Adding two, spring 3yos

[TGJB]

P Eck-- Yes, I looked at adding two points-- it was the only other way to go you could seriously consider. Problem is that it produced a huge percentage of horses running off races and \"X\"\'s-- the distibution would be completely screwy.

By the way, yesterday I tossed off a seat of the pants estimate that stake 3yos at least paired their tops around 50% of the time. Well, George ran it, and here it is:

3yos, March through June, stakes only, same criteria as the Thoro-Pattern (same surface, all sprints or all routes, all races within 6 weeks of each other)--

New tops 22%, \"pairs\" (within a point either way) 28%,  total 50%-- I am the man.

When the study is just colts or geldings, routes only, the breakdown becomes new tops 28%, pairs 33%. Sample size on that one 224.

[BitPlayer]

TGJB -

You may be the man, but you must admit there\'s a certain amount of circularity to your argument.  You make the figures assuming that stake 3yos at least pair their tops around 50% of the time and, lo and behold, the figures show that stake 3yos at least pair their tops 50% of the time.  (Admittedly, that it comes out exactly 50% is pretty amazing.)

Respectfully,

BitPlayer

[TGJB]

So you think I sit there and say, hmmm, there are 8 horses in the race, I\'m going to have to give 4 either pairs or new tops?

I assume nothing about what percentage will run well when I do a day (and contrary to what some people have said I do, most days hold together pretty well, you don\'t break out races all over the place-- both La Derby day and BG day, which I posted sheets for, are examples). You make the figures by using lots of horses of ALL ages. And even if you do a race as a stand alone (only route, weather change, track changing speed), you are restricted by the relationships within the race, defined by beaten lengths, ground, and weight. You can\'t just arbitrarily give horses what you want to give them, and have it come out right.

[HP]

Yes but YOU are the one who thinks it\'s right.  That\'s the circularity thing.  Not that there\'s anything wrong with that.  

HP

[richiebee]

I\'m glad some \"journeymen\" started this thread, because I am still admittedly a ***
10 lb apprentice TG user.

Bellamy Road\'s thoro pattern amazes me. (I know all of the disclaimers concerning patterns and 3YOs at this time of the year.) I know it is based on a tiny (24 start) sample.

Accoding to this limited sample, there is a 25% chance he will reach a new top (breaking Sec\'s Derby record probably) and a 33% chance that he will equal (or worse) his lifetime\'s poorest performance. Fascinating.

Is it counterproductive to publish a pattern based on this small a sample?

[TGJB]

HP-- Don\'t start. I\'m talking about it coming out where a significant pecentage of ALL the horses run in a tight range for themselves. You can\'t make that happen unless it happens-- they are independent of each other.

Richie-- the Thoro-Patterns lump together things which have some differences, so you have to make distinctions, which is a point I made in the introduction to them. I don\'t think anybody would say that a horse making a 5 point jump to a neg 5 was a serious candidate to run a new top.

[JimP]

I think BitPlayer was only refering to your quote from above to P-Eck: \"P Eck-- Yes, I looked at adding two points-- it was the only other way to go you could seriously consider. Problem is that it produced a huge percentage of horses running off races and \"X\"\'s-- the distibution would be completely screwy.\"

That does seem to indicate that a \"screwy distribution\" does influence your thinking on the figures to assign.

[TGJB]

Of COURSE the question of distribution, in general terms, influences my thinking-- that\'s how everyone does figures, not just me. That\'s what the \"projection\" method is all about, whether it\'s stated consciously or not. But that\'s not the same as tying every race to the average, or to your expectation.

This goes to a statement that Friedman made at the Expo, which I didn\'t react to quickly enough to respond to, and which I addressed here once before. Andy and I were explaining why it was right to make race by race variants,and Len said it was wrong because (paraphrasing) in the long run, all permutations (distributions) are possible, and will happen, so it is right to tie every race to the surrounding races (barring extreme weather). There are two problems with this--

1-- as the presentation in the archives shows, it is wrong to make the assumption that the track is staying the same \"speed\".

2-- Yes, all permutaions are possible, BEFORE THEY RUN THE RACE. Once they run the race the relationships between the horses are fixed by lengths beaten, ground, and weight-- you can\'t change one without changing the rest. This means that realistically there are only two or three things you can consider doing with a race-- P Eckharts alternative theory on the Kee races, for example, was the only other one you could seriously consider, and the incredibly high percentage of \"X\"\'s compared to the very low percentage of new tops doing it that way would bring made it unlikely. Along the way, you do come across lots of permutations-- some days (like that nightmarish Kee 4/23 day) you get a few horses who jump out of their skin, AND a lot who X out. I don\'t just decide to have less horses X out to conform with averages, since that would give the ones who ran out of their minds even crazier numbers than I assigned for that day, which were plenty crazy enough.

Of course, you can avoid all this by just using an average variant for the whole day, like Ragozin does-- sprints, routes, start to finish, just throw them together and look at distribution. That way you only have to make one decision.

[DeathBredon]

Considering that Beyer uses the projection method, any idea what he did with 4/23 at Kee?  Would that ever be relevant for you?

[TGJB]

If you mean would it affect how I did the day, no. I did look to see what he did with the 25 LENGTH SPRINT WINNER of the first race (maiden), and he did give it one of the best sprint figures of the year (112). He also robbed him. Try dealing with races where the winner runs 18 points (that\'s EIGHTEEN #&*! POINTS) better than anyone else. Secretariat ran 12 points better in the Belmont.

That place is a bitch to do figures for. The day before was hell too-- and that wasn\'t even a sloppy track, although there were very strong winds.

I once did a study on the average margin between horses on fast tracks vs. very off tracks. The difference was quite significant. The average margin was much larger on very sloppy tracks.

I think you are sometimes screwed no matter how you interpret some very sloppy track results.

If you try to give the horses in the middle of the pack figures that make some sense you will sometimes wildly overrate the performance of the winner.

If you give the winner his expected figure you will sometimes underrate the performance of the mid pack horses.

I once played around with a beaten length adjustment factor that reduced the margins between the horses on a very sloppy track by an amount that made them consistent with typical fast tracks. Then I used the new beaten lengths to make the figures.  

I found it \"very useful\" as a general tool in an inexact game, but it\'s not scientific enough to be appropriate for TG.

I think the reality is probably that a lot of horses just don\'t like extreme slop and others finish well beaten because of all the mud that\'s kicked up into their faces that they have to carry on their bodies. Of course there are some mud freaks out there too.

IMO, you have to take all figures earned under conditions like that with a grain of salt even if you are fairly confident in the methodology.



Post Edited (04-26-05 22:07)

[spa]

Sham broke down in the Belmont, right?

[RICH]

The question now becomes, how does that relate to this years derby? 61% of 3yr colts run new tops or pair right about now. Do they run these pairs/tops after 6+ pt moves.

[gvido]

Sham broke down in the Belmont, right?

Correct.

[RICH]

After seeing some of this years BIG JUMP #\'s., he can possibly jump 5 pts to a 0. What\'s the probability of that?