Figure Study

[gatodelsol]

Is it just me or does anyone else think that by throwing out the worst of the last 3 figs and averaging the other 2 that you are more likely to wind up with bounce candidates than horses that could be expected to run well.  It might be more relevant to throw out the best figure and average the other 2.  Alternatively, it would be more ideal from a statistical point of view to take the last 5 and throw out both the best and worse and average the other 3.  That way you don\'t wind up with a skewing to the high or low end.

[davidrex]

Step up and give out your name if you would switch your allegience if the winner of this contest turns out to be the sheet you are NOT presently using...right.
Jerry should win under stated rules because his #s are alot tighter.
He\'ll also chalk up a larger following because he\'s much more user friendly.Rags has developed over many years an elitist attitude that does not bode well for increased sales to the betting public or the YOUNGER generation of(bettor,trainer,owner).
Why Janis wants to proceed with this charade is beyond my simple comprehension.
SOOO...what if T.G. loses ...anyone SERIOUSLY consider jumping ship?

M.L.

Rags 5/2
T.G. 1/2

[TGJB]

1-- Tony, put them all in one post next time. And you don\'t have to look at the study. As I have said repeatedly, this is NOT supposed to replace handicapping. You seem not to understand at all what we are trying to do.

We are trying to take the judgment (handicapping) OUT, and create a purely mechanical system that will check accuracy, not in individual cases, but in the long run. If you think you can come up with a simple system that will correlate better with the results of races cough it up-- I\'ll bet you sight unseen you can\'t, having spent a lot of time thinking about this. I could go through some individual examples to show why this is the right way, but it would be a waste of my time. Just focus on this part-- in the Bellamy Road example that you used, and in all the others, IT\'S THE SAME FOR BOTH SIDES-- it washes out in the long run.

2- David-- depends what you mean by tightness. If you mean we\'re more accurate, I obviously agree, and that would mean we would do better. If you mean figures in a tighter range, it\'s completely irrelevant-- the horses are ranked based on comparisons to others using the SAME set of figures,not the other guys. Doesn\'t matter at all.

3-- At the very least the study will be pushed back a week. Jimbo had to drop out (he\'s going to be out of the country for two weeks), and Tom and I have been unable to come to terms about how to proceed in his absence. In a perfect world, since Ragozin has the same chance to establish bragging rights from this as I do, Friedman would engage in the process here-- post their ratings and rankings every day, and e-mail me the underlying data so I could check it (and not have to pay for it), and we would do the same in the other direction so they could do the same. But I\'m not holding my breath. I\'m going to try to salvage this thing-- I can do it all here, with everything posted in advance publicly, and the underlying data available to anyone to check, but some yo-yo will still think (or claim) I rigged it. So I\'m going to try to come up with a way where that can\'t be claimed.

[NoCarolinaTony]

That would be my point exactly. I am not sure how omiting a low figure in an array of numbers (3) gets you the best horse or power number. i think if you keep all 3 in and find a power number then you might have something.

I think throwing out the low number will skew the data of some horses. What you want is as an accurate portrial of the horses ability and then you are taking this \"power number \" the lowest  in the field and saying that horse should be the most likely winner based on the lowest figure in a given field. I just don\'t see how throwing out a bad figure hurts the analysis, but rather will improve it.

But either way, I would never Change to Ragozin no matter what the study shows. It is my belief right now the numbers are the best they have been in the last few years.


NC Tony

[Michael D.]

Jerry,

I don\'t like the sample size. I would prefer four out of the last five figures, at least three months.

Also, at the spa, you are going to have to deal with a lot of maidens, off the turfers, and turf sprints. I would imagine a lot of the turf sprint races produce ugly samples (routes to judge sprints, etc).

Too much is going to be left in the wash in this study. Not much interest here.

[TGJB]

Tony-- you have to remember that what we will be measuring success here by is each horse\'s chance of WINNING (who has the winner ranked higher), not head to head. Thus, in a race where there are several horses with tops in the 5 to 7 range:

Horse A: 8,8,8.
Horse B: 20,5,5.

If you average all 3 races A will rank better. Do you really think he has as good (let alone better) a chance to WIN a race that figures to go 7 or better?

Michael-- if you were handicapping, obviously you would prefer the last 5. I would be very surprised if it would matter for a study like this.

What we are trying to do is strike a balance between top ability and consistency (the chance of running that top performance or a good one), in some mechanical way. If this were a betting program where we were going to bet real money, damn right we would tinker with it. But it\'s not-- it\'s a mechanical way of grinding down data over 600 races. The nature of this study is incremental-- nobody gets a big jump because of a $40 winner, there are 600 points to be distributed, one at a time. And the screwing of one side or the other that will come because of the mechanical nature should be roughly evenly distributed over the long run.

[NoCarolinaTony]

Jerry,

So I understand how the scoring will work, regardless of where the winner ranks (ie 6th by TG and 5th by Rags) the winner of the point in this case would be Rags?

Michaels point is correct, in that if you use three figures than all three are better than 2, 5 would be better than 3 and 10 would better yet still a form cycle or two.

I  do understand your point, and could see why you did what you did (your example) but to take it a different way, the horse that has the average of 10 is also more likely to bounce to oblivion more than the horse with the average of 8. The score of 10 reflects that better than a 5 would as compared to the 8. It would be my opinion that the potentially \"better\" horse in this example is also more variable for some reason.

I really would like to understand the scoring part a little bit better.

I strongly believe you need a larger population than 2 of 3 data points to really get this study to be more accurate as a power rating of sorts.  If you took 10 or 20 data points then tossing out a flier wouldn\'t hurt the analysis/projection or power number relecting a horses ability. It\'s not an easy task to do in the first place, and to simplify like this I think will cheapen the study a bit. I really don\'t have anything better to offer than use more data than less.


NC Tony

[TGJB]

Tony--

1-- The scoring is simply, he who has the winner ranked higher gets a point.

2-- We\'re not measuring chance of bouncing, we\'re measuring overall speed, as measured by chance of winning (which is of course is to some degree a function of bouncing, or more precisely, consistency. But a horse with 2 good ones out of three will score higher than one with only one).

3-- The study that Friedman was trumpeting was based on best last figure and best SINGLE figure of last three races. That\'s obviously ridiculous, for all the reasons mentioned times ten. One that goes too far back can bring in races that don\'t have anything to do with a horse\'s development or current form. We tried to reach a happy medium.

And again, I think there will be more correlation between these ratings and results of races than any other simple mechanical system that can be devised.

[Michael D.]

Jerry,

ok, i\'ll take a look next week, but ...

i won\'t be surprised if half of the spa races don\'t register any meaningful comparisons.