Friedman/Trainer Stats

[TGJB]

   Friedman has posted his thoughts on “trainer stats” on the Ragozin site.

1) It seems hard to believe, but Len actually seems to be taking the position that trainers don’t have strengths and weaknesses, that (as he said in a post a few days earlier) it is only worth looking at trainers in terms of overall strength, and noting when a horse changes hands between trainers of different strengths. Putting aside for a moment how we go about measuring them, it should be obvious to any sophisticated handicapper that trainers have characteristics, just as hitters do. Sure, one guy hits .350 and another .250, but that’s not what they talk about in the pitchers meetings—Ordonez will chase stuff up in his eyes, Kent chases sliders in the dirt, Shinjo can’t hit a breaking ball, Ryne Sandberg wanted it middle-in, Albert Bell wanted to extend his arms, etc.
   Likewise trainers. Some wind them up first time out or off layoffs, some don’t. Some have a good opinion about whether a horse can handle grass (and therefore win first turf), and some don’t. Some trainers treat horses who hemorrhage (bleed) differently than others, etc. And then there’s the question of intent—when we first started doing the profiles 5 years ago the first one that caught my eye was that of Bobby Klesaris, who at that time was 0 for many years with first time starters in special weights, but something like 8 for 10 with first time starters in maiden claimers.
   And sure, you have to put the data in context, which is the point I made in my earlier post to Friedman. Some guys get a lot of grass pedigrees, for example, so you have to take that into account. You don’t simply say that since a trainer is 20% in a category this horse has a 20% chance of winning—that would be silly. The best way to use the data is as a supplement—that the horse’s chances of running well are enhanced (or hurt) by the trainer’s tendencies, and to answer questions that can’t be answered by figures alone—first out, second out, bet, not bet, fitness off layoff, etc.  

2) There could very well be merit in discussing what constitutes a meaningful sampling (I never said anything was “overwhelming”), and the Lasix sample on Jory may not be big enough to qualify as such. But Friedman seems to be saying that measuring results by the figures horses run is meaningful, and win and ITM percentages are not—meaning there is no significant correlation between the two. If that is so, what the hell are we all doing making and using the figures to bet with? If done right (see below), studies done using our figures would be even better, and we may do something along those lines at some point. But there are drawbacks to it, and what we have now is definitely worthwhile.

3) Re: Friedman’s comments about our accuracy—we get our underlying data directly from Equibase, who also sells it to whomever Ragozin buys it from. Equibase also directly sends us a maintenance file that cleans up any errors. While certainly some errors can creep in, it is therefore close to impossible that Ragozin is using base data with less errors, since they are one step further from the source.
    As to the profiles, we don’t take them from anybody—we create them ourselves.

4) The “objective” method Friedman used to evaluate the two trainers is similar to the broad averaging Ragozin does to do figures for a day when the variant is changing, and suffers from the same problems. Aggregate points forward or back for a group of horses is not the question—it’s no different than average points forward or back. To oversimplify things, when you are looking at something like first or second Lasix (and many other categories), you are looking to see what percentage of horses jump forward enough to win (or finish ITM).
   Looking at the second Lasix stats for Jory we see he only had 6 runners ITM, but 4 of them won. In the same category Wolfson (with twice as many starters) also had 4  winners, but 15 ITM. It seems likely that a larger  percentage of Jory’s horses make a jump, but a larger percentage of Wolfson’s run about what they usually run. Breaking things down by percentage that go forward 2 points, 4 points etc. would be more meaningful than an average. I have no idea what that study would show with Ragozin figures, and frankly wouldn’t even if Friedman told us (remember his comments about Touch Of The Blues BC ground).

5) Idle thought—I mentioned ten different trainer categories in the ROTW Friedman referred to. Wonder if he ran studies on all ten? As for the $1000 he mentions, what I got for that was Friedman’s admission that track maintenance affects track speed, which contradicted his earlier statements (regarding the 2000 Wood Memorial figures) that it did not.

[TGJB]

Okay, Friedman came back with a response, which judging from the time it took to post (and since he didn\'t respond to 90% of what I said) he probably wrote before I even posted mine. On the specifics of second off a layoff, as used in our profiles it means a race within 30 days of a race after a layoff of 90 days or more, and does not include second out lifetime. And, of course, he is using those large averaging methods again.

Which brings up an interesting point. The point that Friedman makes about each category really being a group of categories within it which have both similarities and differences is correct. Which is why, as we have said before, it is important to know how to use the profiles properly-- to place them in context. We do it that way for a variety of reasons-- space limitations, sample size, etc.. Ultimately they are meant to be used as guides, not hard figures.

But here\'s the interesting part. Ragozin does the same thing (mix unrelated data and average it) to create his figures, and they are sold as \"objective\" truth, not as guides that should be placed in context. For those who are new here, I went into this in detail in several earlier posts that can be found in our archives-- notably \"Now that I have your attention\", 11/14/01, but also \"Changing Track Speeds\", 11/17/01, \"Split variants, Bruno\'s take\", 1/03/02, and \"Figure Methodology II\", 6/10/02.

Most interesting of all is Friedman\'s change of tactics, presumably as a result of their finally starting to sell on-line. If he really does come out to play on a regular basis rather than hide and censor I think you guys are going to be kept entertained for a while, even if it does end up taking another 2 hours a day out of my life.

Jes\' don\'t throw me in the briar patch.

Meanwhile, I know for a fact that there are guys out there who know a whole lot more than me or Friedman about trainer patterns (Chris, Jon), as well as dwarves and assorted others who may as well. Comments?

[Michael D.]

example: Ian Jory

overall ROI: $1.42 in 1095 starts. this guy performs a bit below the market, but probably no way to make money looking at this stat.

routes ROI: $1.13 in 535 starts. his horses don\'t do so well going long. in my opinion, this stat can be useful, as his ROI has a noticable drop in this category.

sprint to route ROI: $0.90 in 122 starts. again, I think this stat is useful, as it proves that he can crank em up sprinting, but may not train as much stamina into his horses as other trainers.

last race 11-29 days ROI: $0.94 in 614 starts.

last race 30-89 days ROI: $1.45 in 268 starts. these last two stats are somewhat useful. most horses run slower on short rest, but Jory\'s run even worse than the market thinks they will. in the 30-89 day category, his ROI is in line with his overall ROI.

4-1 or less ROI: $1.39 in 208 starts
over 4-1 to 10-1 ROI: $1.88 in 312 starts
over 10-1 ROI: $1.17 in 575 starts

I think these last three stats are somewhat uesful. when Jory\'s horses aren\'t taking any money, they perform below the market. of course, most horses who don\'t take money don\'t run well. in Jory\'s case, however, they perform even worse than expected.

STK ROI: $0.70 in 111 starts
a useful stat I guess. his horses underperform when running in stake races.

out of the 27 stats that TG provides on Ian Jory, 15 or so do not mean that much to me, as the samples are too small. of the stats that do have some importance to me, I do not think any of them looked at alone will make anybody rich at the race track. I do think, however, that if you get a general idea of the strengths and weaknesses of each trainer, you have a leg up on your handicapping competition. Ian Jory\'s horses look to be a bit stronger in sprints than in routes, and his horses tend to get overbet in stakes races. sure, Jory\'s next three sprinters may lose, while his next three routers may win, but in the long run, you will have more success betting his sprinters.
in my opinion, if you can get this sort of general feeling about every trainer, you do have a handicapping edge.

[OPM]

JB, I wonder if you ever thought of doing trainer stats in terms of TG #\'s.  For ex.  after a 90 layoff, what percent of trainer horses hit top, new top, within 5 pt of top or some variation of this.  You can then break it down to 3 yr, 4 yr old etc.  Win, ITM may be misleading because the horse might run 4 pt off the # you think it was going to run and still win.  I think basing it on TG # will make is more specific for all users.

[Michael D.]

thoughts on this Alydar?

I used some subliminal messaging to try to get you involved......

[mandown]

I\'d have to agree OPM. The whole raison d\'etre for Thoro-Graph is figures so why measure trainers by a different standard? Might just as well handicap a race by saying the most likely winner is the one with the highest earnings.

[HP]

TG-centric trainer stats would be interesting. You can get some of this from the \"TGI\" but what OPM is describing is a whole new thing.

It might be interesting to design a comprehensive trainer stat product that had both Win/ITM% stats and TG stats and sell it as a separate product for a circuit. Maybe you could add some more specific breakdowns in certain categories. There might be some angles/contrasts in there that could be helpful. I don\'t know if you\'d have enough customers to make this go, but it would be a nice package for those who wanted everything they could get from you on the subject.

A long time ago I used to buy something called \"The Trainer Intention Book\" for the NY circuit. I know there are new services out there now, but I wonder what happened to that guy. HP

[TGJB]

Chris has been talking about this for some time, and at some point we may do some version of this. One of the problems here is logistics-- think about how many categories and sub-categories you would have to put on the sheet. To really do this right we would probably have to create a side product, and we might.

[TGJB]

Mandown needs to recuse himself on this one, since he\'s already on the record, and since there are more immediate programming needs.

[HP]

Yes when I said \"separate product for a circuit\" I meant what you are calling a \"side product.\" New York Trainer Stats, Florida Trainer Stats, etc. All Trainer Stats. Online packages. I\'m sure your illustrious computer consulting and market research teams can handle this from here. HP