Modeling Question

[Tom]

Supoe I have a handicappng factor that shows the following:
W 31%
P 54% (23%)
S 77% (23%)

31% of the horses ranked #1 on this factor win, 54% 1 or 2 win, or 23% ranked 2 win.
Now, how would you weight this to award points to
horses based on their ranking here?
I want to take several facctors and give points for the rank in each and then total them up and make an odds line. So a horse ranked #1 here would get some number of points, ranked 2 would get so many....
Any ideas?
Tom

[andicap]

Tom,
I just don't believe that these models predict the future. How many models have you come up with that ended up running dry holes. Even if you found a couple that did work, how would you know if they would in advance. Seems basing an oddsline on them is pretty dicey.

[GR1@HTR]

Tom,

Just and idea, but creat a consensus fig...Award 1st rank 10 points, 2nd 8 or 9pts, 3rd 6 or 7 and so on...I've never created one but the sw developers might be able to tell you how to do it. Once you get your totals, buy gordon pines value line maker thing(aka Flux Capacitator). Chug and plug your figs into it then you should be good to go.

[Tom]

I used to just give 3-2-1 points to the top three, but looking at the numbers here suggest the 2nd and 3rd ranked have an equal chance of winning, but my old way would give 2 twice the chance as 3 (2-1).
What i did was give 3-2-1 for 5-6 factors, then total up all the pointss, and calculate odds for each.
Probably the whole thing is an exercise in futility, but
it just sounded so logival at the time....
Tom

andicap-


You are so right-if I have a model that says 9 out of 10 winners is ranked 1 on some factor, then I will surely start beting with the 10th.

Tom

[Tom]

What I used to do was give 3-2-1 to each of the top 3 but these numbers suggest that 2-3 have an equal chance of winning. I want to give points that refelct this relationship,, but I am not exactly sure how to do it.
Tom

[Larry Hamilton]

Tom, I think you are setting your sights too low. You are willing to find profit at the 31% win rate level. Just last night, i found 11,000 + spot plays all over 40% win rate and some over 60% AND ALL PROFITABLE....This angle doesn't require trainer data or anything else that has the ability to "learn" and disrupt you using the data as predictive.

A parting shot--I defy you to find, measure and use a catalogue of thousands of spot plays WITHOUT a computer data base

My advice is keep looking....

[Rick Ransom]

Larry,

Surely you don't think that most of those "spot plays" will hold up in the future. You may have found some gold there, but I'll bet there's a lot more iron pyrite to go along with it.

[NoDayJob]

What's the average payout on your top two? -NDJ

[majestyx]

Larry,

I am curious as to how you create and maintain your horse racing database. What program do you use, what data is collected, and how is it compiled (from pp and result files?). I'm assuming you run certain queries on on this data to see how certain scenarios would play out based on win percentage and ROI.

I have limited knowledge of MS Access, but when you start talking about 11,000+ races with numerous pieces of data per horse in each race, I'm guessing that this application isn't quite powerful enough...but then I could be wrong. I'm guessing you may use MS SQL Server or perhaps an Oracle database.

Any insight on how you go about this would be very much appreciated.

Thanks,


Michael

P.S. 11,000 spot plays?! Mind posting one or two of them?

[Rick Ransom]

Tom,

It would be logical to rate them according to the win percentages of each choice, but I don't think they will stay the same over a larger sample. I think you should assume similar percentages to that found on favorites, 2nd favorites, etc. That would be something like 6, 4, 3 and maybe 2, 1 for fourth and fifth also.

[hurrikane]

Majestx.
I run access ..have around 25000 races. some I know have 70K and more. Don't know what you are thinking but if you have access and understand it you have half the battle won. As for data..you will probably have to buy it from one of the several data warehouses on the web. how good is that data..that 's a whole different question..you need to shop around and find something that suites your hadicapping style.

[Larry Hamilton]

yes Access,

The technique is to program with SQL in a macro loop...there is no limit to how many queries you can run. Well, there is one limit, which I have reached one time--when your hard drive gives up the ghost.

Won't share any. Look around, you'll see that it seems my purpose on the net is to provide entertainment to 3 fools. Why should I share?

Don't believe me? That's ok, my only point was that Tom is setting his goals too low, all else is my business.

[Tom]

Not at all...I was just using one factor as an example.
What I want to do is use 5-6 factors and somehow rank each horse by how they are ranked. For example, another factor I am looking at is something like this:
15%
55% 40%
78% 23%

The 2 rank horse is much more likely to win than the top ranked. So is the third. So to start, the group of horse ranked 1-2-3 have a 78% chance of winning and the rest have a 22% collectivley. So I have 78 points and 22 points to award. Forget the 22 points-they are the non-contenders. How do I split 78 points amoung the top three?

I am sure there are valuable spot plays out there, but after two months of maintaining one with HTR data, I can tell you it is really getting old fast. I want to take the model data, which I update every day and use it constructively.

Tom

Tom,
the only problem I see with assigning impact values to certain specialized handicapping factors or "angles" ,is that you can end up with factors that are averaged,instead of fine tuned,or worthless.
Example:
I come up with a factor that says 28% of winners in claiming races are horses that are dropping in class.
I also come up with the statistic that 20% of the winners in claiming races are raising in class.
Now , depending on how this is imputed to your database,you could come up with a factor that says these two factors point out 48% of winners,or if its put in wrong,your computer will tell you 24% . You don't see that 52% of winners do not have either one of these factors.
Thats where I see the rub.There is also the possibility that two factors can statistically erase the benefit of each angle,in effect making both incorrect.This possibility becomes more pronounced when additional factors are added to your database.
I say this not to discourage you,but to warn you to what can happen to statistical databases.