TDN - Using Speed Figures to Track Possible Cheaters

[nicely nicely]

Story begins on the front page of the August 28th edition of the Thoroughbred Daily News

https://www.thoroughbreddailynews.com/pdf/tdn/tdn210828.pdf

[sand1trap]

Is the data sourced from the algorithms available for purchase? Would that not provide a significant edge being able to locate these “new” abusers before they become more “common” knowledge? Ty

[TGJB]

We don’t even have access to it, we just supply the underlying data.

[APny]

But you do have that list of 10 suspect trainers......It would be amazing if that somehow made its way into the public domain. I\'m sure most regular players could name almost all of them.

[Boscar Obarra]

If the problem was subtle , it wouldn\'t be much of a problem.
 
 No finesse in the 21st century , in anything .

[Tavasco]

Boscar Obarra Wrote:
-------------------------------------------------------
> If the problem was subtle , it wouldn\'t be much of
> a problem.
>  
>  No finesse in the 21st century , in anything .

Well said, \"Bridge\" may be the last bastion!

[TGJB]

I saw the test list they produced before the pandemic (and I mean right before, my last in person meeting for a long time). The way the program works is it uses a rolling last few startsâ€" I explained to them that guys run hot and cold, if you use large body samples and averages you won’t get it. (This, by the way, is the mistake in Ragozin’s methodology. It’s right to use large samples and averages- pars- when you first set up a database, because at that point you have nothing else to work with, but wrong to do it after that because you’re dealing with variable conditions throughout the individual days).

In any event, the best that program can do is approximate what a good handicapper who is looking closely at TG for their own circuit can come up with.

[hellersorr]

Well, I\'m sympathetic to the cause - but the powers that be will have to have a helluva lot better than \"guys run hot and cold.\"  Baseball players that hit .300 for the season don\'t hit .300 every seven days.

[TGJB]

Which is exactly why we use figures, not win percentage. That’s where my conversation with the Jockey Club started, and it was also the basis for a fierce argument I had with another well known contest player (who was in a reality show about horseplayers), he thought we should use win percentages.

If win percentages told the tale we would be using them to handicap instead of performance figures.

Keep in mind also this isn’t being used as evidence, just as a means of narrowing down the number of people who need to be focused on, in such a way that they can’t say they’re being picked on.

[colt]

In any event, the best that program can do is approximate what a good handicapper who is looking closely at TG for their own circuit can come up with.

Will you consider expanding the trainer stats (tops/pairs/off/X/win%) to include the Last 30 days (calendar days) at the current track/circuit? The Last 90 days and the cumulative historical track stats are a somewhat useful starting point, but often is not reflective of current trends and especially so for stables that compete on multiple circuits - e.g. Saffie A. Joseph Jr. at Saratoga.

[TGJB]

When you get to something like 30 days on one circuit sample size becomes a problem for most trainers, and a top and an aberrational jump up are not the same thing.

 The best way to do it is to play close attention. That’s what “angle” players like Serling and Hoover do with their home circuits.

[hellersorr]

I\'m not talking about win percentages.

Trainers NOT using illegal drugs are going to go on streaks where their horses\' FIGURES are abnormally good and abnormally bad.  

Right?  In my very best Queens English:  Am I right or am I right?

And, again, I\'m sympathetic to the cause.  I have little doubt that some trainers cheat. (If some trainers didn\'t cheat it would be a very bizarre profession - i.e. a profession not populated by human beings.)

I just don\'t see how hot-&-cold streaks would be probative.

[TGJB]

No. The type of aberrational jump ups we are talking about happen a tiny percentage of the time, like 5%ish. The chance of two in a row happening is 5% of 5%, which is not “all the time”. When a trainer has a grouping with several of those (as happened when Frankel had that one year where everybody in his barn moved up at the same time), the program notes it, and the powers that be, if they are doing their job, look at the NEXT grouping in the rolling sequence. Then if it happens again they do extra surveillance, or whateverâ€" again, it’s not being used as evidence.

[hellersorr]

Okay, got it.

[colt]

Good point regarding the sample size. Having given this some more thought, the last 30 starts on the circuit/track would be a much more worthwhile analysis.