History of the Methodology

[Tim Y]

Originally Posted by turtleman

Hi Tim
What if the winner is not one of my top contenders, do I still put it in the model?

use the line you chose no matter if the one was a contender or not

[Lt1]

Good question. I don't but others will go back and pick a line and then enter it. I'll go back and look at the horse and see if I missed something. In your case you might want to consider marking an x in each factor to indicate you didn't have it as a contender and if you begin to see too many of these you should review you contender and paceline selection. Nothing gets the winner 100% of the time.
Tim

[Lt1]

Tom I also recommend that you keep 20 race cycle forms to let you see from what tier your winners are coming from. The goal is to have most of your winners in the top 3. Too many in tier 4 & 5 will also indicate a contender and paceline selection review.
Tim

[Bill V.]

Originally Posted by turtleman

Hi Tim
What if the winner is not one of my top contenders, do I still put it in the model?

Tom

Hey Tom

Tim G (LT1) is really giving you A1 help ..

I follow Tim's advice about record keeping
I also keep track of the race class , the actual RDSS notations and the TM RC
When I get a race were the winner is not one of my contenders I enter
NA or X
This is a tool I use when structuring my horizontal bets , doubles P3 P4 P5
Its good info to know which races tend to favor the ML favorites and the races that tend to come up with longer price winners.

Good Skill
Bill

[Tim Y]

Sample error. Two years ago I listened to a lecture on line about statistical analysis and it hit home. The lecturer pointed out that looking first at the BEST most outstanding effort (like Frosted, an known in and outer might get you to select), you are getting a skewed idea of what the animal is capable of since we all know the basic statistical principle of returning to the MEAN. One stand out performance is the OUTLIER, not the norm.

The best actuarials (the principle evaluations that insurance companies use to access liabilities) are based upon HUGE samples. The Theory of Large Numbers requires that a large database be evaluated before the results are representative of real life.

We do not have this luxury so we have to find a way around it.

In the Speculator program, there is an area called the Long Shot detector which evaluated the pace of race of each horse. I found it best for Match Up reasons to find similar paces of race (make the horse's be under similar pace pressures before evaluating them). Find several appropriate lines and use them, by themselves, to see how each horse handled that pace of race, particularly asking the question: DID any of these efforts push the horse out of it NORMAL ESP status? Then go back and you compare the other contenders to these lines.

I would try never to wager on horses that did not show at least two similar efforts coming into today's contest.

[Tim Y]

Further on that point, I discovered, INDEPENDENTLY of several other pace handicappers (i.e. Dave Schwartz calls this 2BE or two best ever), the idea I call Two Lines Better than the Field and wrote about it many years ago.

If you can find TWO pace lines and these point to it being the best this race, the chances of that horse doing very well are high. NO ONE GETS RICH on these horses as they are not secrets, but they are solid contenders that I usually use as KEYS in exotics.

Boiling it down, the only differences in one horse or another (and it has NOTHING to do with Man Made class levels) is the ability to handle a narrower (cheap stock) or wider (best stock) range of pace pressures. Horses don't know they are 12.5K or 30K level horses, but the pace of those races will differentiate the two animals.

[turtleman]

Thanks Lt1, TimY and Bill V
I really appreciate all of your input. I will have to reread your posts several times to digest everything. One thing I have realized in the last week is that it takes a lot of discipline to become a good capper.

Thanks again
Tom

[DontSayDont]

"The Theory of Large Numbers requires that a large database be evaluated before the results are representative of real life."

This theory is something I wonder about when it comes to handicapping. In order to have a large db it would have to go back to times when training was done different from today. There is a poster on here with a db with over 1 million entries that show horses that are off over 120 days to win only about 2% iirc. In fact I believe he won't use a horse off over 90 days as a win contender.

Today's training methods with regards to days off is very different from 15 - 25 years ago. Used to be a horse should be coming back in 14-21 days and even have a workout. This is the information that would be stored in the db. Fast forward to today and there are a lot more horses coming from much longer periods between races. It would require many thousands of races before the win % could be reflected in the db to change that 2% to a percentage usable in today's handicapping practice.

Therefore it is MY belief that it is not necessary to have a large amount of numbers to evaluate some profitable handicapping profiles. I don't have any mathematical proof to back my theory.

Ray

[Tim Y]

What do many statistics actually tell us?

I constantly hear conclusions reached on SKEWED inadequate data. Just recently I was discussing how trainer stats (because of their VERY limited and dynamic changes) are so full of holes when viewed objectively.

Trainer Smith is a 45% trainer of 3 year old fillies. He may have 8 3 y/o fillies but ONE of them accounted for just about all the wins. Is HE that great a trainer, or RATHER, does he not have a great FILLY in his barn? Also IF you look at these same data over a long period of time, with the changing nature of many a barn in the make up of their stock running, one season to the other this same trainer may hardly have ANY 3 y/o fillies in the barn.

Troy Taylor, a respected journeyman trainer in Vancouver had never been in the top ten trainers at Hastings since 1965. He was contacted by a generous, horse loving benefactor (Glenn Todd who was the force behind Mario Gutierrez's transition to greatness) who cares a lot for the game and was not afraid to spend money to get many low level stakes horses off the tracks at Pimlico, the Fair Grounds and Saratoga and bring them west. This influx of great stock gave this journeyman trainer a great edge on HIS competition and LO and BEHOLD, ever since the marriage of these two forces, TAYLOR has been leading trainer at Hastings. Did HE, all of a sudden, get to be a prodigy or, did the horses get better? Just like the pass receiving percentage for quarterbacks is too dependent upon their receivers rather than the thrower, statistics on the trainer are overly influenced by the level of stock one has in the barn. As such they should not be trusted.

BASIC statistical analysis has to be viewed in this light or we are lost at making the realistic conclusions.

Read THE DRUNKARD'S WALK: How Randomness Rules Our Lives, and it will make you have a completely new interpretation on the LIMITED conclusions we can take from poorly set up, but somehow trusted statistical evaluations.

[Mitch44]

DontSayDont; IMO you absolutely nailed this. An overly large data base isn't necessary to get good data and the larger it is the harder it is to change it by recent and better data.

Yes todays trainer methods have changed due to better equ.,feed and a host of other reasons that render a large and old database useless. In fact if one were to include class levels(@ least 20 or so levels) the whole study of days between races would change because better class horses can be trained up to a race and cheaper horses generally need to be raced into condition. If your not breaking this stuff down by the basics of horseracing such as conditions of races I.e.( sex, dis., class, surface etc.) your going to end up with a flawed study.

Those with overly large data bases spend little time playing and almost all their time managing their database. HUGE samples??? I don't think so as life actuaries are broken down by many different things, I.e., stress in a large city is a factor as compared to suburban dwellers and different races are subject to different diseases.

Now to each their own however this is how I do it and handle this problem. In the very early 90's I learned from the "Doc" at a seminar at Saratoga that the same thing will happen in a sample of 20 that will happen in a sample of 100 etc. If you want to round off or break ties the "Doc" suggested using 21. This is what I have always used and it has served me very well over the years. Survey a sample of 21 then make a decision to abandon it or confirm with a larger sample by adding to it. I add to it for a sample of 50 and absolutely no more than a sample of 100 for those promising ideas only. Ideas that aren't promising are abandon after 20 or 21 sample size. I can test many more ideas with this technique then obtaining a HUGE data base hoping it may turn into something after it has long proven not to be a worthwhile endeavor. As I said this is how I handle studies, samples, data, factors etc. As "The Hat" would say "Use what you have" whether a trainer stat that all came from one horse or not if that's all you have than use it; It shows that the trainer is capable, if the horse has the ability he'll get it out of the horse. Of course you can refine any trainer stat with your own take on it, if your aware the stat all came from one horse well that's a big edge over the public. Your own studies will always outperform someone else's data. Hope this nugget helps as it has served me well for many years.
Mitch44