PDA

View Full Version : Recency is probably not usefull for betting purposes


DeltaLover
09-11-2013, 10:44 AM
I it amazing how good the crowd is when estimating odds.

Currently I am developing a meta-handicapping system and one of my case studies is researching recency.

As you can see from the results the crowd is extremely accurate in its assessment leaving very little room for profitability from this specific factor.

My sample is covering 15506 races where I am considering only the favorite.

For simplicity I only consider races with no multiple entries.



The expected winners are calculated using the probability without take out. This means that given a 20% takeout an even money shot translates to 60% winning chance.

Take out is calculated by:

race.takeout = 1.0 - 1.0 / sum( [ 1.0 / (1.0 + s.final_odds) for s in race.starters] )

and probability by:

starter.crowd_probability = 1.0 / ( (race.takeout + 1.0) * ( 1.0 + starter.final_odds))

SHORT_LAYOFF_DAYS = 45
LONG_LAYOFF_DAYS = 120



factor expected observed rate roi
================================================== ================================
all_favorites 5576.46 5689.00 1.02 0.86
first_time_out 135.66 135.00 1.00 0.86
layoff 802.53 820.00 1.02 0.87
long_layoff 295.39 287.00 0.97 0.84
second_of_layoff 877.79 875.00 1.00 0.83
third_of_layoff 629.83 677.00 1.07 0.91
deep_form_cycle 2835.25 2895.00 1.02 0.87

number of races: 15506

Ocala Mike
09-11-2013, 11:39 AM
All I get out of your analysis is the understanding that recency factors are efficiently incorporated into the AGGREGATE of bets made by the public. This does NOT, for me. validate the title of your thread.

In other words, recency IS useful for betting purposes in specific individual cases when coupled with other factors such as trainer patterns, workouts, class of race, etc.

It's like someone posting a thread stating "Favoritism is probably not useful for betting purposes," and then showing favorites winning at 33% but paying far less than 2/1 on average.

Jeff P
09-11-2013, 01:03 PM
Delta, you wrote:
Recency is probably not useful for betting purposesBased on the sample results you posted, I might draw an entirely different conclusion as to usefulness of the data.

Looking at the data, favorites 3rd off of layoff and favorites long_layoff produce two distinct outcomes.

Knowing that, the question becomes:

Are these distinct outcomes different enough from public assessment that you can work them into your model in such in a way that your model is incrementally improved?

My answer would be: Maybe (and worth the attempt.)

Perhaps a logit transformation for each? Or maybe unique decision trees?:

1. Favorites 3rd off of layoff.

2. Favorites long_layoff.

3. Favorites not 3rd off of layoff or long_layoff.

Just a suggestion.



-jp

.

DeltaLover
09-11-2013, 01:31 PM
You are absolutely right as far long-layoff and 3rd of the bench are going.

The crowd shows a systematic underestimate bias for these two factors. Unfortunately it is not enough to overcome the takeout.

I will post more related data soon.

Delta, you wrote:
Based on the sample results you posted, I might draw an entirely different conclusion as to usefulness of the data.

Looking at the data, favorites 3rd off of layoff and favorites long_layoff produce two distinct outcomes.

Knowing that, the question becomes:

Are these distinct outcomes different enough from public assessment that you can work them into your model in such in a way that your model is incrementally improved?

My answer would be: Maybe (and worth the attempt.)

Perhaps a logit transformation for each? Or maybe unique decision trees?:

1. Favorites 3rd off of layoff.

2. Favorites long_layoff.

3. Favorites not 3rd off of layoff or long_layoff.

Just a suggestion.



-jp

.

LottaKash
09-11-2013, 02:38 PM
Delta, you wrote:
Based on the sample results you posted, I might draw an entirely different conclusion as to usefulness of the data.

Looking at the data, favorites 3rd off of layoff and favorites long_layoff produce two distinct outcomes.

Knowing that, the question becomes:

Are these distinct outcomes different enough from public assessment that you can work them into your model in such in a way that your model is incrementally improved?

My answer would be: Maybe (and worth the attempt.)

Perhaps a logit transformation for each? Or maybe unique decision trees?:

1. Favorites 3rd off of layoff.

2. Favorites long_layoff.

3. Favorites not 3rd off of layoff or long_layoff.

Just a suggestion.



-jp

.

I would add this to my list:.... Did the horse, before the layoff, have the "ability" to beat these, when in form enough to do so ?....That would be my biggest question...

Robert Goren
09-11-2013, 02:59 PM
Break down your numbers by turf sprints and routes and by dirt sprint and routes and you may see a different set of numbers. The other thing skews numbers is that in the spring it is not unusual to see full fields of horses come off lay offs.

DeltaLover
09-12-2013, 10:12 AM
As a continuation on the topic, here you can see the behavior of layoff days.

Commonly we use specific intervals for layoff and long layoffs, set to 35 (or 45 in some cases) and 180. The objective of this exercise is to clarify if these intervals are statistically significant.

http://www.themindofagambler.com/notes/recency.html

raybo
09-13-2013, 06:07 AM
Recency should be considered with other factors, not as a single factor. Did the horse need rest? Is he/she likely to improve with a layoff? Was the horse truly in need of a layoff, or were there other reasons that the horse was laid off? Etc..

Layoffs, by themselves, mean very little, especially in large samples, like yours, from different tracks, surfaces, distances, classes, ages, time of year, etc., etc., etc.. Also, you're only considering favorites - Why?

Your research is seriously lacking data, and not a little "common sense".

Red Knave
09-13-2013, 08:22 AM
As a continuation on the topic, here you can see the behavior of layoff days.

Commonly we use specific intervals for layoff and long layoffs, set to 35 (or 45 in some cases) and 180. The objective of this exercise is to clarify if these intervals are statistically significant.

http://www.themindofagambler.com/notes/recency.html

The data for Turf only and Dirt only are screwed up.

DeltaLover
09-13-2013, 09:03 AM
Also, you're only considering favorites - Why?


The favoritism of a horse is simply another handicapping factor. Conceptually, layoff and favorite is similar to layoff and long distance or layoff and two years old. What makes favorite such an important component is exactly that it validates the largest portion of the pool making decisions about the creation of overlays easier.

Most handicapping factors behave differently for the favorite(s) than for the higher prices.

As is commonly known the favorite represent a small (but very consistent) overlay (favorites win approximately 1.02 times more than their suggested share). This is a good indicator why they should be examined separately.


Recency should be considered with other factors, not as a single factor. Did the horse need rest? Is he/she likely to improve with a layoff? Was the horse truly in need of a layoff, or were there other reasons that the horse was laid off? Etc..

A handicapping factor is represented by any scenario that can be described by past performance data. It can be something like third of a layoff, switching to an apprentice jockey running in the last day of the month. This factor is either significant or not. Of course after we can always add more components trying to further specify the factor until we reach some significant example. Besides that, I think that questions like Did the horse need rest? Is he/she likely to improve with a layoff? Simply cannot be answered and just create confusion to the whole process. What really matters is the data and not the speculated reasoning.


Your research is seriously lacking data, and not a little "common sense".

The research is covering approximately 15,000 races and has enough matches for the factors I am using. For more you can read here:

https://sites.google.com/a/lakeheadu.ca/bweaver/Home/statistics/notes/chisqr_assumptions


I do not understand what you mean by 'common sense'. Please clarify.

5k-claim
09-13-2013, 09:37 AM
Recency should be considered with other factors, not as a single factor. Did the horse need rest? Is he/she likely to improve with a layoff? Was the horse truly in need of a layoff, or were there other reasons that the horse was laid off? Etc.. And how on earth do you, raybo, know the answers to any of these questions for a given horse?

.

sjk
09-13-2013, 09:45 AM
The way I look at it is to consider a large number of horses that I would bet on after considering lots of information but not considering recency.

Would recency affect my willingness to bet or the minimum acceptable price?

The answer for me is yes. I don't bet on horses that haven't run in the past 120 days.

HUSKER55
09-13-2013, 09:54 AM
I don't bet any horse that hasn't raced in the last 65 days. Do I get clipped once in a while. Who doesn't? The only time I even consider it is if the trainer stat is over 20% and I prefer over 25%.

but that is me.

mountainman
09-13-2013, 12:55 PM
As with any factor, outmaneuvering the (incredibly sophisticated) public on layup horses entails having superior information and handicapping skills.

Rested horses should be evaluated on an individual basis, not just in the context of a trainer's relevant (and thus widely disseminated AND overbet)stats. Conditions, class-moves, and today's placement, amongst other things, must be factored in to make an informed prediction.

Part art/part science. What else is new?

raybo
09-13-2013, 01:15 PM
And how on earth do you, raybo, know the answers to any of these questions for a given horse?

.

By "thinking like a trainer", which has been posted before. I'm not the only one here who can tell when a horse needs a rest. Do a search.

raybo
09-13-2013, 01:16 PM
As with any factor, outmaneuvering the (incredibly sophisticated) public on layup horses entails having superior information and handicapping skills.

Rested horses should be evaluated on an individual basis, not just in the context of a trainer's relevant (and thus widely disseminated AND overbet)stats. Conditions, class-moves, and today's placement, amongst other things, must be factored in to make an informed prediction.

Part art/part science. What else is new?

Exactly!

thaskalos
09-13-2013, 02:18 PM
Having followed this site with some consistency these last few years...I've seen plenty of handicapping factors called "statistically insignificant"...either because they lack in predictive value -- or, because whatever predictive value they have is already accounted for by the wagering crowd.

Recency is no longer significant because laid-off horses do just as well or better...and speed and pace figures are no longer significant because they are greatly overused -- and already reflected on the tote board.

Class and early speed were also once considered potent handicapping factors -- especially when used in combination -- but I suppose the case can be made that these factors too have been "overexposed"...and they too are already reflected on the tote board.

What I haven't seen is an example or two of this "contrarian" type of handicapping I've been reading about...where "underexposed" factors are put to the task of handicapping.

What do the enlightened handicappers of the modern era use...in place of these "inadequate" and "overexposed" factors of yesteryear? Could they supply us with a handicapping example or two...without describing their methods in any detail, of course. Just a general overview would do...just so we can get a general idea.

I, for one, am not looking for a "description" of the process; I just want to see the process at work.

My apologies if my post is not entirely on topic. I was going to start another thread on what is "useful for betting purposes" -- but maybe now I don't have to...

BlueShoe
09-13-2013, 03:36 PM
Recency is no longer significant because laid-off horses do just as well or better...and speed and pace figures are no longer significant because they are greatly overused -- and already reflected on the tote board.

Class and early speed were also once considered potent handicapping factors -- especially when used in combination -- but I suppose the case can be made that these factors too have been "overexposed"...and they too are already reflected on the tote board..
If these factors no longer matter what does? Not much left. Form cycles, trip handicapping, and trainer patterns perhaps? Speed, class, and pace have always been the heart of handicapping, hard to envision these factors to be insignificant. As for recency, still like it regardless of what the data base guys say.

traynor
09-13-2013, 03:46 PM
Having followed this site with some consistency these last few years...I've seen plenty of handicapping factors called "statistically insignificant"...either because they lack in predictive value -- or, because whatever predictive value they have is already accounted for by the wagering crowd.

Recency is no longer significant because laid-off horses do just as well or better...and speed and pace figures are no longer significant because they are greatly overused -- and already reflected on the tote board.

Class and early speed were also once considered potent handicapping factors -- especially when used in combination -- but I suppose the case can be made that these factors too have been "overexposed"...and they too are already reflected on the tote board.

What I haven't seen is an example or two of this "contrarian" type of handicapping I've been reading about...where "underexposed" factors are put to the task of handicapping.

What do the enlightened handicappers of the modern era use...in place of these "inadequate" and "overexposed" factors of yesteryear? Could they supply us with a handicapping example or two...without describing their methods in any detail, of course. Just a general overview would do...just so we can get a general idea.

I, for one, am not looking for a "description" of the process; I just want to see the process at work.

My apologies if my post is not entirely on topic. I was going to start another thread on what is "useful for betting purposes" -- but maybe now I don't have to...

It may be that those who use the processes prefer to restrict access to keep them profitable. It would be rare (or unknown) in business to give away a cash cow--at least until it had run dry. First preference is use it. Second preference is sell it when it is no longer profitable enough to use. Third preference is acquire the good will of (and possible future purchases from) potential customers by giving it away when it is no longer good enough to sell.

Basic business. There is no reason why horse racing should be any different.

Robert Fischer
09-13-2013, 03:47 PM
Form Cycle and Intent offer more clues to recency.

All of these are probably best used as a peripheral.

raybo
09-13-2013, 04:25 PM
If these factors no longer matter what does? Not much left. Form cycles, trip handicapping, and trainer patterns perhaps? Speed, class, and pace have always been the heart of handicapping, hard to envision these factors to be insignificant. As for recency, still like it regardless of what the data base guys say.

Nobody is saying they no longer matter, but generally, they have little value as "stand alone" factors. In combination with other factors they are still very important, and always will be.

You have to remember, databases are only as good as the data that is in them, and what questions you are asking of it. Otherwise they are just stats.

LottaKash
09-13-2013, 04:27 PM
If these factors no longer matter what does? Not much left. Form cycles, trip handicapping, and trainer patterns perhaps? Speed, class, and pace have always been the heart of handicapping, hard to envision these factors to be insignificant. As for recency, still like it regardless of what the data base guys say.

Blue Shoe, all those above mentioned things are still usable and viable, but in today's game, it is, and always has been, about getting value by using those very things...Only today, with so many of those variables being overused at the windows, one must learn to more creative with that same information that is still being offered over the counter....One must learn to use these elements in the "abstract", if one is going to beat the game these days....

Those factors will never go out of play, imo, just utilized more creatively by today's more successful bettor(s)..And, that is the message that I get from certain posters....Others feel this too I perceive...

Many modern day contrarian handicappers, despite what they say, are still using those very basics, whether they will allude to that or not....They are just more creative than some of the others of us, who are still trudging along, searching for the "ultimate-automatics" in our same old way...Fwiw, I am an old guy set in his handicapping way, but I still find enough "overlays" to keep me interested in this ole game of ours, despite being reminded from time to time, that I am at a very serious disadvantage in today's game...Makes me "smile" a bit, as much as I "wonder" as well, about what they are trying to tell me, without really telling me....

I have to admit tho, sometimes after reading certain posts, I come away with a temporary feeling of inadequacy, but that just seems to melt away when I catch another "logical-longshot", using the same old traditional ways of the "olden days n' ways"....

Horseracing has not changed all that much, perhaps, since they first started the game, thousands of years ago....It is still all about Fitness and Proper Classification, and the riding talent to get home first, imo....

raybo
09-13-2013, 04:30 PM
Form Cycle and Intent offer more clues to recency.

All of these are probably best used as a peripheral.

Those 2 things are integral to analyzing the value, or lack thereof, of layoffs, regardless of their length.

raybo
09-13-2013, 04:33 PM
Blue Shoe, all those above mentioned things are still usable and viable, but in today's game, it is, and always has been, about getting value by using those very things...Only today, with so many of those variables being overused at the windows, one must learn to more creative with that same information that is still being offered over the counter....One must learn to use these elements in the "abstract", if one is going to beat the game these days....

Those factors will never go out of play, imo, just utilized more creatively by today's more successful bettor(s)..And, that is the message that I get from certain posters....Others feel this too I perceive...

Many modern day contrarian handicappers, despite what they say, are still using those very basics, whether they will allude to that or not....They are just more creative than some of the others of us, who are still trudging along, searching for the "ultimate-automatics" in our same old way...Fwiw, I am an old guy set in his handicapping way, but I still find enough "overlays" to keep me interested in this ole game of ours, despite being reminded from time to time, that I am at a very serious disadvantage in today's game...Makes me "smile" a bit, as much as I "wonder" as well, about what they are trying to tell me, without really telling me....

I have to admit tho, sometimes after reading certain posts, I come away with a temporary feeling of inadequacy, but that just seems to melt away when I catch another "logical-longshot", using the same old traditional ways of the "olden days n' ways"....

Horseracing has not changed all that much, perhaps, since they first started the game, thousands of years ago....It is still all about Fitness and Proper Classification, and the riding talent to get home first, imo....

Good post, and don't forget, we still have to wager intelligently, and with patience,discipline and consistency, which I feel is a major downfall for the majority of losing players, more so than the handicapping portion.

LottaKash
09-13-2013, 04:41 PM
Good post,

Thanks for saying, Raybo....:jump: ..Nice to have a "feelgood" moment, from time to time...From you, it means something..:ThmbUp:

TrifectaMike
09-13-2013, 06:01 PM
Having followed this site with some consistency these last few years...I've seen plenty of handicapping factors called "statistically insignificant"...either because they lack in predictive value -- or, because whatever predictive value they have is already accounted for by the wagering crowd.

Recency is no longer significant because laid-off horses do just as well or better...and speed and pace figures are no longer significant because they are greatly overused -- and already reflected on the tote board.

Class and early speed were also once considered potent handicapping factors -- especially when used in combination -- but I suppose the case can be made that these factors too have been "overexposed"...and they too are already reflected on the tote board.

What I haven't seen is an example or two of this "contrarian" type of handicapping I've been reading about...where "underexposed" factors are put to the task of handicapping.

What do the enlightened handicappers of the modern era use...in place of these "inadequate" and "overexposed" factors of yesteryear? Could they supply us with a handicapping example or two...without describing their methods in any detail, of course. Just a general overview would do...just so we can get a general idea.

I, for one, am not looking for a "description" of the process; I just want to see the process at work.

My apologies if my post is not entirely on topic. I was going to start another thread on what is "useful for betting purposes" -- but maybe now I don't have to...

Thask, good post (like most of your posts) and you deserve an answer.

Class and early speed were also once considered potent handicapping factors -- especially when used in combination -- but I suppose the case can be made that these factors too have been "overexposed"...and they too are already reflected on the tote board.

Class and early speed are still potent handicapping factors and are also oversexposed... generally reflected on the tote-board. Nevertheless still potent and have to be considered. What has changed drastically is how the factors are used in the abstract.

The change is not a "contrarian type of handicapping", but a more abstract and superior data processing methodologies and abstractions.

For example, speed ratings, approximately 70% of the money bet is based on speed ratings (last race, best of last three, average of last three, best of last 10, etc). This sort of analysis will no longer suffice. However, speed ratings have predictive power. They can't simply be ignored. Instead they need to viewed withing the context of the race and in relation to the other runners.

Without giving away the store (Traynor), here is what I mean.

Take an example race, there is a speed component which is race dependent ( a function of all the horses ). This is the first level of a hierarchy. How each horse will perform on a speed basis is controlled by the hierachy.

Each horse within the race will be affected by this hierachy. Some speed figures will shrink. This shrinkage may cause the speed rating factor to be a meaningless compentent of the race or it might prove to be extremely significant. You don't know until the analysis is performed. It is not something that a individual figure will expose.

Secondly, the amount and type of data analysis performed today by some is much superior and the intergration of the analysis is beyond the average bettor.

Mike

raybo
09-13-2013, 06:18 PM
Thanks for saying, Raybo....:jump: ..Nice to have a "feelgood" moment, from time to time...From you, it means something..:ThmbUp:

You're more than welcome! You stated things very well. Credit where credit is due!

And, thank you -- for the "flattery". :)

raybo
09-13-2013, 06:21 PM
Thask, good post (like most of your posts) and you deserve an answer.

Class and early speed were also once considered potent handicapping factors -- especially when used in combination -- but I suppose the case can be made that these factors too have been "overexposed"...and they too are already reflected on the tote board.

Class and early speed are still potent handicapping factors and are also oversexposed... generally reflected on the tote-board. Nevertheless still potent and have to be considered. What has changed drastically is how the factors are used in the abstract.

The change is not a "contrarian type of handicapping", but a more abstract and superior data processing methodologies and abstractions.

For example, speed ratings, approximately 70% of the money bet is based on speed ratings (last race, best of last three, average of last three, best of last 10, etc). This sort of analysis will no longer suffice. However, speed ratings have predictive power. They can't simply be ignored. Instead they need to viewed withing the context of the race and in relation to the other runners.

Without giving away the store (Traynor), here is what I mean.

Take an example race, there is a speed component which is race dependent ( a function of all the horses ). This is the first level of a hierarchy. How each horse will perform on a speed basis is controlled by the hierachy.

Each horse within the race will be affected by this hierachy. Some speed figures will shrink. This shrinkage may cause the speed rating factor to be a meaningless compentent of the race or it might prove to be extremely significant. You don't know until the analysis is performed. It is not something that a individual figure will expose.

Secondly, the amount and type of data analysis performed today by some is much superior and the intergration of the analysis is beyond the average bettor.

Mike

Well said! :ThmbUp:

DeltaLover
09-13-2013, 07:10 PM
However, speed ratings have predictive power. They can't simply be ignored.

:ThmbUp:

There is no doubt that the final time and its derivatives represent the most influential handicapping factor.

Compare the significance of the layoff based factors to those based in prime power.

I am using Yates correction since we only have 1 degree of freedom.

Expected winners are calculated based in the crowd's odds eliminating the take out.


first_time_out
MATCHES : observed winners: 349.00 expected winners: 356.97
NOMATCHES: observed winners: 1316.00 expected winners: 1308.03
x2: 0.19876 NOT significant
========================================
layoff
MATCHES : observed winners: 1082.00 expected winners: 1080.31
NOMATCHES: observed winners: 3694.00 expected winners: 3695.69
x2: 0.00169 NOT significant
========================================
long_layoff
MATCHES : observed winners: 500.00 expected winners: 499.77
NOMATCHES: observed winners: 2401.00 expected winners: 2401.23
x2: 0.00018 NOT significant
========================================
second_of_layoff
MATCHES : observed winners: 1222.00 expected winners: 1231.73
NOMATCHES: observed winners: 4047.00 expected winners: 4037.27
x2: 0.09034 NOT significant
========================================
third_of_layoff
MATCHES : observed winners: 907.00 expected winners: 897.41
NOMATCHES: observed winners: 3465.00 expected winners: 3474.59
x2: 0.11586 NOT significant
========================================
deep_form_cycle
MATCHES : observed winners: 3947.00 expected winners: 3940.81
NOMATCHES: observed winners: 3826.00 expected winners: 3832.19
x2: 0.01668 NOT significant
========================================
top_two_primer_power
MATCHES : observed winners: 4359.00 expected winners: 4143.89
NOMATCHES: observed winners: 3870.00 expected winners: 4085.11
x2: 22.38871 significant
========================================
top_three_primer_power
MATCHES : observed winners: 5530.00 expected winners: 5336.80
NOMATCHES: observed winners: 2682.00 expected winners: 2875.20
x2: 19.87398 significant
========================================
top_primer_power
MATCHES : observed winners: 2614.00 expected winners: 2509.08
NOMATCHES: observed winners: 5616.00 expected winners: 5720.92
x2: 6.25097 significant
========================================

Overlay
09-13-2013, 07:15 PM
To me, almost any handicapping factor can have value as long as it is given its proper weight (whether in a positive or negative sense), and is used as a means of evaluating a horse's overall chance of winning in comparison to its competition, rather than as a go/no-go elimination tool designed to focus solely on finding the one horse that is most likely to win.

That having been said, I would think that correlating recency with quality of performance would still produce more useful results for that purpose than recency alone.

DeltaLover
09-13-2013, 07:22 PM
To me, almost any handicapping factor can have value as long as it is given its proper weight (whether in a positive or negative sense), and is used as a means of evaluating a horse's overall chance of winning in comparison to its competition, rather than as a go/no-go elimination tool designed to focus solely on finding the one horse that is most likely to win.

That having been said, I would think that correlating recency with quality of performance would still produce more useful results for that purpose than recency alone.

In my opinion:

A handicapping factor can either have value or not. As far as we can describe a factor in terms of past-performance data we can easily conclude if it adds value (as a positive or negative indicator) or not (is perceived by the crowd correctly). The vast majority of any factor we can imagine falls in the later category. A handicapping factor can have extremely high prediction value but still remain neutral for betting purposes if already known to the public.

iceknight
09-13-2013, 08:16 PM
Let me just state that Whiff is not the only one that can sing odes to this site on a regular basis.

There are amazing gems of wisdom and knowledge shared on several threads here. This one is another one where a whole book chapter is succinctly discussed within a few posts. Thanks guys!

Segwin
09-13-2013, 08:39 PM
Let me just state that Whiff is not the only one that can sing odes to this site on a regular basis.

There are amazing gems of wisdom and knowledge shared on several threads here. This one is another one where a whole book chapter is succinctly discussed within a few posts. Thanks guys!

Quite :ThmbUp:

MJC922
09-13-2013, 09:31 PM
After doing some in-depth research on recent action over the past year I can assure you as a signal of intent and of soundness it's underestimated. Now that doesn't mean the underestimation is frequent, in a broad sense you probably won't find any factor which is anymore, however there are spacing sequences interwoven with levels of effort which if examined closely will lead to a few insights. I look at research as a way to lower my standard error in assessing specific factors. We all innately through years of experience have a pretty good feel for what the factors are 'worth' but when a ho hum factor becomes synergistic is what interests me.

thaskalos
09-13-2013, 11:55 PM
Thask, good post (like most of your posts) and you deserve an answer.

Class and early speed were also once considered potent handicapping factors -- especially when used in combination -- but I suppose the case can be made that these factors too have been "overexposed"...and they too are already reflected on the tote board.

Class and early speed are still potent handicapping factors and are also oversexposed... generally reflected on the tote-board. Nevertheless still potent and have to be considered. What has changed drastically is how the factors are used in the abstract.

The change is not a "contrarian type of handicapping", but a more abstract and superior data processing methodologies and abstractions.

For example, speed ratings, approximately 70% of the money bet is based on speed ratings (last race, best of last three, average of last three, best of last 10, etc). This sort of analysis will no longer suffice. However, speed ratings have predictive power. They can't simply be ignored. Instead they need to viewed withing the context of the race and in relation to the other runners.

Without giving away the store (Traynor), here is what I mean.

Take an example race, there is a speed component which is race dependent ( a function of all the horses ). This is the first level of a hierarchy. How each horse will perform on a speed basis is controlled by the hierachy.

Each horse within the race will be affected by this hierachy. Some speed figures will shrink. This shrinkage may cause the speed rating factor to be a meaningless compentent of the race or it might prove to be extremely significant. You don't know until the analysis is performed. It is not something that a individual figure will expose.

Secondly, the amount and type of data analysis performed today by some is much superior and the intergration of the analysis is beyond the average bettor.

Mike

Thanks for your answer, Mike...and I agree with what you say. I just get confused at times when I see people claiming that they use no speed and pace figures at all...simply because these figures happen to be overused. IMO...there is a part of the handicapping puzzle which only these figures can unravel...and to refuse to use them at all seems to me to be a case of throwing out the baby with the bathwater.

I also disagree when I hear people say that the "same old stats" cannot be analyzed in new, more creative ways. This "massaging" of the existing data -- as some sophisticated players call it -- is a very worthwhile pursuit, IMO...because the handicapping information that we have at our disposal cannot be considered to be anything even remotely resembling "absolute truth".

These figures that we use are nothing but "tools"...and tools can be used in different ways -- depending on the skill and the creativity of the person who uses them.

A handicapping factor may appear useless from a wagering standpoint, when we look at it from the wide lens of an extensive database...but this doesn't mean that the competent handicapper cannot apply this "useless" factor in a new and more profitable way.

TrifectaMike
09-14-2013, 12:58 AM
:ThmbUp:

There is no doubt that the final time and its derivatives represent the most influential handicapping factor.

Compare the significance of the layoff based factors to those based in prime power.

I am using Yates correction since we only have 1 degree of freedom. (Nice touch.... I appreciate the inclusion)

Expected winners are calculated based in the crowd's odds eliminating the take out.


first_time_out
MATCHES : observed winners: 349.00 expected winners: 356.97
NOMATCHES: observed winners: 1316.00 expected winners: 1308.03
x2: 0.19876 NOT significant
========================================
layoff
MATCHES : observed winners: 1082.00 expected winners: 1080.31
NOMATCHES: observed winners: 3694.00 expected winners: 3695.69
x2: 0.00169 NOT significant
========================================
long_layoff
MATCHES : observed winners: 500.00 expected winners: 499.77
NOMATCHES: observed winners: 2401.00 expected winners: 2401.23
x2: 0.00018 NOT significant
========================================
second_of_layoff
MATCHES : observed winners: 1222.00 expected winners: 1231.73
NOMATCHES: observed winners: 4047.00 expected winners: 4037.27
x2: 0.09034 NOT significant
========================================
third_of_layoff
MATCHES : observed winners: 907.00 expected winners: 897.41
NOMATCHES: observed winners: 3465.00 expected winners: 3474.59
x2: 0.11586 NOT significant
========================================
deep_form_cycle
MATCHES : observed winners: 3947.00 expected winners: 3940.81
NOMATCHES: observed winners: 3826.00 expected winners: 3832.19
x2: 0.01668 NOT significant
========================================
top_two_primer_power
MATCHES : observed winners: 4359.00 expected winners: 4143.89
NOMATCHES: observed winners: 3870.00 expected winners: 4085.11
x2: 22.38871 significant
========================================
top_three_primer_power
MATCHES : observed winners: 5530.00 expected winners: 5336.80
NOMATCHES: observed winners: 2682.00 expected winners: 2875.20
x2: 19.87398 significant
========================================
top_primer_power
MATCHES : observed winners: 2614.00 expected winners: 2509.08
NOMATCHES: observed winners: 5616.00 expected winners: 5720.92
x2: 6.25097 significant
========================================


As usual good stuff! You made your point.

Mike

TrifectaMike
09-14-2013, 01:57 AM
Thanks for your answer, Mike...and I agree with what you say. I just get confused at times when I see people claiming that they use no speed and pace figures at all...simply because these figures happen to be overused. IMO...there is a part of the handicapping puzzle which only these figures can unravel...and to refuse to use them at all seems to me to be a case of throwing out the baby with the bathwater.

I also disagree when I hear people say that the "same old stats" cannot be analyzed in new, more creative ways. This "massaging" of the existing data -- as some sophisticated players call it -- is a very worthwhile pursuit, IMO...because the handicapping information that we have at our disposal cannot be considered to be anything even remotely resembling "absolute truth".

These figures that we use are nothing but "tools"...and tools can be used in different ways -- depending on the skill and the creativity of the person who uses them.

A handicapping factor may appear useless from a wagering standpoint, when we look at it from the wide lens of an extensive database...but this doesn't mean that the competent handicapper cannot apply this "useless" factor in a new and more profitable way.

It's more than using the "same old stats" and the same tools quite often give rise to different results.

Let me offer a very simple example.

There is an upcoming race in which you have no opinion. So, you elicit the opinion of varied ( varied by method ), but good horse players. The opinion is in the form of a probability.

Player 1 says .4 (40%) and Player 2 says .20 (20%).

You respect both opinions and have no reason to believe one player over the other. Therefore, you average the 2 and will accept .3 (30%). Have you improved your position by averaging? Well, it depends.

If Player 1 or Player 2 is correct, your position is worse.

If both players are incorrect, but still reasonably calibrated, you may have improved your position.

If both players are well calibrated, you position may indeed be uncalibrated and definitely lack sharpness.

Let's replace the Players with models:

Model 1 is an Eary Speed Model and Model 2 is an Ability Model. Each model provides you with the most probable winner and associated probability. You are faced with the same problem as before with the elicitation from Player 1 and Player 2.

Solutions to this type of problem can lead to very interesting methodologies. It allows you to work in well established frameworks. The point as Thask stated same stats, same data, same tools, but different results.

Note: From the posts I've read over the years, I'm convinced that at least 2 posters here have explored this area, but I believe one is much further ahead than the other. I can tell you that this methodology crudely implemented can break even at all tracks, all surfaces, all distances and without specifically playing overlays ( simply highest ranked probability horse).

Mike

PICSIX
09-14-2013, 07:39 AM
It's more than using the "same old stats" and the same tools quite often give rise to different results.

Let me offer a very simple example.

There is an upcoming race in which you have no opinion. So, you elicit the opinion of varied ( varied by method ), but good horse players. The opinion is in the form of a probability.

Player 1 says .4 (40%) and Player 2 says .20 (20%).

You respect both opinions and have no reason to believe one player over the other. Therefore, you average the 2 and will accept .3 (30%). Have you improved your position by averaging? Well, it depends.

If Player 1 or Player 2 is correct, your position is worse.

If both players are incorrect, but still reasonably calibrated, you may have improved your position.

If both players are well calibrated, you position may indeed be uncalibrated and definitely lack sharpness.

Let's replace the Players with models:

Model 1 is an Eary Speed Model and Model 2 is an Ability Model. Each model provides you with the most probable winner and associated probability. You are faced with the same problem as before with the elicitation from Player 1 and Player 2.

Solutions to this type of problem can lead to very interesting methodologies. It allows you to work in well established frameworks. The point as Thask stated same stats, same data, same tools, but different results.

Note: From the posts I've read over the years, I'm convinced that at least 2 posters here have explored this area, but I believe one is much further ahead than the other. I can tell you that this methodology crudely implemented can break even at all tracks, all surfaces, all distances and without specifically playing overlays ( simply highest ranked probability horse).

Mike

Good stuff :ThmbUp:

Capper Al
09-14-2013, 09:52 AM
I'm okay with my use of recency. It should be used in combinations with other factors. Granted alone it doesn't have the impact it once had.

Tom
09-14-2013, 10:26 AM
By "thinking like a trainer", which has been posted before. I'm not the only one here who can tell when a horse needs a rest. Do a search.

Didn't we used to call that handicapping? ;)

Tom
09-14-2013, 10:31 AM
You do your work, run your queries, analyze the readouts, factor in the weather and track condition, and you have you statistical, mathematical, geometrical,calculated odds line and you bet accordingly.

Then the gate opens and half the pinheads grab.

mountainman
09-14-2013, 10:53 AM
I have to admit tho, sometimes after reading certain posts, I come away with a temporary feeling of inadequacy

Don't be. Most of these guys are just striving to say very (very) old things in the newest, most self-confident sounding way.

And never equate articulance or writing ability with actual knowledge or skill. I once read-and still somewhat believe-that the sharpest, most successful people are rarely the best spoken. Not only are they too busy DOING to pontificate, but their acumens put them in a place where exceptional communication skills just aren't necessary.

MJC922
09-14-2013, 12:17 PM
You do your work, run your queries, analyze the readouts, factor in the weather and track condition, and you have you statistical, mathematical, geometrical,calculated odds line and you bet accordingly.

Then the gate opens and half the pinheads grab.

Haha, yeah like the Fish Hooks race on Aug24th -- and who should complain when the winner made perfect sense. $7 exacta freebie for the 'team'.

thaskalos
09-14-2013, 02:35 PM
Don't be. Most of these guys are just striving to say very (very) old things in the newest, most self-confident sounding way.

And never equate articulance or writing ability with actual knowledge or skill. I once read-and still somewhat believe-that the sharpest, most successful people are rarely the best spoken. Not only are they too busy DOING to pontificate, but their acumens put them in a place where exceptional communication skills just aren't necessary.
I would rather make money than be brilliant. :)

raybo
09-14-2013, 03:37 PM
Didn't we used to call that handicapping? ;)

Yes, but not too many players nowadays want to spend the time to do it. I might add that you should "think like a good trainer".

raybo
09-14-2013, 03:45 PM
I would rather make money than be brilliant. :)

And yet, you are one of the "best spoken" members here. :confused:

Maybe "best spoken" was not exactly what Mountainman meant?

thaskalos
09-14-2013, 04:11 PM
And yet, you are one of the "best spoken" members here. :confused:

Maybe "best spoken" was not exactly what Mountainman meant?
I don't know what mountainman meant...but I've read enough of his postings to know that I would gladly trade my mastery of the English language for his.

I don't write to impress; I write exactly the way I talk...and I suspect the same is true for mountainman.

Magister Ludi
09-16-2013, 12:06 AM
I can tell you that this methodology crudely implemented can break even at all tracks, all surfaces, all distances and without specifically playing overlays ( simply highest ranked probability horse).Mike

Using BLP to combine models?

TrifectaMike
09-16-2013, 01:04 AM
Using BLP to combine models?

Yes Sir

Mike

HUSKER55
09-16-2013, 07:20 AM
BLP???

mountainman
09-16-2013, 08:54 AM
I don't know what mountainman meant...but I've read enough of his postings to know that I would gladly trade my mastery of the English language for his.

I don't write to impress; I write exactly the way I talk...and I suspect the same is true for mountainman.

Much respect for you. You know that.

classhandicapper
09-16-2013, 11:50 AM
I think there is still value in the "exceptions" within each factor.

1. Class is "generally" fully reflected on the board, but there are still occasions when the public in aggregate is not sure if a horse is dropping or moving up in class (and by how much) because of all the expanded and confusing class designations and conditions, shipping, state bred racing, casino money impacting purses etc...

Most people seem to understand that "trip" matters to figures etc..., but some don't make the leap that trips are subtly tougher and more competitive as you move up in class (and vice versa).

There are occasions when a field of horses can be identified as vastly superior/inferior to the official class designation before horses start running back.

2. Everyone has a good set of speed figures these days, but many handicappers view them very literally or assume their accuracy is beyond reproach.

I am a huge proponent of time based pace and final time figures and think they are especially indispensable in races for lightly raced horses that haven't sorted themselves out yet, but I could write a pamphlet on everything that can go wrong with them on any specific day or for any specific race. IMO there is value in knowing when a specific figure is a nonsensical reflection of ability or even inaccurate.

I think if you attack any factor (class, pace, speed figures, trip, consistency, recency, bias, trainer, jockey, etc..) the value is now in understanding the exceptions to the rules.

mountainman
09-16-2013, 12:06 PM
I think there is still value in the "exceptions" within each factor.

1. Class is "generally" fully reflected on the board, but there are still occasions when the public in aggregate is not sure if a horse is dropping or moving up in class (and by how much) because of all the expanded and confusing class designations and conditions, shipping, state bred racing, casino money impacting purses etc...

Most people seem to understand that "trip" matters to figures etc..., but some don't make the leap that trips are subtly tougher and more competitive as you move up in class (and vice versa).

There are occasions when a field of horses can be identified as vastly superior/inferior to the official class designation before horses start running back.

2. Everyone has a good set of speed figures these days, but many handicappers view them very literally or assume their accuracy is beyond reproach.

I am a huge proponent of time based pace and final time figures and think they are especially indispensable in races for lightly raced horses that haven't sorted themselves out yet, but I could write a pamphlet on everything that can go wrong with them on any specific day or for any specific race. IMO there is value in knowing when a specific figure is a nonsensical reflection of ability or even inaccurate.

I think if you attack any factor (class, pace, speed figures, trip, consistency, recency, bias, trainer, jockey, etc..) the value is now in understanding the exceptions to the rules.

Beyond raceflow and fractions, I think it's important to look at the quality of timed-increments-not just whether the pace was contested, but how fast, on form, was the first flight or outright leader? However accurate and correctly adjusted for track speed, certain splits, at some distances over certain tracks, tend to cluster, and for me, it's as much the quality as the speed of the fractions that matters.

Cratos
09-16-2013, 04:10 PM
BLP???

I believe that Magister Ludi and Trifecta Mike are speaking of using a Bayesian Logic Program (BLP) which is a statistical relational model.

However I could be wrong?

Cratos
09-16-2013, 04:29 PM
I think there is still value in the "exceptions" within each factor.

1. Class is "generally" fully reflected on the board, but there are still occasions when the public in aggregate is not sure if a horse is dropping or moving up in class (and by how much) because of all the expanded and confusing class designations and conditions, shipping, state bred racing, casino money impacting purses etc...

Most people seem to understand that "trip" matters to figures etc..., but some don't make the leap that trips are subtly tougher and more competitive as you move up in class (and vice versa).

There are occasions when a field of horses can be identified as vastly superior/inferior to the official class designation before horses start running back.

2. Everyone has a good set of speed figures these days, but many handicappers view them very literally or assume their accuracy is beyond reproach.

I am a huge proponent of time based pace and final time figures and think they are especially indispensable in races for lightly raced horses that haven't sorted themselves out yet, but I could write a pamphlet on everything that can go wrong with them on any specific day or for any specific race. IMO there is value in knowing when a specific figure is a nonsensical reflection of ability or even inaccurate.

I think if you attack any factor (class, pace, speed figures, trip, consistency, recency, bias, trainer, jockey, etc..) the value is now in understanding the exceptions to the rules.

I find your postulations about “Class” to be intriguing although it is confounding to me how “Class” can be uniquely quantified.

To me, “Class” can be pre-race defined, but always post-race determined and even then the “Class” attribute in horseracing is fleeing unless it becomes consistent at a given level through the winning efforts of the horse.

If I have missed the thesis of your argument about “Class”; please explain.

HUSKER55
09-16-2013, 05:04 PM
thank you cratos!

classhandicapper
09-16-2013, 09:41 PM
I find your postulations about “Class” to be intriguing although it is confounding to me how “Class” can be uniquely quantified.

To me, “Class” can be pre-race defined, but always post-race determined and even then the “Class” attribute in horseracing is fleeing unless it becomes consistent at a given level through the winning efforts of the horse.

If I have missed the thesis of your argument about “Class”; please explain.

I see "class" in a different way than some other handicappers.

To me, class handicapping is looking at field quality and performance in a non speed figure way. It's looking at who beat who, by how much, with what trips etc.. It's an alternative to looking at those same horses in a time based speed and pace figure way trying to do the same thing.

To me, both approaches have advantages and disadvantages.

With a time based approach, you are typically dealing with issues of accuracy. That becomes apparent when you look at the same race using multiple sets of figures. They often disagree...and sometimes by a lot even though both are looking at the same race times and theoretically adjusting for speed of track and other factors that impact time.

With a class based approach, those accuracy problems vanish, but you are often dealing with insufficient information. Most commonly, there is a wide variation of quality among maidens and limited winners, there are crop quality issues, circuit quality issues, and other potentially difficult subjective judgments a class handicapper has to make whose answers leap off the pages when you simply look at times.

Beyond that, I think field quality impacts the times horses run also.

So they are two distinct ways of measuring the much of same thing, but they are also interrelated.

To find value, you have to understand where the other school is going wrong.

DeanT
09-17-2013, 08:39 AM
This is an enjoyable high level thread, imo. Lots of good posts.

I reiterate what Mike and a few others are alluding. In the "old days" we were concerned with picking the 32% of winners with predictive techniques like a fig because that's where the money was. Today, those predictive techniques are still picking 32% of the winners - virtually nothing has changed in the grand scheme of things. But in the 2000's, we've had to look more closely at the other 68%, because that's where the money is.

It makes handicapping a really interesting game, imo. The tenets can stay the same, but how we bet them or fade them move constantly.

DeltaLover
09-17-2013, 09:15 AM
I do not like the separation to class, speed or form factors. A factor starts with some handicapping hypothesis but to presents real value should be converted to a numerical scale hiding the reasoning behind it. We can have infinite factors each of them describing any handicapping aspect, most of them are useless for betting purposes as they are incorporated in the right proportion to the betting pools. The objective is to discover those factors that are miscalculated by the crowd.

traynor
09-17-2013, 09:42 AM
I do not like the separation to class, speed or form factors. A factor starts with some handicapping hypothesis but to presents real value should be converted to a numerical scale hiding the reasoning behind it. We can have infinite factors each of them describing any handicapping aspect, most of them are useless for betting purposes as they are incorporated in the right proportion to the betting pools. The objective is to discover those factors that are miscalculated by the crowd.

Or that are unknown to, unavailable to, or ignored by the crowd, possibly because of a lack of knowledge (by that crowd) of their importance (significance) in the outcome of races.

classhandicapper
09-17-2013, 10:42 AM
I do not like the separation to class, speed or form factors. A factor starts with some handicapping hypothesis but to presents real value should be converted to a numerical scale hiding the reasoning behind it. We can have infinite factors each of them describing any handicapping aspect, most of them are useless for betting purposes as they are incorporated in the right proportion to the betting pools. The objective is to discover those factors that are miscalculated by the crowd.

There is often no value in a single factor overall, but a lot of value if you have superior understanding of a specific circumstance or race using that factor.

For example, you may have the greatest model in the world over a series of 100K races, but if I know your speed figure source made a mistake in a specific race, your model won't value that specific race correctly and neither will the crowd.

So the question becomes, how do I find those races.

1. You can understand the different methodologies of each of the figure makers well (because they often have some wildly different philosophies on some things which leads to different figures for the same race) and know their strengths of weaknesses.

2. You can evaluate the race in a non time based way (like comparative class)

3. You can find a specific subset of races where all speed figures fail badly for whatever reason

4. etc...


That's what I'm getting at.

The public at large and even most advanced statistical modeling has a very good sense of the overall values, but not necessarily all specific race issues.

Capper Al
09-17-2013, 10:47 AM
Agree.

For me, factors work much better in pairs.

traynor
09-17-2013, 10:56 AM
There is often no value in a single factor overall, but a lot of value if you have superior understanding of specific a circumstance or race using that factor.

For example, you may have the greatest model in the world over a series of 100K races, but if I know your speed figure source made a mistake in a specific race, your model won't value that specific race correctly and neither will the crowd.

So the question becomes, how do I find those races.

1. You can understand the different methodologies of each of the figure makers well (because they often have some wildly different philosophies on some things which leads to different figures for the same race) and know their strengths of weaknesses.

2. You can evaluate the race in a non time based way (like comparative class)

3. You can find a specific subset of races where all speed figures fail badly for whatever reason

4. etc...


That's what I'm getting at.

The public at large and even most advanced statistical modeling has a very good sense of the overall values, but not necessarily all specific race issues.

Perhaps. It may also be that such an approach would tend to diminish the value of a (generic) factor (derived from a large sample of races) by interpreting it differently in a specific race or set of races.

The basic fallacy is in expecting a generic factor to apply to a specific race or set of races. That is not what factors do. Factors only suggest possible trends--not specific outcomes in specific races.

None of which is to say that interpretation of data in specific races is not a reasonable thing to do--only that generic factors are (in general) most usefully applied generically.

classhandicapper
09-17-2013, 11:08 AM
Perhaps. It may also be that such an approach would tend to diminish the value of a (generic) factor (derived from a large sample of races) by interpreting it differently in a specific race or set of races.

The basic fallacy is in expecting a generic factor to apply to a specific race or set of races. That is not what factors do. Factors only suggest possible trends--not specific outcomes in specific races.

None of which is to say that interpretation of data in specific races is not a reasonable thing to do--only that generic factors are (in general) most usefully applied generically.

My thinking is probably just simpler.

I'm looking at every factor trying to find subsets within it where the public typically misunderstands the situation.

Another example is track biases.

1. I can find value by being better at identifying the biases

2. If everyone knows that a particular day was biased, I can find value by understanding that not all horses are impacted equally by the same track bias and knowing which horses were impacted greatly and which were not on that day (same going forward if I identify one early in the race card.

The factor is track bias.

Lots of people look for them.

Lots of people keep notes on horses that were hurt of benefitted from a bias.

It gets built into the price to some extent.

In some small number of cases within it I have an insight that is not commonly shared.

TrifectaMike
09-17-2013, 11:56 AM
A major difficulty encountered by most bettors is their use of an inconsistent framework. The constant comparison of apples and oranges. The inconsistency of scales. And the inability of realizing the synergy of factors.

Predicting the outcome of a horse race is a stochastic process. With any framework that is employed, that stochastic nature has to be understood and taken in to account.

An ad-hoc approach is akin to a small sample size problem. The signal in the data doesn't exist and the noise is continually applified.

Mike

classhandicapper
09-17-2013, 12:35 PM
A major difficulty encountered by most bettors is their use of an inconsistent framework. The constant comparison of apples and oranges. The inconsistency of scales. And the inability of realizing the synergy of factors.

Predicting the outcome of a horse race is a stochastic process. With any framework that is employed, that stochastic nature has to be understood and taken in to account.

An ad-hoc approach is akin to a small sample size problem. The signal in the data doesn't exist and the noise is continually applified.

Mike

Knowing that a particular speed figure that a lot of large bettors or modelers are looking at is wrong is not noise.

Knowing that on a particular biased day that horse "X" did not benefit from riding the golden rail that everyone else will be downgrading him for, is not noise.

The idea is to develop the skill to do that with enough accuracy to have an edge. It's just another path to the money.

TrifectaMike
09-17-2013, 01:12 PM
Knowing that a particular speed figure that a lot of large bettors or modelers are looking at is wrong is not noise.

Knowing that on a particular biased day that horse "X" did not benefit from riding the golden rail that everyone else will be downgrading him for, is not noise.

The idea is to develop the skill to do that with enough accuracy to have an edge. It's just another path to the money.

Knowing that a particular speed figure that a lot of large bettors or modelers are looking at is wrong is not noise.

Smart big bettors or good modelers don't base their models on a particular speed rating. (The same for any factor)

Let me try this:

Player 1 has access to superior speed ratings.

Player 2 has access to reasonably good speed ratings and le's add in a bias (for fun).

So, Player 1 has superior

Horse 1..... 90
Horse 2..... 87
Horse 3..... 85

Player 2 has inferior and biased ratings

Horse 1.... 86
Horse 2.... 83
Horse 3.... 82

What is the impact of less accurate and biased ratings on the toteboard.....

ABSOLUTELY NOTHING!

Mike

classhandicapper
09-17-2013, 01:51 PM
Knowing that a particular speed figure that a lot of large bettors or modelers are looking at is wrong is not noise.

Smart big bettors or good modelers don't base their models on a particular speed rating. (The same for any factor)

Let me try this:

Player 1 has access to superior speed ratings.

Player 2 has access to reasonably good speed ratings and le's add in a bias (for fun).

So, Player 1 has superior

Horse 1..... 90
Horse 2..... 87
Horse 3..... 85

Player 2 has inferior and biased ratings

Horse 1.... 86
Horse 2.... 83
Horse 3.... 82

What is the impact of less accurate and biased ratings on the toteboard.....

ABSOLUTELY NOTHING!

Mike


The problem is that your example does not reflect the reality I am talking about.

Group 1 is an influential group of bettors with a bad figure for one key horse in the race, but adequate and accurate information on all other aspects of the race.

You have an accurate figure for the race in question and adequate and accurate information on all other aspects of the race.

The betting lines each makes will be different, but all else being equal yours will be vastly superior.

thaskalos
09-17-2013, 02:12 PM
It is my contention that, during the critical final stage of our handicapping and while we are constructing our bets, we are driven by instincts that we cannot readily explain...and may even be unable to understand. We may have different tools at our disposal which mechanically lead us to a particular point, but, at the critical stage of the handicapping and betting process, I believe our knowledge and experience kick in and instinctively take over...causing us to isolate on particular things, while pushing other things to the background. Whether you call it instinct or intuition...it is the culmination of where we are as players...and it guides us during the final stages of the decision-making process.

At least that's what seems to happen with me...

Capper Al
09-17-2013, 02:15 PM
It is my contention that, during the critical final stage of our handicapping and while we are constructing our bets, we are driven by instincts that we cannot readily explain...and may even be unable to understand. We may have different tools at our disposal which mechanically lead us to a particular point, but, at the critical stage of the handicapping and betting process, I believe our knowledge and experience kick in and instinctively take over...causing us to isolate on particular things, while pushing other things to the background. Whether you call it instinct or intuition...it is the culmination of where we are as players...and it guides us during the final stages of the decision-making process.

At least that's what seems to happen with me...

And this intuition is combining factors that come up with our wager.

TrifectaMike
09-17-2013, 02:15 PM
The problem is that your example does not reflect the reality I am talking about.

Group 1 is an influential group of bettors with a bad figure for one key horse in the race, but adequate and accurate information on all other aspects of the race.

You have an accurate figure for the race in question and adequate and accurate information on all other aspects of the race.

The betting lines each makes will be different, but all else being equal yours will be vastly superior.

Very, very doubtful, if you have good models. (one bad figure or good figure doesn't impact the line to any great degree).

Mike

TrifectaMike
09-17-2013, 02:19 PM
It is my contention that, during the critical final stage of our handicapping and while we are constructing our bets, we are driven by instincts that we cannot readily explain...and may even be unable to understand. We may have different tools at our disposal which mechanically lead us to a particular point, but, at the critical stage of the handicapping and betting process, I believe our knowledge and experience kick in and instinctively take over...causing us to isolate on particular things, while pushing other things to the background. Whether you call it instinct or intuition...it is the culmination of where we are as players...and it guides us during the final stages of the decision-making process.

At least that's what seems to happen with me...

I don't recall the last time I actually thought about a bet.

Mike

thaskalos
09-17-2013, 02:22 PM
I don't recall the last time I actually thought about a bet.

Mike

Can you recall the last time you thought about a race?

TrifectaMike
09-17-2013, 02:34 PM
Can you recall the last time you thought about a race?

Races? All the time, when calibrating my models or developing different metrics or studying effects, but never during the "handicapping" or betting phase.

Mike

raybo
09-17-2013, 02:38 PM
I think that there is a misconception between those who employ models/computer method play, and those who don't. Those who don't employ them must make the kinds of decisions and maybe rely on intuition, etc., that Thasklos states, in order to make smart (or dumb) wagers. Those who do employ them do not have to make those kinds of decisions, and don't have to rely on intuition, those things have been previously employed within the model/computer method. For some of these players, the only decision to be made before placing the bet, on a race that the model/method has shown playable and long term profitable, is whether or not there is an odds requirement involved. For me, in my win play, if the method says play, I play, if the odds requirement has been met. I rarely think about races or individual horses, in win play. My superfecta play is a different story, there is much more to consider, and programming the method for these wager types is much too complex, at least for me.

And, I agree with Mike, races/types of races must be considered during recalibrations and the never ending job of evolving one's methods in order to keep up with the constantly changing racing environment, whether in the game, as a whole, or at particular tracks or times of year/meet, etc..

thaskalos
09-17-2013, 03:37 PM
Races? All the time, when calibrating my models or developing different metrics or studying effects, but never during the "handicapping" or betting phase.

Mike

The bulk of my play is comprised of vertical wagering...and I cannot possibly imagine a way of doing this efficiently without "thinking" during the handicapping and betting phases -- especially when it comes to trifectas and superfectas.

I don't doubt that there is a "mechanical" way of doing it, mind you...it's just beyond my comprehension.

Tom
09-17-2013, 03:53 PM
Very, very doubtful, if you have good models. (one bad figure or good figure doesn't impact the line to any great degree).

Mike

But it can impact the outcome of the race.

classhandicapper
09-17-2013, 04:44 PM
It is my contention that, during the critical final stage of our handicapping and while we are constructing our bets, we are driven by instincts that we cannot readily explain...and may even be unable to understand. We may have different tools at our disposal which mechanically lead us to a particular point, but, at the critical stage of the handicapping and betting process, I believe our knowledge and experience kick in and instinctively take over...causing us to isolate on particular things, while pushing other things to the background. Whether you call it instinct or intuition...it is the culmination of where we are as players...and it guides us during the final stages of the decision-making process.

At least that's what seems to happen with me...

I agree with that completely, but generally my plays are derived from one piece of information (a bias, pace influenced figure, a bogus figure, a trainer pattern, a probable race development, a trip, a horse coming out of deceptively strong field etc..) that gets me interested in the race to begin with. I go into it thinking I may know something about the race that will not be fully reflected on the board. I still go through the full handicapping process and evaluate the price, but typically there is a core reason behind wanting to make the play.

traynor
09-17-2013, 05:23 PM
My thinking is probably just simpler.

I'm looking at every factor trying to find subsets within it where the public typically misunderstands the situation.

Another example is track biases.

1. I can find value by being better at identifying the biases

2. If everyone knows that a particular day was biased, I can find value by understanding that not all horses are impacted equally by the same track bias and knowing which horses were impacted greatly and which were not on that day (same going forward if I identify one early in the race card.

The factor is track bias.

Lots of people look for them.

Lots of people keep notes on horses that were hurt of benefitted from a bias.

It gets built into the price to some extent.

In some small number of cases within it I have an insight that is not commonly shared.

That makes a lot of sense. Thanks for explaining.

traynor
09-17-2013, 05:26 PM
A major difficulty encountered by most bettors is their use of an inconsistent framework. The constant comparison of apples and oranges. The inconsistency of scales. And the inability of realizing the synergy of factors.

Predicting the outcome of a horse race is a stochastic process. With any framework that is employed, that stochastic nature has to be understood and taken in to account.

An ad-hoc approach is akin to a small sample size problem. The signal in the data doesn't exist and the noise is continually applified.

Mike

Exactly. That difficulty is made even worse by the fact that the inconsistency prevents "learning" (in the sense of repeatable processes).

LottaKash
09-17-2013, 05:43 PM
Predicting the outcome of a horse race is a stochastic process.
Mike

I learned a new word today....:jump:

Stochastic comes from the Greek (http://en.wikipedia.org/wiki/Greek_language) word στόχος, which means "aim". It also denotes a target stick; the pattern of arrows around a target stick stuck in a hillside is representative of what is stochastic.

raybo
09-17-2013, 07:32 PM
That word has been thrown around threads here for quite sometime. I wonder how many people ever looked it up? I wonder how many people really know what it means, regarding horse racing? What is a racing example of a stochastic approach? I wonder how many people were already applying it and didn't even realize they were? :lol:

barn32
09-17-2013, 10:49 PM
After reading this thread I feel like Woody Allen sitting at a bar with Warren Beatty, when all of a sudden Marilyn Monroe comes by and starts chatting...with whom do you think she shall go home?

thaskalos
09-17-2013, 11:03 PM
After reading this thread I feel like Woody Allen sitting at a bar with Warren Beatty, when all of a sudden Marilyn Monroe comes by and starts chatting...with whom do you think she shall go home?

With John F. Kennedy

Magister Ludi
09-17-2013, 11:26 PM
A major difficulty encountered by most bettors is their use of an inconsistent framework. The constant comparison of apples and oranges. The inconsistency of scales. And the inability of realizing the synergy of factors.

Predicting the outcome of a horse race is a stochastic process. With any framework that is employed, that stochastic nature has to be understood and taken in to account.

An ad-hoc approach is akin to a small sample size problem. The signal in the data doesn't exist and the noise is continually applified.

Mike

That's an interesting observation. The majority of traders are noise traders. They move all financial markets. The relatively few signal traders correct the mispricings of the noise traders.

DeltaLover
09-18-2013, 08:02 AM
That word has been thrown around threads here for quite sometime. I wonder how many people ever looked it up? I wonder how many people really know what it means, regarding horse racing? What is a racing example of a stochastic approach? I wonder how many people were already applying it and didn't even realize they were? :lol:


Very simple explanation:

- Deterministic: An event who's outcome can be determined in advance. For example after throwing a coin in the air we can be sure that it will fall to the ground

- Stochastic: An event that we can not predict it's outcome. Despite the fact we can be sure that the coin will fail we can not be sure if it will be phase up or not

- Chaotic: An event that in theory can be considered deterministic but based in the number of factors affecting its output, it behaves as stochastic. Such an example is weather forecast.

DeltaLover
09-18-2013, 08:37 AM
As several posters mentioned above, it is impossible for a specific factor to show profitability despite the fact that some of them might behave better that the others. Some of the posters are mentioning that we need to consider class movements, workouts, weight and other changes as additional filters to recency to derive some valuable results.

Of course this is correct

The question though is, what kind of methodology should we use towards this escalation? We are facing several challenges, starting from the description of each scenario going to the rejection methodology and more important to the data availability (a very complex factor will not provide enough research data even when a complete data base is available).

It becomes apparent, that what we need is to create race level ratings, measuring attributes of the race and test the fundamental factors using groups based in them.

A very simple example of this might be testing a factor (like recency) against distance. Of course distance is a binary attribute that can either be route or sprint thus the methodology might be relatively simple.

But, how we treat a case where we want to test recency against not only distance but the amount of 'speed' that exists in the race? Measuring the cumulative speed of the race as a whole is not such a simple task. All the time I hear at the track people stating that this race has a lot of speed or has no speed at all, the question is how can we create an objective rating to classify a race as slow or fast?

I am convinced that the methodology needed for this definition, is one of the most important aspects of value betting and the one that is (almost) completely ignored by the crowd.

Defining such a methodology (or a meta handicapping level as a I call it) allows us to automatically generate race based metrics and have a valid opinion about their effectiveness something that can lead to a such a selection process to ensure profitability 'in the long run'...

classhandicapper
09-18-2013, 09:43 AM
All the time I hear at the track people stating that this race has a lot of speed or has no speed at all, the question is how can we create an objective rating to classify a race as slow or fast?

I am convinced that the methodology needed for this definition, is one of the most important aspects of value betting and the one that is (almost) completely ignored by the crowd.

.

Most people look at prior fractions, pace figures, running styles, jockey tendencies, how certain horses have been successful in the past, how the track is playing and whether the jockeys know it, blinker changes, etc...

The thing is, almost every race is so unique in its combination of these things because the specific horses, jockeys, and conditions are always slightly different, most people go with various major lessons and insights they have learned over time and intuit the rest.

It doesn't lend itself perfectly to an automated systematic approach, though the TimeformUS pace projector is darn good and will do WAY better than most people do on their own.

The difference here is that some people are building models and systems to evaluate every race and try to find value while other people are looking for specific races where they have an insight that potentially makes the race a good source of value.

DeltaLover
09-18-2013, 10:01 AM
Most people look at prior fractions, pace figures, running styles, jockey tendencies, how certain horses have been successful in the past, how the track is playing and whether the jockeys know it, blinker changes, etc...


The problem with this approach is that it is highly subjective and open to many interpretations. We need a systematic approach providing a valid opinion about the attribute(s) in question. It has to be simple and generic but valid. We should start with very simple models, gradually adding constraints until we reach an acceptable behavior. The objective should be the method and not the specific angles (like blinker changes or pace figures).

TrifectaMike
09-18-2013, 10:42 AM
As several posters mentioned above, it is impossible for a specific factor to show profitability despite the fact that some of them might behave better that the others. Some of the posters are mentioning that we need to consider class movements, workouts, weight and other changes as additional filters to recency to derive some valuable results.

Of course this is correct

The question though is, what kind of methodology should we use towards this escalation? We are facing several challenges, starting from the description of each scenario going to the rejection methodology and more important to the data availability (a very complex factor will not provide enough research data even when a complete data base is available).

It becomes apparent, that what we need is to create race level ratings, measuring attributes of the race and test the fundamental factors using groups based in them.

A very simple example of this might be testing a factor (like recency) against distance. Of course distance is a binary attribute that can either be route or sprint thus the methodology might be relatively simple.

But, how we treat a case where we want to test recency against not only distance but the amount of 'speed' that exists in the race? Measuring the cumulative speed of the race as a whole is not such a simple task. All the time I hear at the track people stating that this race has a lot of speed or has no speed at all, the question is how can we create an objective rating to classify a race as slow or fast?

I am convinced that the methodology needed for this definition, is one of the most important aspects of value betting and the one that is (almost) completely ignored by the crowd.

Defining such a methodology (or a meta handicapping level as a I call it) allows us to automatically generate race based metrics and have a valid opinion about their effectiveness something that can lead to a such a selection process to ensure profitability 'in the long run'...

Hierachical Bayes Models.

DL,

Do yourself a favor and google: Hierachical Bayes Normal-Normal Model. It is probably the simplest model to understand, especially since a closed form solution exist for the posterior distribution.

Mike

Cratos
09-18-2013, 10:45 AM
Very, very doubtful, if you have good models. (one bad figure or good figure doesn't impact the line to any great degree).

Mike

Would horseracing be the type of stochastic process that is a random field? Also do you see today’s common used handicapping processes as leaning too much toward a deterministic process and this is causing the “noise” in the current process?

TrifectaMike
09-18-2013, 10:58 AM
Would horseracing be the type of stochastic process that is a random field? Also do you see today’s common used handicapping processes as leaning too much toward a deterministic process and this is causing the “noise” in the current process?

Stochasticity in horse racing prediction is due to uncertainity in the handicapping process and data analysis and not due to physical characteristics of the race. The physical characteristics of the race are chaotic and not random.

The noise in the handicapping process is due to predictions based on noise in the data which impacts the odds. Which has the effect of amplifying the noise as the system feeds back on itself.

Mike

raybo
09-18-2013, 11:05 AM
Yup, as I suspected, some of us are using a stochastic approach, or chaotic if you prefer, by creating individual groupings of "race" types, including the amount of speed in the field, as well as the individual amounts of speed/fractional velocities and preferred running style each horse in the race has shown in the past, all broken out by track, surface, and distance, without using the terms deterministic, stochastic, or chaotic. There is no need to use specific terms to label something, all that is needed is the approach, the individual is free to call it what they will.

In other words, if your goal is to impress others with your vocabulary, then use the 3 terms mentioned here, but if your goal is to help others, then describe the approach instead of simply using a single word. But, then maybe some small, highly educated group of people prefer to try to impress only the other members of that small group. This appears to be the case in these instances, IMO. I often wonder why the members of that small group don't just cut to the chase, instead of continually running around in circles, creating subterfuge and smoke. But then, how else would they justify all those years of pursuing higher education?

TrifectaMike
09-18-2013, 11:59 AM
Yup, as I suspected, some of us are using a stochastic approach, or chaotic if you prefer, by creating individual groupings of "race" types, including the amount of speed in the field, as well as the individual amounts of speed/fractional velocities and preferred running style each horse in the race has shown in the past, all broken out by track, surface, and distance, without using the terms deterministic, stochastic, or chaotic. There is no need to use specific terms to label something, all that is needed is the approach, the individual is free to call it what they will. If that is what you do, that is fine. Those three words have a very specific meaning to me and it is not what you say it is. Nor do I function in the framework you have described.

In other words, if your goal is to impress others with your vocabulary, then use the 3 terms mentioned here, but if your goal is to help others, then describe the approach instead of simply using a single word. But, then maybe some small, highly educated group of people prefer to try to impress only the other members of that small group. This appears to be the case in these instances, IMO. I often wonder why the members of that small group don't just cut to the chase, instead of continually running around in circles, creating subterfuge and smoke. But then, how else would they justify all those years of pursuing higher education?

Mike

raybo
09-18-2013, 12:35 PM
Mike

That's the way it has been described here. Now, if the word does not really mean that, then you made my point. Thanks!

TrifectaMike
09-18-2013, 01:12 PM
That's the way it has been described here. Now, if the word does not really mean that, then you made my point. Thanks!

I gave my definition of stochastic as applied to horse racing:

"Stochasticity in horse racing prediction is due to uncertainity in the handicapping process and data analysis and not due to physical characteristics of the race. The physical characteristics of the race are chaotic and not random."

Do you disagree?
Mike

Cratos
09-18-2013, 01:41 PM
Stochasticity in horse racing prediction is due to uncertainity in the handicapping process and data analysis and not due to physical characteristics of the race. The physical characteristics of the race are chaotic and not random.

The noise in the handicapping process is due to predictions based on noise in the data which impacts the odds. Which has the effect of amplifying the noise as the system feeds back on itself.

Mike

Thanks for a very explanatory response. Your explanation about the physical characteristics of the horserace should be enlightening to those of us who thought of it as random.

Therefore am I correct is assuming that the "noise" in part is driven by the bettors' emotipns.

MJC922
09-18-2013, 03:23 PM
When you use favorites as a source, you're bucketing the comprehensive opinion of the crowd which already in part reflects exactly what it is you're trying to measure.

What I would suggest is that you design and or obtain a performance rating (either class or speed based) for every horse in the race. This should ideally win as often as the favorite does or nearly as often, that way you can measure the impact of any factor for how it differs from the overall. When you bucket the top ranked horse for recency in thousands of races you will see where recency might matter.

If your studies are like mine you will find profitability rarely comes as result of adding other independently potent factors to the mix, i.e top ranked horses do not become more profitable with top trainers, top riders, top pace, you name it, yeah they might win more often but that isn't profit, as you often say the crowd is very good. What I'm seeing is profit only comes from overlays which you need some measure of an accurate line to identify. Overlays typically don't win a very high percentage of races (unfortunately).

raybo
09-18-2013, 05:06 PM
I gave my definition of stochastic as applied to horse racing:

"Stochasticity in horse racing prediction is due to uncertainity in the handicapping process and data analysis and not due to physical characteristics of the race. The physical characteristics of the race are chaotic and not random."

Do you disagree?
Mike

I might agree if you explained what "uncertainty in the handicapping process and data analysis" means. That quoted phrase, to me, means that the handicapper is uncertain, not the process. In a legitimate systematic process, there is no uncertainty, the uncertainty is in whether or not the race will run as it should. We all know that races do not always run the way we have predicted, that is why they are called "predictions" and not "guarantees".

TrifectaMike
09-18-2013, 07:19 PM
Administrator, why was my last post deleted?

Mike

TrifectaMike
09-18-2013, 07:29 PM
I might agree if you explained what "uncertainty in the handicapping process and data analysis" means. That quoted phrase, to me, means that the handicapper is uncertain, not the process. In a legitimate systematic process, there is no uncertainty, the uncertainty is in whether or not the race will run as it should. We all know that races do not always run the way we have predicted, that is why they are called "predictions" and not "guarantees".
I'll let this post stand without a response.

Mike

PaceAdvantage
09-19-2013, 12:36 AM
I'll let this post stand without a response.

MikeI'll take a stab as to why it was deleted (you recreated the deleted post here I see).

Ummmmmm...it's totally redundant? The above simply could have been achieved by not replying at all.

Plus, it's a tad self-serving, don't you think? :lol:

TrifectaMike
09-19-2013, 06:50 AM
I'll take a stab as to why it was deleted (you recreated the deleted post here I see).

Ummmmmm...it's totally redundant? The above simply could have been achieved by not replying at all.

Plus, it's a tad self-serving, don't you think? :lol:

So, my remark was censored because I breached a redundancy rule or did you decide to censor the post, because it was a TM post?

Let's see. My post:

"I'll let this post stand without a response."

is censored, but this post:

"In other words, if your goal is to impress others with your vocabulary, then use the 3 terms mentioned here, but if your goal is to help others, then describe the approach instead of simply using a single word. But, then maybe some small, highly educated group of people prefer to try to impress only the other members of that small group. This appears to be the case in these instances, IMO. I often wonder why the members of that small group don't just cut to the chase, instead of continually running around in circles, creating subterfuge and smoke. But then, how else would they justify all those years of pursuing higher education?"

is just fine.

Hmm...

Mike

Magister Ludi
09-19-2013, 07:23 AM
So, my remark was censored because I breached a redundancy rule or did you decide to censor the post, because it was a TM post?

Let's see. My post:

"I'll let this post stand without a response."

is censored, but this post:

"In other words, if your goal is to impress others with your vocabulary, then use the 3 terms mentioned here, but if your goal is to help others, then describe the approach instead of simply using a single word. But, then maybe some small, highly educated group of people prefer to try to impress only the other members of that small group. This appears to be the case in these instances, IMO. I often wonder why the members of that small group don't just cut to the chase, instead of continually running around in circles, creating subterfuge and smoke. But then, how else would they justify all those years of pursuing higher education?"

is just fine.

Hmm...

Mike

I was amazed when Mr. Raybo's post to which Trifecta Mike refers was not deleted, or at least censured by the administrator. I started to frame a response three times to said post. However, I thought better of it since I didn't want to stoop to the same ad hominem arguments to which Mr. Raybo had fallen. Trifecta Mike is one of the precious few posters on this site who do "cut to the chase" with intelligent and logical arguments couched in consistant and precise terminology. Perhaps those who dismiss his and similar missives as intellectual elitism should spend some years and considerable dollars "pursuing higher education".

I'll leave all of you now, not that anyone will care, to your "handicapping" social club, where "subterfuge and smoke" is the order of the day.

PaceAdvantage
09-19-2013, 07:43 AM
So, my remark was censored because I breached a redundancy rule or did you decide to censor the post, because it was a TM post?

Let's see. My post:

"I'll let this post stand without a response."

is censored, but this post:

"In other words, if your goal is to impress others with your vocabulary, then use the 3 terms mentioned here, but if your goal is to help others, then describe the approach instead of simply using a single word. But, then maybe some small, highly educated group of people prefer to try to impress only the other members of that small group. This appears to be the case in these instances, IMO. I often wonder why the members of that small group don't just cut to the chase, instead of continually running around in circles, creating subterfuge and smoke. But then, how else would they justify all those years of pursuing higher education?"

is just fine.

Hmm...

Mike
First off, you're wrongly assuming that I deleted your post. I did not. There are other moderators here.

Second, your repost is still standing, is it not?

I suggest you and Magister Ludi ease up a bit...it's not the end of the world, there is no club, and life goes on here...

PaceAdvantage
09-19-2013, 07:49 AM
You know, it always amazes me how some guys on here can have the fortitude to beat the game, or become master statisticians or master complex theories, yet one little wrong move on a message board by an admin sends them fleeing for the exits.

It's actually quite interesting to see time and again how some people can't take the littlest of things....

cj
09-19-2013, 08:57 AM
The above simply could have been achieved by not replying at all.

Plus, it's a tad self-serving, don't you think? :lol:

Exactly, that is why it was deleted.

burnsy
09-19-2013, 09:23 AM
You know, it always amazes me how some guys on here can have the fortitude to beat the game, or become master statisticians or master complex theories, yet one little wrong move on a message board by an admin sends them fleeing for the exits.

It's actually quite interesting to see time and again how some people can't take the littlest of things....

C'mon. I get a kick out of reading this stuff. Not that i would use it betting in a million years. I might as well be at work. Spread sheets, vocabulary and stats.......Gee, why even go to the racetrack at all or gamble for that matter?

I'll step in with my "street" track edumacation.........:lol: . Thats how i apply my real education to gambling. Of course recency matters, especially when you compare it to the price they are offering. Thats all you have to know. What am i risking??? and what will they pay me if this horse wins (or loses)? If the horse has not been racing regularly...you better be getting paid......really simple, not rocket science or research material. Third off the layoff...you know you are not going to get what you want (price wise) but the horse is primed to run. I've been doing this for so long i don't need proof...it works.

My next book.....as per the "book" thread. How to Read a Racing Form/Simple Track Street Smarts. It will explain how to handicap without a PhD, endless programs and buying crap from other people. Everyone bitches about the take out, how much time, effort and money do they spend on this garbage and products? I can look at most races and come up with contenders in about 5 minutes. Then its all about what they are paying....really, really simple. In case you've never seen my selections, they are pretty good. This should be fun folks, i guess if you are a geek...this garbage is fun to you. I guess my approach is different but it has deadly results and i can chill out, drink a beer and have fun at the same time. Good luck, but i miss the entire point when i can do well the way i do it. I got a headache reading alot of this but it was funny. "Self serving" is my middle name..if it ain't fun, i'm not buying...:)

traynor
09-19-2013, 09:56 AM
C'mon. I get a kick out of reading this stuff. Not that i would use it betting in a million years. I might as well be at work. Spread sheets, vocabulary and stats.......Gee, why even go to the racetrack at all or gamble for that matter?

I'll step in with my "street" track edumacation.........:lol: . Thats how i apply my real education to gambling. Of course recency matters, especially when you compare it to the price they are offering. Thats all you have to know. What am i risking??? and what will they pay me if this horse wins (or loses)? If the horse has not been racing regularly...you better be getting paid......really simple, not rocket science or research material. Third off the layoff...you know you are not going to get what you want (price wise) but the horse is primed to run. I've been doing this for so long i don't need proof...it works.

My next book.....as per the "book" thread. How to Read a Racing Form/Simple Track Street Smarts. It will explain how to handicap without a PhD, endless programs and buying crap from other people. Everyone bitches about the take out, how much time, effort and money do they spend on this garbage and products? I can look at most races and come up with contenders in about 5 minutes. Then its all about what they are paying....really, really simple. In case you've never seen my selections, they are pretty good. This should be fun folks, i guess if you are a geek...this garbage is fun to you. I guess my approach is different but it has deadly results and i can chill out, drink a beer and have fun at the same time. Good luck, but i miss the entire point when i can do well the way i do it. I got a headache reading alot of this but it was funny. "Self serving" is my middle name..if it ain't fun, i'm not buying...:)

"Pretty good" and "fun" are not that difficult to accomplish. Unfortunately, it is going a little beyond that point--to the area of consistent profit--that takes all the research, study, and effort.

I think it would be useful to take your approach a step further--to making it explicit, and then teaching it to others. One of the things that seems to elude most people is that the transition from "almost-even recreational bettor" to "consistently profitable bettor" is almost impossible. That is why the overwhelming majority of bettors bet modest amounts, and write off losses as "entertainment expenses."

I think everyone pretty much understands that. Few would argue that drinking beer, betting a few bucks, winning some, losing some is all great entertainment. Fewer still would believe that with a "little improvement" their approach(es) to handicapping horse races could be profitable. Those who believe so are highly motivated to emphasize the recreational aspects of wagering, and to avoid the (possible) profit-making aspects. It is simple human nature.

burnsy
09-19-2013, 10:40 AM
"Pretty good" and "fun" are not that difficult to accomplish. Unfortunately, it is going a little beyond that point--to the area of consistent profit--that takes all the research, study, and effort.

I think it would be useful to take your approach a step further--to making it explicit, and then teaching it to others. One of the things that seems to elude most people is that the transition from "almost-even recreational bettor" to "consistently profitable bettor" is almost impossible. That is why the overwhelming majority of bettors bet modest amounts, and write off losses as "entertainment expenses."

I think everyone pretty much understands that. Few would argue that drinking beer, betting a few bucks, winning some, losing some is all great entertainment. Fewer still would believe that with a "little improvement" their approach(es) to handicapping horse races could be profitable. Those who believe so are highly motivated to emphasize the recreational aspects of wagering, and to avoid the (possible) profit-making aspects. It is simple human nature.

Yeah, I am so far ahead this year........it will be profitable no matter what happens from here to Jan 1. Pretty good, means i am ahead, not behind. My argument is that math and numbers is only part of winning but having the track sense to know risk (chance of winning) vs. price is everything when you actually place that bet. You think i don't study and reasearch when i'm reading a racing form? It took years for me to learn how to "feel out" a race. Plus, i only bet certain tracks that i know very well.

I'm 50 years old, worked hard and raised a family for 25 years. Got divorced 13 years ago. Been betting for 35 years. If i didn't know what i was doing, i would not be betting horses at this age. I'd be broke by now. I don't bet every race or all the time. I play in spots. Thats the key, thats why i can drink and have fun. I bet when i see my advantage and not until then. I'm not compulsive or in the need for "action". I only work about 15 - 20 hours a week now..........betting horses and football are my other "occupations". So if you think i'm talking about "entertainment" and i am not serious about winning...you have not read my posts or seen my selections.......I don't need "expert advice" or "sheets" or "programs". I know how to handicap, if i didn't, i would of had to quit years ago lad. I'm not saying all these people are losing....but its not for me and just from being around the block and in the streets gambling for so long. Its too much work for me when i can read in between the lines for myself without complicated programs or buying products. I'll put my skills against anyone while i'm drinking a beer and well, that part is censored.....at this point in my life, if i'm not having fun while i am working....why get out of bed in the morning? This is what the numbers crunchers will never understand and believe me i can crunch numbers if need be. A good handicapper in ANYTHING can master their skills like an artist or a craftsman by being focused, unbiased and in control of their bankroll. I'm not saying the other way can't work....but if you think all gamblers that have success operate like geeks on the The Big Bang Theory...you are nuts.

traynor
09-19-2013, 11:29 AM
Yeah, I am so far ahead this year........it will be profitable no matter what happens from here to Jan 1. Pretty good, means i am ahead, not behind. My argument is that math and numbers is only part of winning but having the track sense to know risk (chance of winning) vs. price is everything when you actually place that bet. You think i don't study and reasearch when i'm reading a racing form? It took years for me to learn how to "feel out" a race. Plus, i only bet certain tracks that i know very well.

I'm 50 years old, worked hard and raised a family for 25 years. Got divorced 13 years ago. Been betting for 35 years. If i didn't know what i was doing, i would not be betting horses at this age. I'd be broke by now. I don't bet every race or all the time. I play in spots. Thats the key, thats why i can drink and have fun. I bet when i see my advantage and not until then. I'm not compulsive or in the need for "action". I only work about 15 - 20 hours a week now..........betting horses and football are my other "occupations". So if you think i'm talking about "entertainment" and i am not serious about winning...you have not read my posts or seen my selections.......I don't need "expert advice" or "sheets" or "programs". I know how to handicap, if i didn't, i would of had to quit years ago lad. I'm not saying all these people are losing....but its not for me and just from being around the block and in the streets gambling for so long. Its too much work for me when i can read in between the lines for myself without complicated programs or buying products. I'll put my skills against anyone while i'm drinking a beer and well, that part is censored.....at this point in my life, if i'm not having fun while i am working....why get out of bed in the morning? This is what the numbers crunchers will never understand and believe me i can crunch numbers if need be. A good handicapper in ANYTHING can master their skills like an artist or a craftsman by being focused, unbiased and in control of their bankroll. I'm not saying the other way can't work....but if you think all gamblers that have success operate like geeks on the The Big Bang Theory...you are nuts.

Have you ever heard of the Texas Sharpshooter Fallacy? It is the base of the saying, "Shoot first. Whatever you hit, call it the target."

Everyone likes to believe their way is the only way. The bottom line is that those who emphasize the recreational aspects of wagering tend to be either losers, or to only win trivial amounts when they win. There is nothing wrong with that, and I do not for a moment denigrate the activities of recreational bettors.

However, there is quite a difference between those who bet primarily for fun, and those who bet primarily for profit.

Tom
09-19-2013, 11:38 AM
And don't forget those who only talk about betting.

raybo
09-19-2013, 12:13 PM
And don't forget those who only talk about betting.

Yeah, it's amazing that some can talk all this amazing theory and yet, never post a pick or an analysis, and then play it off as "I have no interest in doing so". Theory and "book learning" is one thing, actually showing the theory in practice is another.

traynor
09-19-2013, 12:27 PM
The Texas sharpshooter fallacy is an informal fallacy in which pieces of information that have no relationship to one another are called out for their similarities, and that similarity is used for claiming the existence of a pattern This fallacy is the philosophical/rhetorical application of the multiple comparisons problem (in statistics) and apophenia (in cognitive psychology). It is related to the clustering illusion, which refers to the tendency in human cognition to interpret patterns where none actually exist.

The name comes from a joke about a Texan who fires some shots at the side of a barn, then paints a target centered on the biggest cluster of hits and claims to be a sharpshooter.
http://en.wikipedia.org/wiki/Texas_sharpshooter_fallacy

mountainman
09-19-2013, 02:19 PM
Lots of winners on the internet. And they use such great words.

thaskalos
09-19-2013, 02:35 PM
It's getting pretty hard to find a losing horseplayer on the internet...

traynor
09-19-2013, 04:57 PM
More should read:
http://psych.colorado.edu/~vanboven/teaching/p7536_heurbias/p7536_readings/kruger_dunning.pdf

and take the advice to heart. Those least able to recognize their own deficiencies are those most likely to believe they see deficiencies in others. It is pointless, as well as boring.

PaceAdvantage
09-19-2013, 06:57 PM
C'mon. I get a kick out of reading this stuff. Not that i would use it betting in a million years. I might as well be at work. Spread sheets, vocabulary and stats.......Gee, why even go to the racetrack at all or gamble for that matter?

I'll step in with my "street" track edumacation.........:lol: . Thats how i apply my real education to gambling. Of course recency matters, especially when you compare it to the price they are offering. Thats all you have to know. What am i risking??? and what will they pay me if this horse wins (or loses)? If the horse has not been racing regularly...you better be getting paid......really simple, not rocket science or research material. Third off the layoff...you know you are not going to get what you want (price wise) but the horse is primed to run. I've been doing this for so long i don't need proof...it works.

My next book.....as per the "book" thread. How to Read a Racing Form/Simple Track Street Smarts. It will explain how to handicap without a PhD, endless programs and buying crap from other people. Everyone bitches about the take out, how much time, effort and money do they spend on this garbage and products? I can look at most races and come up with contenders in about 5 minutes. Then its all about what they are paying....really, really simple. In case you've never seen my selections, they are pretty good. This should be fun folks, i guess if you are a geek...this garbage is fun to you. I guess my approach is different but it has deadly results and i can chill out, drink a beer and have fun at the same time. Good luck, but i miss the entire point when i can do well the way i do it. I got a headache reading alot of this but it was funny. "Self serving" is my middle name..if it ain't fun, i'm not buying...:)Nice mini-rant...too bad it pretty much had nothing to do with what I wrote... :lol:

wiffleball whizz
09-19-2013, 07:05 PM
It's getting pretty hard to find a losing horseplayer on the internet...


So true......nobody has the balls to come on here and say they got banged for $1200 over the weekend or they made a big play and lost....

Also impossible to find losing poker players on the Internet too!!!!

On a different note the owner of this site is tooting his own horn about that horse in the 7th at Belmont.....what's wrong with a pm to the whizz saying hey I like a horse. :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol:

Getting a horse from people around here is like giving salt in a salt mine

DeltaLover
09-19-2013, 07:39 PM
I am really surprised by the negativity expressed in this and in other similar threads. I do not think that anyone claimed that he is a winner nor that he posses the absolute handicapping truth, at least not in this thread so I find comments like 'Lots of winners on the internet' to not be a good fit for our discussion.

The fact that opinions differ so greatly, is one of the characteristics of the game and the main reason for its existence. Personally, I have absolutely no intention of changing anyone's opinion especially when he appears to be negative and stubborn about his approaches and not willing to listen and comprehend. I am coming to the realization that this game cultivates very strong egos that are inflated by the lack of determination for analytical thinking and hard work.

For example, Trifecta Mike made a great suggestion about Hierachical Bayes Normal-Normal Model, a very interesting and fascinating topic that might present solutions to quite a few of the issues associated with horse betting. Instead of becoming more curious and interesting about the topic, I can detect some hostility from the posters, that I do not believe is constructive for our improvement and evolution as bettors.

Why it is so difficult for some of us to accept the fact that some others know more than us and we should try to learn from them instead of pretending that we know better?

I am perfectly fine with someone who is convinced that data mining is useless and solely relies to his memory and intuition to make decisions; to be honest this is the profile of the betting opponent where I am going to make a profit from.

What I do not understand though, is how he can respond to an opinion that is based in cold data with an aphorism, like 'Of course recency matters' without providing any kind of evidence or though process.

mountainman
09-19-2013, 08:00 PM
to be honest this is the profile of the betting opponent where I am going to make a profit from.



And this bravado informs the discussion?

TheEdge07
09-19-2013, 08:04 PM
So true......nobody has the balls to come on here and say they got banged for $1200 over the weekend or they made a big play and lost....

Also impossible to find losing poker players on the Internet too!!!!

On a different note the owner of this site is tooting his own horn about that horse in the 7th at Belmont.....what's wrong with a pm to the whizz saying hey I like a horse. :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol:

Getting a horse from people around here is like giving salt in a salt mine

Never bring sand to the beach.

barn32
09-19-2013, 08:53 PM
Back in the 70s and 80s I made some good coin betting on horses at cheap tracks that were running three days back.

traynor
09-19-2013, 09:01 PM
I am really surprised by the negativity expressed in this and in other similar threads. I do not think that anyone claimed that he is a winner nor that he posses the absolute handicapping truth, at least not in this thread so I find comments like 'Lots of winners on the internet' to not be a good fit for our discussion.

The fact that opinions differ so greatly, is one of the characteristics of the game and the main reason for its existence. Personally, I have absolutely no intention of changing anyone's opinion especially when he appears to be negative and stubborn about his approaches and not willing to listen and comprehend. I am coming to the realization that this game cultivates very strong egos that are inflated by the lack of determination for analytical thinking and hard work.

For example, Trifecta Mike made a great suggestion about Hierachical Bayes Normal-Normal Model, a very interesting and fascinating topic that might present solutions to quite a few of the issues associated with horse betting. Instead of becoming more curious and interesting about the topic, I can detect some hostility from the posters, that I do not believe is constructive for our improvement and evolution as bettors.

Why it is so difficult for some of us to accept the fact that some others know more than us and we should try to learn from them instead of pretending that we know better?

I am perfectly fine with someone who is convinced that data mining is useless and solely relies to his memory and intuition to make decisions; to be honest this is the profile of the betting opponent where I am going to make a profit from.

What I do not understand though, is how he can respond to an opinion that is based in cold data with an aphorism, like 'Of course recency matters' without providing any kind of evidence or though process.

That is at the core of what causes people who lose to continue losing--while strongly advocating that everyone else should do it "their way." The only thing more foolish that I can imagine offhand is someone accepting those views as "how things work in the real world." It seems some of those least capable of recognizing "reality" are those most vocal about how "reality really works."

However, as you point out, that is the opposition in betting. If one chooses to believe that statistics don't apply in the "real world," or that backfitting algorithms to last week's or last month's races is "the way its done in the real world"--I congratulate them, and hope they never, ever learn how foolish such notions are.

Does recency matter? Practically, rather than philosophically, I have no definitive answer. Entries with good recent races tend to be overbet. Attempts to rate entries based on performances in older races may not be what it seems on the surface. That is, the entry may win or lose because of many other circumstances, and selecting a horse because it ran well at this class/distance whatever six months ago or a year ago may seem like magic (or handicapping brilliance) when it seems to work, but detailed research will indicate other factors (or an absence of other factors) are stronger indicators.

Ignoring pace, speed, and class levels entirely, does recency matter? Only to the extent that it indicates tne animal is at least ambulatory. Some of the best scores I have ever made were on horses that had not raced in the last 6 to 12 months. No bullet works, none of the typical indicators favored by many bettors. So unless one is playing with pace, speed, and class ratings for comparisons, I would have to say that recency could easily be ignored without suffering bankroll depletion as a result.

traynor
09-19-2013, 09:06 PM
Back in the 70s and 80s I made some good coin betting on horses at cheap tracks that were running three days back.

Still works.

dkithore
09-19-2013, 09:26 PM
If it helps I will say, all opinions are needed and because it rings true to some people and that makes this a great site.

To say, if you disagree, "I will take my marbles and go home" is not an adult response, imo. I have high respect for T.M. and M.L. admittedly I do not always understand their views. so what? Some one else does. So, I urge you both to stay and engage.

Thanks.

traynor
09-19-2013, 09:45 PM
If it helps I will say, all opinions are needed and because it rings true to some people and that makes this a great site.

To say, if you disagree, "I will take my marbles and go home" is not an adult response, imo. I have high respect for T.M. and M.L. admittedly I do not always understand their views. so what? Some one else does. So, I urge you both to stay and engage.

Thanks.

It might be useful to understand the notion of "discourse communities." It is not so much trying to be a SA as that if one is accustomed to communicating with people for whom precise language is a mandatory requirement, it is frustrating when people want to make words mean (to them) whatever they want them to mean (or not) depending on the mood of the moment.

I think both TM and ML are accustomed to discourse communities in which statements like "small samples might not work in snob land, but they sure work in lala land" would not only be unacceptable, they would make the speaker (or writer) seem incompetent and unaware of the basics of the topic being discussed.

classhandicapper
09-20-2013, 08:52 AM
The problem with this approach is that it is highly subjective and open to many interpretations. We need a systematic approach providing a valid opinion about the attribute(s) in question. It has to be simple and generic but valid. We should start with very simple models, gradually adding constraints until we reach an acceptable behavior. The objective should be the method and not the specific angles (like blinker changes or pace figures).

IMO we don't necessarily need anything that's objective, subjective, simple, or complex. We don't need something that must handicap every single race in a systematic way or one that uses superior intuition.

We need something that has an edge. It doesn't matter how you get there or whether it works for all races or just for a few a year. It just has to work.

I try to find edges by looking at small subsets of factors that are generally well known and understood at higher levels. People may understand that trainer "X" has a great overall record with first time starters, but they may not understand that he's way better on turf, in routes, for a specific owner.

So while others are betting all of them or just all the turfers etc... I am only looking at the turf routers for a specific owner because that's where the value is.

All that's left is to handicap the rest of the race to see if I think it's a good spot and then see if the price looks right.

When things change, I'll change.

The idea would be to have a portfolio of "potential plays".

But that's what works for me. Others will and should take another path.

DeltaLover
09-20-2013, 08:59 AM
We need something that has an edge. It doesn't matter how you get there or whether it works for all races or just for a few a year. It just has to work.


How you can be sure that it is working? You need some criteria for that. An easy answer might be positive ROI, but keep in mind that this can be deceiving.

Any approach might show some profitability over a specific period of time, to be sure about its long term behavior you need to establish an objective methodology eliminating noise leading to more secure conclusions.

traynor
09-20-2013, 09:09 AM
IMO we don't necessarily need anything that's objective, subjective, simple, or complex. We don't need something that must handicap every single race in a systematic way or one that uses superior intuition.

We need something that has an edge. It doesn't matter how you get there or whether it works for all races or just for a few a year. It just has to work.

I try to find edges by looking at small subsets of factors that are generally well known and understood at higher levels. People may understand that trainer "X" has a great overall record with first time starters, but they may not understand that he's way better on turf, in routes, for a specific owner.

So while others are betting all of them or just all the turfers etc... I am only looking at the turf routers for a specific owner because that's where the value is.

All that's left is to handicap the rest of the race to see if I think it's a good spot and then see if the price looks right.

When things change, I'll change.

The idea would be to have a portfolio of "potential plays".

But that's what works for me. Others will and should take another path.

There is a subtle point here. Some may believe you are simply "layering" into progressively finer detail, until something seems to "have value." That is little more than what others are doing with (very) small samples, with the built-in problem that repetition is unlikely.

However, what you suggest implies cause-and-effect relationships that don't exist in simple associations. It is finding the right subsets--ones that actually exert some kind of influence on the outcomes of races--that makes your approach different. And better.

Tom
09-20-2013, 09:14 AM
What other criteria is there beside ROI?
Your process produces money.
You continually tweak it and evolve it and your ROI tells you where your stand.

This is not a long term game - it is a short term activity. No one remains static enough in their approach to get to the long run.
It is not the process that matters, it is the results.

classhandicapper
09-20-2013, 09:18 AM
How you can be sure that it is working? You need some criteria for that. An easy answer might be positive ROI, but keep in mind that this can be deceiving.

Any approach might show some profitability over a specific period of time, to be sure about its long term behavior you need to establish an objective methodology eliminating noise leading to more secure conclusions.

In a case like this, I'd look at the win%, sample size, ROI, intuitively observe the betting patterns on the horses as time passes to see if others were catching on and the prices were falling, make sure there wasn't a single big winner distorting the ROI, and look at all the details of the horses to make sure the one I'm considering is as good as the ones in the sample.

This is the process for identifying a "prospect" for a bet.

I still handicap the race in a more fundamental way to determine if this is a good spot and the price is right.

Sometimes I find these gems without stats or data. I'll simply make an observation because there happened to be a cluster of examples that were similar and it happened to register in my head. I may not have long term data and it may not even exist. So I'll tip toe into playing some of them with small bets, refine my thinking, and see how I'm doing. I do go up blind alleys sometimes. Then I stop.

I would have loved to have a huge database to assist in the research at times, but when I started in this game, it was all memory, and paper and pencil. Much I what I know started as trial and error.

If I was forced to handicap every race in public (like Andy Serling for example), I'd pick a lot of winners on fundamentals, but I don't think I'd do very well or be overly impressive because there are still way too many situations I don't understand well enough to have an edge.

I allowed the flowers to grow and pulled out the weeds until I eventually had an edge. I love that saying. ;) The process continues.

DeltaLover
09-20-2013, 10:11 AM
What other criteria is there beside ROI?
Your process produces money.
You continually tweak it and evolve it and your ROI tells you where your stand.

This is not a long term game - it is a short term activity. No one remains static enough in their approach to get to the long run.
It is not the process that matters, it is the results.

Even a negative EV strategy will have cases where it appears to be profitable and a positive EV might appear to be loosing. Of course this happens due the variance of the strategy.

To make my case lets assume a strategy having an EV per a block of 10 races is $10 while your standard deviation is $100. If you bet 1000 races (100 sessions of 10)then your expected profit will be 10*100 = $1000. Chances are that this will not happen this way. The standard deviation of your sample (100 session) will be inversely proportional to the square root of its size. For our example, the stdev will be: 100 / 10 = 10 (the denominator is the square root of the 100).

Assuming the results are contained within 3 sigmas, this means that your EV for the sample of the 100 sessions of 10 races each can have an EV rangining from -20 to +40. (-$20 = $10 - 3 * 10 $40 = $10 + 3 * 10)

So it is quite possible to be loosing 20 * 100 = $2,000 after 1000 races and still have a positive expectation! The same example can be given for the reverse case of a minus EV that can show positive ROI after a large sequence of bets.

cj
09-20-2013, 10:50 AM
Let me ask a couple questions to think about.

1. If a horse's recent running lines are bad, would you prefer he raced recently or was coming off of a break?

2. If a horse's recent running lines are good, would you prefer he raced recently or was coming off a break?

Seems to me people underbet horses coming off a layoff that went bad before it, and overbet horses that were going well and were given time off when they return.

Tom
09-20-2013, 11:21 AM
Case in point - check out the dogwood Saturday at CD.
The bottom horse, used to be trained by Pletcher, good form going into a layoff, then a noticeable lack of workouts.

I forget the name, but I was looking at that one last night.

Who Bet has the PPs.

What is the call on this one?

DeltaLover
09-20-2013, 11:59 AM
Let me ask a couple questions to think about.

1. If a horse's recent running lines are bad, would you prefer he raced recently or was coming off of a break?

2. If a horse's recent running lines are good, would you prefer he raced recently or was coming off a break?

Seems to me people underbet horses coming off a layoff that went bad before it, and overbet horses that were going well and were given time off when they return.

I will run the simulation later tonight and post the results

raybo
09-20-2013, 12:06 PM
Let me ask a couple questions to think about.

1. If a horse's recent running lines are bad, would you prefer he raced recently or was coming off of a break?

2. If a horse's recent running lines are good, would you prefer he raced recently or was coming off a break?

Seems to me people underbet horses coming off a layoff that went bad before it, and overbet horses that were going well and were given time off when they return.

CJ, of course, the answers to those 2 questions are obvious, however, you know it's not that simple.

raybo
09-20-2013, 12:10 PM
I will run the simulation later tonight and post the results

Your sample will not be valid unless you add many other factors to the equation. Is the horse coming off a layoff simply in the race for post-layoff conditioning? Is the horse who has been running well going to continue running well or did his last race take too much out of him? There are other things that must be considered.

DeltaLover
09-20-2013, 12:15 PM
Is the horse coming off a layoff simply in the race for post-layoff conditioning?

Is the horse who has been running well going to continue running well or did his last race take too much out of him?

Sorry.

These are not factors that can be determined by the data. These are just rhetoric questions that cannot be answered. A factor is a scenario that can be descibed by a decission tree.

mountainman
09-20-2013, 12:16 PM
Let me ask a couple questions to think about.

1. If a horse's recent running lines are bad, would you prefer he raced recently or was coming off of a break?

2. If a horse's recent running lines are good, would you prefer he raced recently or was coming off a break?

Seems to me people underbet horses coming off a layoff that went bad before it, and overbet horses that were going well and were given time off when they return.

Agree completely. Even when a horse's last race was months ago, and potentially irrelevant, the performance is given considerable weight by the public. Consequently, an ugly line should afford more value than a competitive effort. Rested horses heal, and can reverse form.

In my opinion, it's more ominous when a runner in good form is sidelined. Horsemen are reluctant to shut down a thoroughbred making money. Thus, when they do, it's a hint something has gone seriously wrong.

Take the start of a spring meet, for instance, when lots of horses haven't started since late fall. I've found it good policy to wager mainly on runners showing poor performances at the top of their pp's.

raybo
09-20-2013, 01:01 PM
Sorry.

These are not factors that can be determined by the data. These are just rhetoric questions that cannot be answered. A factor is a scenario that can be descibed by a decission tree.

So, are you saying that you don't have decision trees built into your method that answers those questions, long term? Decision trees are great things, if they are comprehensive enough.

Tom
09-20-2013, 01:26 PM
In the Dogwood case, you have a layoff, a lack works in the middle, a trainer change from a top guy....how many times has that happened that you can use as a reference?

If the horse is 5-2, do you use her?
If she is 10-1 do you use her?

Show Me the Wire
09-20-2013, 05:20 PM
In my opinion, it's more ominous when a runner in good form is sidelined. Horsemen are reluctant to shut down a thoroughbred making money. Thus, when they do, it's a hint something has gone seriously wrong.

Unless there is no race in the condition book, or the races didn't fill, or not wanting to chance losing the horse in an o/c, etc.

But I do understand what you are saying and not necessarily disagreeing with you, just trying to point out a variety of reasons exist, which can not be accounted or discounted through dbs or pps.

MJC922
09-20-2013, 05:32 PM
Case in point - check out the dogwood Saturday at CD.
The bottom horse, used to be trained by Pletcher, good form going into a layoff, then a noticeable lack of workouts.

I forget the name, but I was looking at that one last night.

Who Bet has the PPs.

What is the call on this one?

On generic recency alone I wouldn't be overly concerned. A 3yo filly in a graded stake off for 12 weeks with spacings of 8 and 9 weeks prior to that doesn't worry me in the least, doesn't mean I'm right, obviously others will differ maybe due to workouts, the game is rich in complexity. Trainer change is another issue altogether. 25% trainer overall, how bad can he be? I never seem to win with Court up though.

MJC922
09-20-2013, 06:16 PM
Sorry.

These are not factors that can be determined by the data. These are just rhetoric questions that cannot be answered. A factor is a scenario that can be descibed by a decission tree.

They are not easy to determine but they can prove to be relevant at times because they can place the recency factor into context. I will give you an example of a spacing and effort sequence:

CLM_4UP_W2-B-M1-F

We have an older claiming horse showing a poor effort 'F' two races ago. Trainer rests the horse for 1 month and the horse responds well to the freshening and posts a 'B' grade effort. Now today he's back in on two weeks rest. This angle is not rocket science but has a profitable ROI after 999 occurrances. Makes perfect sense. So the problem with studies that don't dig very deep is there's a million variations on that W2 spacing, some are better than others. I'm not saying studies like this are in any way adequate. I don't play these things myself but it's interesting to look at what seems to be undervalued, understand it, and let it seep into the thought processes.

DeltaLover
09-20-2013, 10:03 PM
Let me ask a couple questions to think about.

1. If a horse's recent running lines are bad, would you prefer he raced recently or was coming off of a break?

2. If a horse's recent running lines are good, would you prefer he raced recently or was coming off a break?

Seems to me people underbet horses coming off a layoff that went bad before it, and overbet horses that were going well and were given time off when they return.


The results are the following:



layoff_and_recent_races_are_bad
MATCHES : observed winners: 153.00 expected winners: 142.45
NOMATCHES: observed winners: 1738.00 expected winners: 1748.55
x2: 0.76653
NOT significant
========================================
layoff_and_recent_races_are_good
MATCHES : observed winners: 450.00 expected winners: 482.90
NOMATCHES: observed winners: 1749.00 expected winners: 1716.10
x2: 2.78640
NOT significant
========================================
long_layoff_and_recent_races_are_bad
MATCHES : observed winners: 61.00 expected winners: 64.13
NOMATCHES: observed winners: 981.00 expected winners: 977.87
x2: 0.11529
NOT significant
========================================
long_layoff_and_recent_races_are_good
MATCHES : observed winners: 142.00 expected winners: 151.26
NOMATCHES: observed winners: 709.00 expected winners: 699.74
x2: 0.61757
NOT significant
========================================
second_of_layoff_and_recent_races_are_bad
MATCHES : observed winners: 149.00 expected winners: 166.98
NOMATCHES: observed winners: 2262.00 expected winners: 2244.02
x2: 1.96558
NOT significant
========================================
second_of_layoff_and_recent_races_are_good
MATCHES : observed winners: 465.00 expected winners: 454.26
NOMATCHES: observed winners: 1547.00 expected winners: 1557.74
x2: 0.29793
NOT significant
========================================
third_of_layoff_and_recent_races_are_bad
MATCHES : observed winners: 123.00 expected winners: 111.82
NOMATCHES: observed winners: 1646.00 expected winners: 1657.18
x2: 1.08860
NOT significant
========================================
third_of_layoff_and_recent_races_are_good
MATCHES : observed winners: 480.00 expected winners: 463.96
NOMATCHES: observed winners: 1613.00 expected winners: 1629.04
x2: 0.66871
NOT significant
========================================
deep_form_cycle_recent_races_are_bad
MATCHES : observed winners: 345.00 expected winners: 376.68
NOMATCHES: observed winners: 4677.00 expected winners: 4645.32
x2: 2.79079
NOT significant
========================================
deep_form_cycle_and_recent_races_are_good
MATCHES : observed winners: 2046.00 expected winners: 2012.11
NOMATCHES: observed winners: 5306.00 expected winners: 5339.89
x2: 0.76276
NOT significant
========================================

DeltaLover
09-20-2013, 10:33 PM
So, are you saying that you don't have decision trees built into your method that answers those questions, long term? Decision trees are great things, if they are comprehensive enough.


Any handicapping factor can be considered a decision tree, this is not the point I am trying to make with your statements:

- Is the horse coming off a layoff simply in the race for post-layoff conditioning?

- Is the horse who has been running well going to continue running well or did his last race take too much out of him?

Both of them are asking for an opinion rather than a fact that can be determined based on the past performances. The problem with both of them is simply that they cannot be validated after the race.

A question like:

- What is the probability that this horse will be leading the race at the second call?

makes sense, since we can just check the chart of the race and create a fitness function


while:

- What is the probability that this horse is running only for preconditioning?

represents a question open to interpretation, since we do not have a concrete scenario to verify it. Even if you use the fraction calls you can never be sure why the horse was never a factor, it can be due to preconditioning, to bleeding or other injury, track condition, bad start or anything else.

mountainman
09-20-2013, 10:34 PM
Unless there is no race in the condition book, or the races didn't fill, or not wanting to chance losing the horse in an o/c, etc.

But I do understand what you are saying and not necessarily disagreeing with you, just trying to point out a variety of reasons exist, which can not be accounted or discounted through dbs or pps.

Of course. In fact, I regularly advise viewers to obtain a condition book, collect overnites and do the gumshoe work before concluding anything about readiness or trainer intent. And REALLY serious handicappers should familiarize themselves with various preference systems used by different racing offices. In addition, UPCOMING conditions that a horse wasn't yet eligible for can be cause to excuse a horse's absence-or even interpret it as a positive.

As with most factors, this tips from science to art at some point. No computer or database can entirely teach you how trainers think. That's a matter not only of doing your homework, but of local knowledge and lots of experience.

DeltaLover
09-20-2013, 10:43 PM
They are not easy to determine but they can prove to be relevant at times because they can place the recency factor into context. I will give you an example of a spacing and effort sequence:

CLM_4UP_W2-B-M1-F

We have an older claiming horse showing a poor effort 'F' two races ago. Trainer rests the horse for 1 month and the horse responds well to the freshening and posts a 'B' grade effort. Now today he's back in on two weeks rest. This angle is not rocket science but has a profitable ROI after 999 occurrances. Makes perfect sense. So the problem with studies that don't dig very deep is there's a million variations on that W2 spacing, some are better than others. I'm not saying studies like this are in any way adequate. I don't play these things myself but it's interesting to look at what seems to be undervalued, understand it, and let it seep into the thought processes.


How do you assign the grade of the effort (like 'F' or 'B')? Do you only consider fraction calls or you also use class movements, distance changes, weight, or any other change? Can you describe the model that shows a profitable ROI after 999 occurrences?

Without considering the race schema I am reluctant to accept that a single handicapping factor can show profitability.

I know that some of the posters here like Traynor or Pondman have already expressed similar views, meaning to not consider the other starters of the race but only concentrate to the profile of each one individually, but I cannot discover any factor having the ability to be profitable by its own.

raybo
09-20-2013, 11:06 PM
Any handicapping factor can be considered a decision tree, this is not the point I am trying to make with your statements:

- Is the horse coming off a layoff simply in the race for post-layoff conditioning?

- Is the horse who has been running well going to continue running well or did his last race take too much out of him?

Both of them are asking for an opinion rather than a fact that can be determined based on the past performances. The problem with both of them is simply that they cannot be validated after the race.

A question like:

- What is the probability that this horse will be leading the race at the second call?

makes sense, since we can just check the chart of the race and create a fitness function


while:

- What is the probability that this horse is running only for preconditioning?

represents a question open to interpretation, since we do not have a concrete scenario to verify it. Even if you use the fraction calls you can never be sure why the horse was never a factor, it can be due to preconditioning, to bleeding or other injury, track condition, bad start or anything else.

Such subjective decisions may indeed be subjective, but they can be tested just like a hard racing data factors. The proof is in the long term. When you use these "subjective" decision trees, does the long term result improve or decline. That's all the proof one needs. It matters not whether your decisions are objective or subjective, as long as they can be recorded and compared, in their totality. I use "subjective" decision tools heavily in my method, and long term they improve the results. Just because you can't pull that data directly from the race data, doesn't mean it doesn't exist, or can't be used to make viable decisions in individual races.

Tom
09-20-2013, 11:17 PM
On generic recency alone I wouldn't be overly concerned. A 3yo filly in a graded stake off for 12 weeks with spacings of 8 and 9 weeks prior to that doesn't worry me in the least, doesn't mean I'm right, obviously others will differ maybe due to workouts, the game is rich in complexity. Trainer change is another issue altogether. 25% trainer overall, how bad can he be? I never seem to win with Court up though.

You got me interested, so I looked it up.

Gr3
Dirt
Sprint
Layoff 80-120 days

45 races 2005-present
11 horses fit the layoff criteria and their record 11 2-2-1 ROI 1.25

cj
09-20-2013, 11:34 PM
The results are the following:




I appreciate you taking the time to look this up, but winners vs expected winners is not what I am interested in unless it is related to the price they pay. The point of my post was that one factor gets overbet, and one gets underbet, not whether they win more or less than expected.

DeltaLover
09-21-2013, 12:05 AM
I appreciate you taking the time to look this up, but winners vs expected winners is not what I am interested in unless it is related to the price they pay. The point of my post was that one factor gets overbet, and one gets underbet, not whether they win more or less than expected.

Absolutely yes.

The expected winners are derived by the odds after a correction for the takeout is applied. In other words I do not really care about a factor's impact as far as winning frequency is going. What I am interested is the margin of error the crowd is making for every factor.

traynor
09-21-2013, 02:42 AM
How do you assign the grade of the effort (like 'F' or 'B')? Do you only consider fraction calls or you also use class movements, distance changes, weight, or any other change? Can you describe the model that shows a profitable ROI after 999 occurrences?

Without considering the race schema I am reluctant to accept that a single handicapping factor can show profitability.

I know that some of the posters here like Traynor or Pondman have already expressed similar views, meaning to not consider the other starters of the race but only concentrate to the profile of each one individually, but I cannot discover any factor having the ability to be profitable by its own.

That is not quite how it goes. I use models, not individual factors--the models are comprised of various combinations of factors. I may have inadvertently created the impression that I was viewing each entry in isolation--somewhat like a "spot play." That is not the case. I use clusters of factors ("models") that are highly specific for each track and for each category of race. However, that is only to determine likely contenders--not final selections.

I use two fundamentally different approaches. Volume wagers are made primarily on the basis of models, but the models are applied to races--not just to individual entries. Specifically, the various factors that go into each model are weighted, and the application evaluates the race using comparisons of ranges of values, rather than simplistic ranking. That process is now almost entirely automated.

For more serious wagers, I use other approaches in combination with the models.

For both types of wagers, every horse in the race is initially considered--factors ("attributes") are good indicators, but only a part of the picture. Most of the factors/attributes I use are only significant (for wagering) when all the other entries are considered. Specifically, each cluster of factors has an overall value that must be considered relative to the presence or absence of other clusters of factors--and/or the same cluster of factors--in the other entries. None can be used in isolation. I apologize if I created that impression.

Data mining to locate factors/attributes and creating profiles or models based on those factors/attributes is only the start of the analytical process.

traynor
09-21-2013, 02:55 AM
One thing I might mention is my prejudice against relying on databases of old races to do more than find potential indicators. I don't accept those indicators as anything more than interesting possibilities until they have been coded into models and used (productively) to analyze current races.

The simple version is that the past results were not known at the time of the events from which they were derived. Expecting repetition of those past results--now that the results of the events are known--may not be overly useful. That is why applying models derived from old races to current and future races rarely lives up to the high hopes and expectations of the models.

The only ROI I consider meaningful is the one based on the races I actually bet on--not wishin' and hopin' from something that happened last month or last year on races I didn't bet on.

raybo
09-21-2013, 03:02 AM
That is not quite how it goes. I use models, not individual factors--the models are comprised of various combinations of factors. I may have inadvertently created the impression that I was viewing each entry in isolation--somewhat like a "spot play." That is not the case. I use clusters of factors ("models") that are highly specific for each track and for each category of race. However, that is only to determine likely contenders--not final selections.

I use two fundamentally different approaches. Volume wagers are made primarily on the basis of models, but the models are applied to races--not just to individual entries. Specifically, the various factors that go into each model are weighted, and the application evaluates the race using comparisons of ranges of values, rather than simplistic ranking. That process is now almost entirely automated.

For more serious wagers, I use other approaches in combination with the models.

For both types of wagers, every horse in the race is initially considered--factors ("attributes") are good indicators, but only a part of the picture. Most of the factors/attributes I use are only significant (for wagering) when all the other entries are considered. Specifically, each cluster of factors has an overall value that must be considered relative to the presence or absence of other clusters of factors--and/or the same cluster of factors--in the other entries. None can be used in isolation. I apologize if I created that impression.

Data mining to locate factors/attributes and creating profiles or models based on those factors/attributes is only the start of the analytical process.

Hmmmm - quite similar to what I do. Quite similar. For win betting, it's fully automated, a true black box. For superfecta betting, it's just the beginning of the process.

raybo
09-21-2013, 03:29 AM
One thing I might mention is my prejudice against relying on databases of old races to do more than find potential indicators. I don't accept those indicators as anything more than interesting possibilities until they have been coded into models and used (productively) to analyze current races.

The simple version is that the past results were not known at the time of the events from which they were derived. Expecting repetition of those past results--now that the results of the events are known--may not be overly useful. That is why applying models derived from old races to current and future races rarely lives up to the high hopes and expectations of the models.

The only ROI I consider meaningful is the one based on the races I actually bet on--not wishin' and hopin' from something that happened last month or last year on races I didn't bet on.

I use past races to test a track for profit potential. I record the preferred track tested methods into current time, if they are producing results similar to those during the testing period then I bet the method. But, I continue tracking those wagered races and retesting all methods, in real time, modifying the method as required. Again all this is fully automated. So, I may begin wagering with one method but change betting methods during the meet. Every method is updated automatically as new races are recorded. I can manually look at each race for each method, knowing which horses were bet, what their odds were, each method's accumulated hit rate, ROI, and profit/loss, at any minimum odds range, up to 11/1 and higher. I can see if one or two hits provided most of the profit (outliers), or if there were no outliers (I can even delete the outliers and evalute the method without them), which tells me if the method is likely to continue profitably into the near future. This also tells me if things are changing at the track, in which case I can check the recent results of all the other methods and switch to one that is handling those changes better.

Capper Al
09-21-2013, 06:11 AM
Hey, Dave has a pretty good video on recency on his site.

MJC922
09-21-2013, 09:22 AM
How do you assign the grade of the effort (like 'F' or 'B')? Do you only consider fraction calls or you also use class movements, distance changes, weight, or any other change? Can you describe the model that shows a profitable ROI after 999 occurrences?

Without considering the race schema I am reluctant to accept that a single handicapping factor can show profitability.

I know that some of the posters here like Traynor or Pondman have already expressed similar views, meaning to not consider the other starters of the race but only concentrate to the profile of each one individually, but I cannot discover any factor having the ability to be profitable by its own.

As stated in prior posts you need a very good measure of a horse's efforts to obtain answers from studies like this. I use my own performance ratings which aren't time-based and which folks out here wouldn't have (well ok they have them if they're subscribed through TM but they won't have the 5 or 6 year denormalized database needed to do a study like this). Some might use Beyer, whatever, might be ok, but obviously GIGO applies, if the numbers people use for this type of study are junk forget it, don't bother going there. What you're calling 'race schema', I'm taking to mean how the pace sets up and the class of the other horses in the race, I agree with you that's far too important to ignore. Studies like this are just another path I've traveled down in a quest to leave no stone unturned. The findings from a study like this (the angles) aren't the basis for a methodology, several dozen do crush the game when bet blindly, unfortunately plays are relatively rare so frankly I don't even look for them. Take the example, 999 plays over 4 or 5 years, what's that 20 per month? Maybe 20% wins if we're lucky and maybe an 8% ROI. Not worth my time. Currently tracking a methodology based upon the same ratings which is nearly getting that many plays per day at 13% ROI.

DeltaLover
09-21-2013, 09:35 AM
As stated in prior posts you need a very good measure of a horse's efforts to obtain answers from studies like this. I use my own performance ratings which aren't time-based and which folks out here wouldn't have (well ok they have them if they're subscribed through TM but they won't have the 5 or 6 year denormalized database needed to do a study like this). Some might use Beyer, whatever, might be ok, but obviously GIGO applies, if the numbers people use for this type of study are junk forget it, don't bother going there. What you're calling 'race schema', I'm taking to mean how the pace sets up and the class of the other horses in the race, I agree with you that's far too important to ignore. Studies like this are just another path I've traveled down in a quest to leave no stone unturned. The findings from a study like this (the angles) aren't the basis for a methodology, several dozen do crush the game when bet blindly, unfortunately plays are relatively rare so frankly I don't even look for them. Take the example, 999 plays over 4 or 5 years, what's that 20 per month? Maybe 20% wins if we're lucky and maybe an 8% ROI. Not worth my time. Currently tracking a methodology based upon the same ratings which is nearly getting that many plays per day at 13% ROI.


- Beyond a certain level of accuracy, the quality of ratings is not critical anymore. We can easily establish a methodology to derive a valid opinion, within a specific degree of confidence that a specific rating is sufficiently enough or it needs more refinement.

- The 'race schema' is not limited to pace considerations. We can use parameters like recency, class movements, workout activity, shippers and pretty much anything to create a race-level rating that can describe the 'race-schema'. It is true that the most significant attribute of a race is its pace schema although there might be cases where other attributes might present more value.

- I completely agree that a low frequency betting strategy is not useful for many reasons. Not only such a strategy leaves little room for profitability, it also always suffer from possible over fitting effects.

MJC922
09-21-2013, 09:59 AM
- Beyond a certain level of accuracy, the quality of ratings is not critical anymore. We can easily establish a methodology to derive a valid opinion, within a specific degree of confidence that a specific rating is sufficiently enough or it needs more refinement.

- The 'race schema' is not limited to pace considerations. We can use parameters like recency, class movements, workout activity, shippers and pretty much anything to create a race-level rating that can describe the 'race-schema'. It is true that the most significant attribute of a race is its pace schema although there might be cases where other attributes might present more value.

- I completely agree that a low frequency betting strategy is not useful for many reasons. Not only such a strategy leaves little room for profitability, it also always suffer from possible over fitting effects.

We'll have to agree to disagree on the importance of quality ratings. The day that's no longer critical is the day I'll know the game is unbeatable. Fortunately the data I have suggests otherwise.

traynor
09-21-2013, 10:10 AM
To be useful for wagering, race schemas should evaluate each component in relation to the others in the race, and enable prediction of the outcome. The relative value of any single component (or factor, or attribute), or of any particular combination, is only that--relative.

It seems more useful to consider the circumstances and constraints involved in how each piece of the model was earned, and how each piece is applied, rather than isolating factors or clusters of factors and applying them to individual entries. That is, the same techiques used to create and analyze pace scenarios can be usefully applied to other criteria.

Simplistically, unless one accounts for the confounding variables (and the effect(s) of those variables on the outcome of the race(s)), one is facing a very steep uphill climb in understanding. Factor A for any one entry can only be meaningfully evaluated (for race wagerihng purposes) when it is studied relative to the presence or absence of Factor(s) Whatever (the confounding variables) in other entries.

That is why I think data mining is useful to establish baselines and areas of further inquiry, but much less so for deciding which entry to bet. Before I bet on any model, I want to see the results of that model applied to races that were still in the future at the time the model was developed. I think that is the area where most database handicappers go astray--over-reliance on past events being replicated in the future.

I don't want The Answer. I want the money.