Does your application or handicapping do this?

[Capper Al]

Just wondering how many of your commercial or homegrown apps or methods would tag a horse Not To Win (NTW). An NTW might even have the best speed figure. This saved my day yesterday from a disaster. I noticed that the favorite was marked NTW by my app. I wheeled my top three picks in a double and pulled myself to almost even for the day.

Please indicate the vendor app or method used to mark NTW if you will. I call my current app Charlie, and it usually marks NTW on form and/or class. I still have to perfect this, but it's doing okay.

[RaceBookJoe]

My normal handicapping ( pen/paper ), will give me an NC ( non-contender/no chance) if the horse hasnt shown any life lately with poor jockey/trainer and a running style that i feel is wrong for winning the race. Not sure that is what you are looking for?

I try to group horses, like i would think most players do, A- horses that CAN win or at worst place, B- horses that most likely wont win but have a chance to finish in the money or 4th if looking for the superfecta, and C- horses with no chance of getting any purse money. Im liking your recent threads.

[Capper Al]

My normal handicapping ( pen/paper ), will give me an NC ( non-contender/no chance) if the horse hasnt shown any life lately with poor jockey/trainer and a running style that i feel is wrong for winning the race. Not sure that is what you are looking for?

I try to group horses, like i would think most players do, A- horses that CAN win or at worst place, B- horses that most likely wont win but have a chance to finish in the money or 4th if looking for the superfecta, and C- horses with no chance of getting any purse money. Im liking your recent threads.

My groupings are

Contenders to Win
Able to come In The Money but not to win
Not even good enough to come In The Money.

[Dave Schwartz]

IMHO, the purpose of the contender selection process is to separate the "likely to win" horses from the "not likely to win" horses.

That should make your n/c horses NTW.

Or am I missing your meaning?

[JustRalph]

Negative UDM's in Jcapper

[Capper Al]

IMHO, the purpose of the contender selection process is to separate the "likely to win" horses from the "not likely to win" horses.

That should make your n/c horses NTW.

Or am I missing your meaning?

Call it what you will, N/C or NTW, it flags horses not likely to win regardless of their speed or pace figures.

[Capper Al]

Negative UDM's in Jcapper

Do UDM's rate from positve to negative with the postive being better or more likely to win than the negatives? Is it possible to have one of the better speed or pace ratings and still have a negative UDM? Is UDM a final evaluation score?

Thanks

[raybo]

In RS if a horse doesn't get highlighted it is unlikely to win.

[traynor]

The app I use rates horses on a number of attributes. The attributes are connected to average mutuels, as well as performance. The "top" ratings in a given category are those most likely to produce a profit. The algorithms use the low-end range of each rating to generate the equivalent of a "NTW" rating. In both cases, the ratings are track, distance, class, and surface specific.

For example, a given rating at Track A may have a profit potential of 36%. However, if an entry with that rating has one or more of the negative attributes (potential less than 10%) it is removed as a contender from the race analysis portion of the application.

That is based on a lot of testing that convinced me the best thing to do with entries with certain attributes is to ignore them--that is what the application does. Including entries with certain negative attributes confounded race projections because those entries would often rate high in other attributes, but rarely win.

[Dan Montilion]

PK... Possible Key
NVC..No Value Contender
BO....Bottom of Exotics only.

[Dan Montilion]

Also: Elm... Elimination.

[JustRalph]

Do UDM's rate from positve to negative with the postive being better or more likely to win than the negatives? Is it possible to have one of the better speed or pace ratings and still have a negative UDM? Is UDM a final evaluation score?

Thanks

Sure, because you build them. I have had horses who were 1st or 2nd choice who had 5 positive udm's and 1 or 2 bad/neg udms at the same time.

You then decide how much of a factor the negatives will be by comparing horses as a whole. You get a bigger picture

Some negative udm's make horses complete throw outs, some just reveal weaknesses compared to others in todays race. Here is some info from jeff

http://www.jcapper.com/podcasts/NegExpectationUDM.mp3

[zerosky]

technically all horses entered for a race have a 'chance.
I separate the contenders into possibles, probables, and outsiders.
my bet always come from the probables.

[Capper Al]

We all seem to be on the same page. For example- I have a total comprehensive point score, but let's say a graded winner ends up in a claiming race. This horse hasn't run in a year and has a trainer with a 5% win ratio and a jockey with a 3% win ratio. With his back class and speed this horse might have the highest total score but would get bounced as NTW because of the other factors.

[barn32]

Just wondering how many of your commercial or homegrown apps or methods would tag a horse Not To Win (NTW). What percentage of the time do your NTW horses actually win?

[Capper Al]

What percentage of the time do your NTW horses actually win?

Actually, one won with the above scenario but it came from an Allowance race. I would say my NTWs win about 5% or less. It's been a while since I researched it. It seems less than 5% now.

[traynor]

What percentage of the time do your NTW horses actually win?

This is an interesting question that may not have a quick and easy answer. The issue--like most other issues involving ratings and rankings--is more complex than it seems on the surface, because the answer should be specific to each (distinct and separate) combination of factors.

It is easy to say that X% of entries with attribute Z fail to win. It is much, much more useful to know that X% of the entries with attribute Z that also rank highest or second highest in primary factor W fail to win. Specifically, an entry that has one or more positive attributes may offset the disadvantages of the negative attribute(s), while another entry with a different combination of positive and attributes may lose.

Simplistic addition and subtraction of variable values is worthless--one negative factor may be more "influential" in the race outcome than a numer of positive factors, and vice versa. The only way that the actual cause-and-effect relationship of any combination of positive and negative attributes of an individual entry can be accurately determined is detailed research in which each combination of attributes is considered separately.

For example, suppose you have records that indicate the best average of the last two Beyer numbers win 35% of the six furlong sprints at Laurel. You also have records that indicate entries with jockeys with less than 10% win rates lose 87% of their races. What is the win probability of an entry with the best average of the last two Beyer numbers when ridden by a jockey with a win rate of less than 10%?

The only way you can answer that question in any useful, meaningful way is if you have detailed, accurate records of a sufficient number of similar scenarios to be "statistically significant." That is, records of situations in which both factors were evident in an entry.

There are few things that will empty your piggy bank faster than massaging a few simplistic numbers with a computer app and believing you have discovered Truth-With-a-Big-T about horse racing. It is not quite that easy.

[TrifectaMike]

This is an interesting question that may not have a quick and easy answer. The issue--like most other issues involving ratings and rankings--is more complex than it seems on the surface, because the answer should be specific to each (distinct and separate) combination of factors.

It is easy to say that X% of entries with attribute Z fail to win. It is much, much more useful to know that X% of the entries with attribute Z that also rank highest or second highest in primary factor W fail to win. Specifically, an entry that has one or more positive attributes may offset the disadvantages of the negative attribute(s), while another entry with a different combination of positive and attributes may lose.

Simplistic addition and subtraction of variable values is worthless--one negative factor may be more "influential" in the race outcome than a numer of positive factors, and vice versa. The only way that the actual cause-and-effect relationship of any combination of positive and negative attributes of an individual entry can be accurately determined is detailed research in which each combination of attributes is considered separately.

For example, suppose you have records that indicate the best average of the last two Beyer numbers win 35% of the six furlong sprints at Laurel. You also have records that indicate entries with jockeys with less than 10% win rates lose 87% of their races. What is the win probability of an entry with the best average of the last two Beyer numbers when ridden by a jockey with a win rate of less than 10%?

The only way you can answer that question in any useful, meaningful way is if you have detailed, accurate records of a sufficient number of similar scenarios to be "statistically significant." That is, records of situations in which both factors were evident in an entry.

There are few things that will empty your piggy bank faster than massaging a few simplistic numbers with a computer app and believing you have discovered Truth-With-a-Big-T about horse racing. It is not quite that easy.

You make some interesting points. I have made the some of the very same points. As I have said on numerous occasions Bayes Theorem is the gateway to understanding horse racing analytics.

Functioning in a Bayesian framework leads to an understanding of casuality.

One way to succeed in horse racing is to have a firm understanding of joint (joint distributions) probabilities, prior probabilities and conditional probability. If you study and understand these concepts the racing world becomes a lot less complex.

Mike (Dr Beav)

[speculus]

........As I have said on numerous occasions Bayes Theorem is the gateway to understanding horse racing analytics.

Functioning in a Bayesian framework leads to an understanding of casuality.

..............
Mike (Dr Beav)

Imagine Trifecta Mike was the guy who, donning the garb of a staunch "frequentist", killed a good thread (http://www.paceadvantage.com/forum/showthread.php?t=62877) started by me only because he suspected my thinking was "Bayesian"?

I just can't believe it!

[Capper Al]

You make some interesting points. I have made the some of the very same points. As I have said on numerous occasions Bayes Theorem is the gateway to understanding horse racing analytics.

Functioning in a Bayesian framework leads to an understanding of casuality.

One way to succeed in horse racing is to have a firm understanding of joint (joint distributions) probabilities, prior probabilities and conditional probability. If you study and understand these concepts the racing world becomes a lot less complex.

Mike (Dr Beav)

Once a system is running and collecting data, then the Bayesian theories can verify the results. Until then the Bayesians know nothing.

[TrifectaMike]

Imagine Trifecta Mike was the guy who, donning the garb of a staunch "frequentist", killed a good thread (http://www.paceadvantage.com/forum/showthread.php?t=62877) started by me only because he suspected my thinking was "Bayesian"?

I just can't believe it!

Spec, with all due respect....You still don't get it!

If you believe I was taking a frequentist position, you understand less about this subject than I gave you credit for.

Mike (Dr Beav)

[lansdale]

Imagine Trifecta Mike was the guy who, donning the garb of a staunch "frequentist", killed a good thread (http://www.paceadvantage.com/forum/showthread.php?t=62877) started by me only because he suspected my thinking was "Bayesian"?

I just can't believe it!

Hi Spec,

I know you're a smart guy, but you're really wrong about this. TM has made probably hundereds of posts in support of Bayesian handicapping here, and even tried teaching it at times. If he was playing frequentist in one post, maybe he was just messing with you. If you don't believe me, google his last year or so of posts.

Cheers,

lansdale

[lansdale]

Once a system is running and collecting data, then the Bayesian theories can verify the results. Until then the Bayesians know nothing.

Hi Al,

Hope TM makes a further reply here, but you seem not to get Bayesian thinking. It doesn't require that a system be 'up and running and collecting data' to be of value. The essence of Bayes is that it can make excellent predictions using a minimum of data.

Cheers,

lansdale

[Capper Al]

Hi Al,

Hope TM makes a further reply here, but you seem not to get Bayesian thinking. It doesn't require that a system be 'up and running and collecting data' to be of value. The essence of Bayes is that it can make excellent predictions using a minimum of data.

Cheers,

lansdale

That's the problem. We need the minimum of data first. The system needs to be built before we can have a minimum of data. Not here's the data, build the system. Building systems is still an intuitive process. We build it and then we see if it works. I get the essence of Bayes. When my sytem is finish then I'll verify the elements of it with the stats.

[traynor]

That's the problem. We need the minimum of data first. The system needs to be built before we can have a minimum of data. Not here's the data, build the system. Building systems is still an intuitive process. We build it and then we see if it works. I get the essence of Bayes. When my sytem is finish then I'll verify the elements of it with the stats.

Sometimes statistics are a bit like the Oracle at Delphi ... the answers are in there somewhere, but you have to know which questions to ask to get the answers you need.

[traynor]

You make some interesting points. I have made the some of the very same points. As I have said on numerous occasions Bayes Theorem is the gateway to understanding horse racing analytics.

Functioning in a Bayesian framework leads to an understanding of casuality.

One way to succeed in horse racing is to have a firm understanding of joint (joint distributions) probabilities, prior probabilities and conditional probability. If you study and understand these concepts the racing world becomes a lot less complex.

Mike (Dr Beav)

Horse racing is one of those areas that is not just intellectually stimulating enough to make the acquisition of real knowledge on the topic a pleasure, but quite rewarding to those willing to invest the time and effort needed to go beyond the superficial.

[Gamblor]

This is an interesting question that may not have a quick and easy answer. The issue--like most other issues involving ratings and rankings--is more complex than it seems on the surface, because the answer should be specific to each (distinct and separate) combination of factors.

It is easy to say that X% of entries with attribute Z fail to win. It is much, much more useful to know that X% of the entries with attribute Z that also rank highest or second highest in primary factor W fail to win. Specifically, an entry that has one or more positive attributes may offset the disadvantages of the negative attribute(s), while another entry with a different combination of positive and attributes may lose.

Simplistic addition and subtraction of variable values is worthless--one negative factor may be more "influential" in the race outcome than a numer of positive factors, and vice versa. The only way that the actual cause-and-effect relationship of any combination of positive and negative attributes of an individual entry can be accurately determined is detailed research in which each combination of attributes is considered separately.

For example, suppose you have records that indicate the best average of the last two Beyer numbers win 35% of the six furlong sprints at Laurel. You also have records that indicate entries with jockeys with less than 10% win rates lose 87% of their races. What is the win probability of an entry with the best average of the last two Beyer numbers when ridden by a jockey with a win rate of less than 10%?

The only way you can answer that question in any useful, meaningful way is if you have detailed, accurate records of a sufficient number of similar scenarios to be "statistically significant." That is, records of situations in which both factors were evident in an entry.

There are few things that will empty your piggy bank faster than massaging a few simplistic numbers with a computer app and believing you have discovered Truth-With-a-Big-T about horse racing. It is not quite that easy.


This is a good post, but the more you combine factors, the less data on the factor you end up with. Where do you strike the balance between enough data and enough factor definition?

[pondman]

I can tell you the confidence level of a horse shipping between Souhthern and Northern California. It ranges between + 46 and - 84. The public often will over bet the -80 horse and ignore th +43 horse. I'm on it if it + 38 and has the odds.

[dkithore]

technically all horses entered for a race have a 'chance.
I separate the contenders into possibles, probables, and outsiders.
my bet always come from the probables.

How do you differentiate between possibles and probables? Can u pl. clarify for me? Thanks.

DK