Here is the entire original thread:

Figure Making Methodology

Posted by Jerry Brown on May 02, 2000 at 12:43:10:

Recently David Patent posted a question on this site which, although barbed, raised some important questions concerning the making of figures, and the differences in methodolgy between Thoro-Graph and Ragozin. For those unaware of the background, the issue at hand is the dramatically different figures assigned by yours truly and Ragozin to the Wood Memorial - we had the race going fast, with several horses running new tops while Ragozin had it going even faster, substantially, with 5 of the 12 horses jumping to big new tops.

The point I made in earlier postings is that since the track was sealed early in the card and opened just before the Wood, since it rained during the card, and since every race after the third was around one turn, using these races to make the Wood variant and figures was crazy, and would result in giving out crazy numbers. In response to a question on Ragozin's site Friedman said that was exactly what they had done, although he didn't seem to think there was anything wrong with it.

David then posted his question, most of which follows here:

I finally think I understand how TGJB got the variant for the Wood -- he made it up.

By this I mean that he apparently decided what the a particular horse or horses should have run in the Wood and constructed the variant to fit the pattern that he believed should have held for those horses in that race. Brown ridiculed the use of surrounding races as "silly," so he clearly did not use the surrounding races to get the variant. So, the question is, if not the surrounding races, what can one use?

I have read before comments by TGJB to the effect that he will at times fit the numbers to the pattern he believes should prevail rather than accepting what appears to be an unbelievable number.

Now, that may or may not be a sound way of handicapping, but how can you possibly prove or disprove such a method? As such, one who makes numbers by this method has no grounds to criticize another maker of numbers except on the basis that horses cannot move or should not move forward or backward more than a particular amount in a given race. That seems to defy logic and history. Horses do crazy things all the time. It seems to me that we must have some anchor to tie our numbers to, and if not, then it is pure guesswork.

For starters, let me say this: Even if it were impossible to make figures based on one race, it would be wrong to tie it to surrounding races in circumstances like this - you would be no better off than tieing it the same days races at Keeneland, or the races of the same day a year earlier at Aqueduct. I mean this literally.

Aside from the weather,aside from the other races being one turn, the fact they sealed and unsealed the track especially with water in it, makes comparison a no-go. It is, in effect, two different tracks.In a situation where it actually is impossible to do a figure - like Chilukki's debut last year, when the track was obviously changing speed and the field was all first-timers - you are better off not making a figure than making a misleading one, and we left boxes for that race.

But it's seldon impossible - most of the time it's fairly sraightforward. David is right - horses do crazy things all the time. But groups of horses seldom do. That is the whole theory behind their "projection" method of making figures, used by Beyer, Myself, Ragozin, etc. - That the past histories of horses can be used to predict future figures. If that premise is wrong, not only can't you make figures, you can't use them to predict the outcome of races. And if you make a mistake,like Ragozin did with the Wood, it compounds itself because you will use those figures to make later figures.

Where David is wrong is that he says I use the patterns of the horses - I do that minimally. Mostly you use the speed (ability) of the horses. To ovesimplify it, if you have a race where a lot of older horses with tops of 10 are running 5, you did something wrong. Obviously, if you have a situation with lightly raced horses and a weather change (or it's the only grass race of the day, after a dark day, and it rains that night), it's gonna get tricky. That's where judgement comes in. And I feel compelled to point out that at this point I've done a lot more track days than Ragozin has - for the last decade. I'm pretty sure he's only been doing two circuits, and even before that never more than six. I've been averaging around 10 for 18 years. My judgement is pretty good.

David's other question was, in effect, how do you know who is right. Well, you don't find out by blindly following dogma. You find out by checking to see which figures correlate well with results in the long run, and by using common sense. For the former, we have set up the Red Board Room. For the latter, I raised this issue to begin with, and I think anyone who is not blindly partisan gets it. But I'll go further.

We're going to post the entire Wood field, with the numbers they ran. David, why don't you see if you can get Friedman to do the same, and tell me which makes more sense. Assuming they won't do it, look at ours and tell me how they would look if they were two points faster (keep in mind our figures run a couple of points faster than Ragozin's).

We're also going to re-post a response of mine from a year ago, that deals with these and related subjects in more depth.

From posting 2.25.99

Back in the '80s, when I had only been making figures a few years, the Belmont meet began. The Widener and inner grass courses, which for years had running at the same variant as each other, suddenly split in no consistent pattern. One day one would be five to six points faster, the next day the other would, it would stay that way for a couple of days, then they would be at the same speed. It was clear from looking at the horses at the time something was going on, but what? It made no sense.

So I sent my top track man (Bill Spillane) out with a tape measure, figuring something was wrong with the turf rails (0, 9 foot & 18 foot), or the starting gate was being positioned incorrectly. No dice. Spillane said everything was right. I had no explanation, but since it was clear from looking at the horse's figures I continued to split the variants anyway.

Connie Merjos was ragozin's NY track man at the time, and a friend of mine from my Ragozin days. A couple of weeks intot he meet he stopped by to say hello, and I asked him what the hell was going on. He had no idea what I meant--Ragozin had not been splitting the variants. A few days later he called me back to say that he had talked to other figure makers (Lawton, etc.) and every one but Ragozin was splitting. And now Ragozin had sent him out with a tape measure, but he had found nothing, so Ragozin was continuing to use one variant for both courses.

I wonder if anyone has figured out the punch line.

A few days later Spillane had to hang around the track late, and happened to be there when they turned on the sprinklers--Widener course only. The courses were being watered on different days.

The day after I found this out I told the story to Julian Weinberg, who had switched from Ragozin to Thoro-Graph (he later worked for us) but still had friends in their office. Julian went over to 11th Street to say hello, and told the story to Bob Beanish (for which I threatened to remove his windpipe). According to Julian, Bobby thought it was silly, called Ragozin over, and told him the story. To which Len replied, "So that's what's going on."

The point is this: make as few assumptions as possible. In Robert Heinlein's Stranger In A Strange Land, a character, when asked what color a house is, says "the two sides I can see are white."

I make no assumptions about the relationship between one and two turn races because I have found the relationbship changes often at virtually at all race tracks (see Monmouth)--I hardly look at one when doing the other. I make no assumption that track speed is constant throughout the day even without weather change because I have found (contrary to a statement in Ragozin's book) that it often does change, especially at tracks near large bodies of water (see Bay Meadows).

Here's what we know: horse's past figures can be used as a guide to what they will run in the future, and to what they ran on the day for which you are doing variants. How do we know? It's the premise. If it is wrong, not only all speed figure methods invalid, whether projection or class level, but the whole exercise would be pointless--they would be useless as a predictive tool for betting races.

How do we know the premise (and figures) are right? By the results of races. If the results correlate with the figures, the figures are right. If not, they are not.

Here's what we don't know: we don't "know" the relationship between sprints/routes is a constant, we don't "know" two contiguous grass courses have the same variant, we don't "know" track speed stays constant all day because the weather doesn't change, let alone stay constant one day to the next (an argument I had with ragozin when I was there--if there was no moisture between days he assumed no change). There's a lot people "know" that we don't know.

We do know that much of what we use as underlying data is inherently inaccurate. Wind speed is estimated by a human being who is looking at flags (on the roof or infield, not the track itself) before and after the race, not during. And the true effect of wind has a lot to do with direction, since there's a very large structure (grandstand) right there. Incidentally, Ragozin uses airport wind at some tracks--hourly readings taken miles from the track.

We also know that 5 pounds = 1 point, which both Ragozin and Thoro-Graph use, while a good estimate, is not accurate. But since we can't get the body weights of the individual horses, it will have to do.

We know ground loss is estimated by humans, and some of these guys tend to use more of the "fan out" (top of the stretch) than others, and some are more conscientious than others.

So, given the above, we know that claims of super accuracy due to video frame counting etc. is the equine equivalent of counting angels dancing on the head of a pin. Accuracy, indeed, depends more on the judgment and common sense of the figure maker than on the raw data, as you said.

TGJB



Posted by The Fat Man on May 03, 2000 at 00:03:55:

Discussion provides much food for thought!

Love those Rags people, but now respect you

too Mr. Brown. Now off to the Salad Bar.



Posted by David Patent on May 02, 2000 at 20:12:58:

First, I want to thank Jerry for a temperate and well-reasoned response to my question. You were very helpful in illuminating some of the more subtle differences in approach between you and Ragozin (although I am not sure what the turf course watering anecdote has to do with different methodologies. Sounds like Ragozin goofed and once he knew the facts I would assume he split the variant).

Actually, both sides agree that you can use the past to predict the future and that horses do generally run within a certain range of their ability. That holds more true, however for 25K older routers than for 3 y.o.s in April. In other words, I am much more willing to believe a big jump up in the majority of a field of top 3 y.o.s in the Wood than I am 25K claimers (although I have seen that happen too occasionally).

Where I disagree in terms of provability is Jerry's point that you can track performance by seeing which figures correlate well with results in the long run, and by using common sense. When you say results do you mean betting results? Can't be that because money management is everything in betting. Could you set up a program to bet Rag and Thoro using the exact same betting criteria and bet the same amounts based on some pre-ordained formula? Maybe but I doubt it. Also, I'm all for common sense but I don't know how you quantify that.

To me the real question of whether to use Rag or Thoro is kind of sheet you are comfortable looking at. (I'm serious here). Thoro sheets look much "prettier." Lines are smoothed. Horses run little tops and bounce a little here and there relative to Ragozin, which often produces very ugly lines on horses -- corresponding to the view that horses more often do crazy things (for which they usually pay the price in the next race or two).

That said, thanks again Jerry. I'd love to see the Wood numbers up on the Rag board. Although I am a Ragozin customer 99% of the time I am often frustrated by the nonsensical distribution system they cling to (i.e. no downloads) and the sometimes Red Square like approach to dissent.



Posted by Jerry B on May 03, 2000 at 20:15:02:

TG--First of all, did David and JR Litt compare notes before posting or what?

To clarify a couple of points:

1. I used the example of the grass courses to show the danger of assuming correlation between two things (two grass courses, consecutive days, one and two turn races, sealed and unsealed tracks) and making figures combining both, as opposed to doing them independently. Incidentally, Friedman said in his post they did the day on a "slide", which by definition means they did correlate the sealed track to the unsealed one. I look at all the sorrounding info too--I just don't make as many assumptions about using it.

2. Yes, 3yos in April are more likely to run new tops than older 25K claimers--that was why I was willing to give 3 horses new tops and two others pair-ups. But once you know to disregard the surrounding races that became by far the most likely scenario to be correct--horses sometimes run 6 point new tops too, but it would be unreasonable to have a whole bunch of stake 3yos in April doing it. The only way you can come up with the scenario Ragozin did is to tie it to surrounding races--otherwise there is no argument for it, other than "anything is possible".

3. There are various ways to attempt to check which figures are best--Sportstat just finished a 6 month mechanical study, for example, which will be published soon (we have no idea how it came out). But when you are dealing with an individual race where lots of horses get huge numbers, the best way is to see what they do over a period of time.

If several horses who ran big numbers fall apart for a long period of time (as opposed to bounce), if they pair up a lot on one set of numbers or the other, or if they settle in at least 2 points off that number--any of these would be something to consider.

4. As far as tighter (prettier) lines, the way I do it, using horses previous figuures, that's going to happen. But since I don't fudge within a race, the fact that I can pair up so many horses is evidence that the earlier numbers and the new one I'm making are likely to be right. Think about it.

5. If JR is referring to Skip Away's Blue Grass, we gave him a 1-3/4 that day. I personally bet out on him that day, and against him in the Derby. And yes, future figures gavve evidence that the number was right.

TGJB



Posted by JRLitt on May 02, 2000 at 20:09:31:

I appreciate this explanation because it is well reasoned and is devoid of some of the animosity and defensiveness often displayed when discussing the differences in methodologies. I have a couple of points.

(1) I don't think anyone disagrees that past performance is predictive of future performance. I do, however, dispute the notion that it is necessarily true of top three year-olds running in April. It would surprise me to see several 25K claimers jumping up to big new tops. It would not surprise me in the least to see that happen with the 5 horses in the Wood. That says nothing about whether Ragozin is correct in his numbers, but I do not believe that one can look at the Wood in isolation and predict a number based solely on past performance and relative order of finish.

(2) I dispute the notion that one can determine the validity of one method over the other by examining subsequent races. JB used the example of Chiluki's number of last year, which I believe Ragozin gave a 5 (I guess Thoro simply made no number). Seeing that number, I drew several conclusions: (a) Chiluki was unlikely to repeat that number anytime soon (b) was likely a better sprinter and (c) would likely not have a very short-lived career. All of these turned out to be true. I could also say the number was clearly wrong because she never ran close to it again and that it is highly unlikely that a 2 year-old filly would run such a number in May. There is simply no way from the subsequent data to "prove" which is correct.

Similarly, there will be no way to know based on the next few races, who was right in the Wood (though I would say if FP runs terrible as I suspect it would lend credence to the Ragozin number. [take a look at Skip Away's Ragozin sheet sometime as an example where JB would probably have disputed the number because it was too big, but is perfectly consistent with his future performance). If these horses run worse over the next few races, they could all be reacting to the number. If they hold the numbers, they will have reached new levels. Both are consistent with typical 3 year-old patterns. I would love an explanation as to how the subsequent races are going to "prove" one methodology over the other.



Posted by Dan on May 02, 2000 at 14:34:42:

This is one of the reasons I really prefer Thorograph. As an owner, I have had two horses race on one day at Bay Meadows (which you mention in your post) and have had the slower one run a faster final time. Just get the results charts from BM, somedays you'll see 6,250 claimers go 6f in 1:09 flat and later in the day Alw NW1 horses go in 1:09.2. Whatever you think of NorCal allowance horses almost without exception they are much faster than our 6,250 claimers. How else can you explain this phenomena except to say the track changes over the course of the day. As for sprints and routes, if you don't believe in this, all you have to do is look at the DRF variants. Look, I know they are badly outdated and almost useless as a handicapping tool, but if you don't believe that there is a difference in sprint and route track speeds then the DRF variant, as basic as it is, should be identical or close to it, for sprints and routes. It rarely is. If you average it over extended periods of time you'll see the routes and sprints have completely different variants. Ask any trainer, the horses are more consistant than the track condition, and thus, they are the best guide of how fast they run. Thanks again for explaining all this JB, those of us who believe in TG really appreciate it.



Posted by Richie on May 02, 2000 at 19:55:16:

Jerry, I have been using you for probably 2yrs, I don't bet serious money unless I have the sheets. You sound pretty smart. I am a happy man. Thanks



Posted by Lost Mountain on May 22, 2000 at 08:45:27:

Here's my problem with TG; horses run the same # all the time. Like they are machines. The other guys don't flat line nearly as often. Sorry guys, but there is no comparison.