture, with each variable you consider, you are making a decision; at the least, you are forming an opinion. And this opinion is based upon the knowledge (garnered from years of experience I'd guess) that resides in your memory. Furthermore, the decision is based on the opinion that some attribute or another is better than another attribute and importantly by some discernable margin. But if you think about it, what you are doing is the following: Knowing that factor A is good (+), factor B is bad (-), etc. and that the degree of these factors 0-10 or 0-100 or 0-1000 (it doesn't matter the scale of range) is relevant and important.

The essential point is that the mind is measuring and measurements are a function of numbers at their essence. So these non-verbal thoughts which occur during the handicapping process can be accurately described as numbers. In fact they can be more accurately described as numbers than as either verbal or non-verbal thought. The struggle of understanding verbal measurements was something I addressed in my earilier post. How much is "a little bit; a lot; a great deal; a smidgen; a pinch; a tad; a medium-lot; a whole-lot"?

The fact is, if you dissected your handicapping and assigned numerical values to all the variables you consider, you would become a much better handicapper in the long run. Why? Because you would create a record of the measurements in which you are engaged and a subsequent tweaking might reveal areas where improvement can be achieved.

Measurement-based decisions occur when the conglomeration of measurements reach some tipping point. Since numbers can accurately describe the measurements, they can also accurately describe the tipping points which create decisions.

And so, all of this can be done. I didn't say it was easy. In fact, I characterized it as a "struggle." In this struggle, you will need to deconstruct your process, step by step. But even if you do this not with the intention of converting your handicapping into computer software, you should get a nice benefit from the exercise. As far as making a betting line, I agree with you that this is unnecessary. However, I think it's hard to dispute the counter-argument that any measurement of value and what may constitue a trigger-point for a wager has merit. At the least, it gives you a historic record which you may then assess as you continue the quest for profits in a very tough game.

[CincyHorseplayer]

That is not what Cramer's point is at all.Why is there this urge to eliminate thinking?"If I can only quantify it and put it in a program".It isn't about that it can't be done,but why do it?Making an oddsline is about forcing thought into a why bet this horse??Why am I making this horse 7/2.It eliminates objectivity and that is the point.

[mountainman]

Of course computers can assign reasonably- accurate probabilities based on much the same data that an intuitive handicapper sorts through. But the nuances of racing are too subtle for a machine to interpret as efficiently as the best human mind. In my opinion, the real contribution of computer-oriented handicappers is the objective research they sometimes bring to the table. On this site, for instance, when the geeks put out stats that break handicapping into small pieces and examine relevant factors, i'm ALL eyes. And when the research is especially well-conceived, perhaps outside-the-box, or straight up challenges traditional belief, pace advantage really comes to life. But when discourse becomes so esoteric as to espouse or disseminate an all- numbers approach , or attempts to reduce handicapping to a single rating, I tune it out. Like many members here, I've handicapped, wagered on, and watched 10's of thousands of races, and my computer is between my ears. I'm here looking for weapons I can take into battle, not some number to fight those battles for me.

[markgoldie]

Originally Posted by mountainman

Of course computers can assign reasonably- accurate probabilities based on much the same data that an intuitive handicapper sorts through. But the nuances of racing are too subtle for a machine to interpret as efficiently as the best human mind. In my opinion, the real contribution of computer-oriented handicappers is the objective research they sometimes bring to the table. On this site, for instance, when the geeks put out stats that break handicapping into small pieces and examine relevant factors, i'm ALL eyes. And when the research is especially well-conceived, perhaps outside-the-box, or straight up challenges traditional belief, pace advantage really comes to life. But when discourse becomes so esoteric as to espouse or disseminate an all- numbers approach , or attempts to reduce handicapping to a single rating, I tune it out. Like many members here, I've handicapped, wagered on, and watched 10's of thousands of races, and my computer is between my ears. I'm here looking for weapons I can take into battle, not some number to fight those battles for me.

Don't get me wrong. I'm not telling anyone what they should do or how they should do it. All I'm saying is that whatever you do is capturable by numbers logic.

Probably the best example in recent history was the development of computer chess. I have mentioned this before. The great world champion Garry Kasparov said emphatically that he believed that a computer would never be able to beat the best human player. Now. With due respect to the great intelligent people we have on the forum, I would venture that Kasparov is considerable smarter than any of us. And that this could be verified by any number of IQ tests.

Kasparov based his statement on the fact that at it's ultimate highest levels, chess is more art than science. The level of possibilities in a chess game quickly enter a realm of infinity that is far beyond anything that can be assigned to horse race handicapping. And so he played the machine in a series of matches and games. Ultimately, he lost. This was a bigger blow to a verifiable genius than any insult could possibly affect a comparative moron like myself. Kasparov was forced to admit that he was wrong.

But why? Why and how could he be wrong? How could a machine, programmed by men who could not begin to play with him over the board, beat him?

It turned out that everything this genius of science and art combined did over the chess board was reduceable to numbers. Every decision he made was based on an assessment probabilities whose calculations reached a tipping point. These assessments were informed by the data base that he held in his mind from a lifetime devoted to the game. But the computer held a vast data base as well. And the computer's data base had been dissected though the use of numbers.

In a pivotal game, Kasparov offerred a pawn sacrifice. The expert commentators (all chess grandmasters) allowed how no human player in the world would accept this sacrifice from Kasparov, knowing that he would follow it up with an attack that would be unstoppable. The computer looked at it and took the pawn. Kasparov reacted by throwing up his hands in disgust. By now he knew that there was no forced attack that he could mount in the ensuing game that would give him a checkmate. And so he would have to play the computer a pawn down. He lost.

In a post-match news conference, Kasparov said that when you play another human, you can take advantage of mistakes. But the computer never makes a mistake.

How about us when we handicap? Do we make mistakes? Do we overlook things? Are there things in our experience which we do not relentlessly apply? I think we all know the answer to that.

And so we do what we do. And yet I have a small problem with those who say the art of handicapping will never be captured by computers and their cold, inhuman number-crunching. I'm not sure, but I think Kasparov would agree with me.

[mountainman]

Chess players formulate moves as the event itself plays out, while handicappers are restricted to sizing things up in advance. Thus chess can more easily be reduced to science-meaning an infallible series of 'correct' responses as possible outcomes and permutations become fewer in number and easier to weigh. Handicappers may tackle a race one factor at a time, but the process is based purely on supposition, and it's progress purely a leap of faith lacking linear sequence or fixed points to proceed from. Thanks very much for the response.

[Robert Fischer]

as far as line-making is concerned, I believe heavy favorites to be many times more important than anything else in the race.

[thaskalos]

Originally Posted by mountainman

Chess players formulate moves as the event itself plays out, while handicappers are restricted to sizing things up in advance. Thus chess can more easily be reduced to science-meaning an infallible series of 'correct' responses as possible outcomes and permutations become fewer in number and easier to weigh. Handicappers may tackle a race one factor at a time, but the process is based purely on supposition, and it's progress purely a leap of faith lacking linear sequence or fixed points to proceed from. Thanks very much for the response.

Does this mean that it is easier to beat Garry Kasparov in chess...than it is to out-play the public at the racetrack?

[mountainman]

Originally Posted by thaskalos

Does this mean that it is easier to beat Garry Kasparov in chess...than it is to out-play the public at the racetrack?

No, it means that a world-class handicapper could MORE than hold his own against any computer program.

[markgoldie]

Originally Posted by mountainman

Handicappers may tackle a race one factor at a time, but the process is based purely on supposition, and it's progress purely a leap of faith lacking linear sequence or fixed points to proceed from.

Consider a data base of say 10,000 races as a method of weighing the relationship of factors. This is something of a "fixed point" from which a program might proceed. If you add to that the constant updating of data as new races enter the base and the ability to "learn" from mistakes and/or new information, you can wind up with a formidable handicapping tool, possibly beatable in the short term, but very difficult to outperform over a long haul. But frankly, this is an area where Dave Schwartz or Jeff P. are far beyond me. My only point is and was that the processes of the mind can be broken down into numbers.

[Fastracehorse]

Originally Posted by markgoldie

Maybe my post wasn't quite clear enough. Okay. Let me expand a bit.

You say you make your determination based on a host of different factors. Fine. Presumably you know what these factors are and they are not simply absorbed into your consciousness by an unknown cosmic force of some sort.

So you engage in the process of handicapping, which (correct me if I'm wrong) is sort of a non-verbal assimilation of different facts regarding the past performances of the horses in question. During this process, plusses and minuses strike you about each horse. As you continue the process, the race as a whole begins to take shape; that is, the interactions of these horses with their attributes and shortcomings, fall into a picture in your mind as to how the race will unfold. As this process becomes more clear, you consult the odds' board to see if and how your handicapping opinion can be exploited through some sort of overlay- a flaw in the crowd's thinking that offers you a porthole to profit.

This, you (rightfully) feel is art. Reliant on science, of course, but pure art nonetheless. How in the world could anyone in their right mind think that this elegant intellectual process of the human mind might be broken down into numbers, such that a cold, insensitive machine could possibly reproduce what's going on here?

Well? I do... and here's why. If you slow down your handicapping process, I mean way down, you will see something interesting. At every juncture, with each variable you consider, you are making a decision; at the least, you are forming an opinion. And this opinion is based upon the knowledge (garnered from years of experience I'd guess) that resides in your memory. Furthermore, the decision is based on the opinion that some attribute or another is better than another attribute and importantly by some discernable margin. But if you think about it, what you are doing is the following: Knowing that factor A is good (+), factor B is bad (-), etc. and that the degree of these factors 0-10 or 0-100 or 0-1000 (it doesn't matter the scale of range) is relevant and important.

The essential point is that the mind is measuring and measurements are a function of numbers at their essence. So these non-verbal thoughts which occur during the handicapping process can be accurately described as numbers. In fact they can be more accurately described as numbers than as either verbal or non-verbal thought. The struggle of understanding verbal measurements was something I addressed in my earilier post. How much is "a little bit; a lot; a great deal; a smidgen; a pinch; a tad; a medium-lot; a whole-lot"?

The fact is, if you dissected your handicapping and assigned numerical values to all the variables you consider, you would become a much better handicapper in the long run. Why? Because you would create a record of the measurements in which you are engaged and a subsequent tweaking might reveal areas where improvement can be achieved.

Measurement-based decisions occur when the conglomeration of measurements reach some tipping point. Since numbers can accurately describe the measurements, they can also accurately describe the tipping points which create decisions.

And so, all of this can be done. I didn't say it was easy. In fact, I characterized it as a "struggle." In this struggle, you will need to deconstruct your process, step by step. But even if you do this not with the intention of converting your handicapping into computer software, you should get a nice benefit from the exercise. As far as making a betting line, I agree with you that this is unnecessary. However, I think it's hard to dispute the counter-argument that any measurement of value and what may constitue a trigger-point for a wager has merit. At the least, it gives you a historic record which you may then assess as you continue the quest for profits in a very tough game.

You can argue that the handicapping process can have the layers peeled away and each individual layer attributed a #; so that you have a finite measurement of each part that joins to form the whole.

In some extent you are right ; but you are missing the handicapping point all together. Finite measurements are important only to the level of 'good enoughness.' Most speed fig players know that their fig only has to be good enough - 90% accurate.

I think most players realize this when experimenting with times of races. They are only a measuring stick of the past. They aren't good enough.

What a player needs to know is what leads to winners. A speed fig maybe helps some: but experience from seeing the circumstances unfold doesn't require a formulation of a # churned out by a computer after data has been enterd by the player.

What is important to picking winners?

A horse looks incredibly sharp on the track - you can't know that beforehand - and you don't have time to punch this in the PC - and most of all, you don't need to - and why would I waste the time?

fffastt

[Lasix67]

Cramer has allowed me the chance to look at things outside of the box, but I am a computer handicapper that applies different angles according to the data. Works well for me because most everyone is looking at the same pp's from drf or bris and it is public against one another so you need a different perspective in my opinion.

[fmolf]

Originally Posted by Lasix67

Cramer has allowed me the chance to look at things outside of the box, but I am a computer handicapper that applies different angles according to the data. Works well for me because most everyone is looking at the same pp's from drf or bris and it is public against one another so you need a different perspective in my opinion.

remember that a computer is only as good as the person inputting the raw data!...the old "garbage in garbage out" syndrome.Computer handicappers whether you like it or not have to tell the machine what factors to look for and what value to assign to each factor.Where computers really help in handicapping is sorting through types of races and/or finding spot angle plays that might offer good value.

[markgoldie]

Originally Posted by fmolf

remember that a computer is only as good as the person inputting the raw data!...the old "garbage in garbage out" syndrome.Computer handicappers whether you like it or not have to tell the machine what factors to look for and what value to assign to each factor.Where computers really help in handicapping is sorting through types of races and/or finding spot angle plays that might offer good value.

Exactly. And that's what Lasix is saying. Computers don't cause you to lose control over anything, they enable you to do things faster. They force you to put a solid number on opinions which you only have floating around in your head in some soup of impressions.

Computers are simply mega-sized, mega-fast calculators, data-storers, and retrievers. But when you mention them in the context of handicapping, right away you get people who think you've lost your soul or something. Well, how about the guy who uses a hand calculator and a room-full of old racing forms? Is he okay? Does he pass the test of humanistic artist/scientist? And suppose he had a crew of a thousand employees all with calculators and old racing forms? Is he still okay? Sure, the ideas must come from the maestro. But the same is true with using a computer.

Yes, there are some things like paddock and track inspections that are not amenable to last minute incorporation into a handicapping program (although with notebook computers I'm not sure that even this could not be accomodated). Even so, having the handicapping done and ready waiting for last-minute inspection notes makes a lot of sense. Otherwise, you're really taxing the old noggin to perform under fire.