Class Pars - Convergence?

[Brian Flewwelling]

i know this topic has been thrashed, but i have two questions:

1) does hard thinking when you are confused cause strokes?

2) does the following procedure (described by others, including MV McKee-sp?) converge to a consistent set of Pars that are independent of the starting point for the pars?

a- calc Raw Speed = 90 + (T90-Trace)xBLfactor

b- ClassPars = Avg(Raw Speeds) by class

c- DailyTrackVariant = Avg(RawSpeed-ClassPar) separated by R/S, and Turf/Dirt

d- better ClassPars=Avg(Raw Speeds + DTV) by class

return to c, and repeat at least 10 times, maybe several hunderd.


notes (or confessions) - T90 is the time for the 'universal' 10K claimer for that track, surface, distance
- i assume a class of horses gets the same speed at different distances, so Clm10K uses speeds for all distances... surfaces are kept separate


Sub Question1: if i cause a perturbation, like change the T90 for one or more distances at a track ... thus the Raw Speeds ... will running some more cycles create properly adjusted Pars?

Sub Question2: if the average DTV for a particular distance is noticeably off 0, should that not be rectifiable by adjusting the T90 for that distance (and hence the raw speeds)?

You input would be appreciated.

now i will check the latest 30 laps

Fleww

[Brian Flewwelling]

corollary question:
By Convergence do i mean an average DTV of 0?

If so What Average?

Fleww

[Macdiarmadillo]

1. hard thinking when you are confused may lead to 'roids.

2. GIGO.

You're missing steps in the procedure and adding parts c and d that appear unnecessary. Are you looking just for a set of pars for one track, universal pars, or what? You are also so far ignoring or not mentioning factors such as age, sex, purse value, track condition, race conditions. Pars will be different for each group for the first 2 factors, pars will be thrown way off by not accounting for the last 3 factors.

sub 1. I don't see the point of adjusting T90 for one track. either that's what it is or it's wrong.

sub 2. I though the whole point of pars was to calculate the DTV, which is often noticeably off 0. If you adjust T90 then you alter everything else, then you must adjust T90 again and so forth. Plus the calculation there for DTV is w-a-y oversimplified.

And now, what have the 30 iterations/tests told you?

[Brian Flewwelling]

Originally posted by Macdiarmadillo
1. hard thinking when you are confused may lead to 'roids.

So That is what causes that pain there. Now i know why Joe said Speed figs were a pain in the ....


2. GIGO.

You're missing steps in the procedure and adding parts c and d that appear unnecessary. Are you looking just for a set of pars for one track, universal pars, or what? You are also so far ignoring or not mentioning factors such as age, sex, purse value, track condition, race conditions. Pars will be different for each group for the first 2 factors, pars will be thrown way off by not accounting for the last 3 factors.

i am determining Class Pars, at least that is the goal. i get a separate par for each 'condition' so age, sex, purse, and race conditions are there.
Track condition is to be handled by the DTV (we hope)

sub 1. I don't see the point of adjusting T90 for one track. either that's what it is or it's wrong.

sub 2. I though the whole point of pars was to calculate the DTV, which is often noticeably off 0. If you adjust T90 then you alter everything else, then you must adjust T90 again and so forth. Plus the calculation there for DTV is w-a-y oversimplified.

And now, what have the 30 iterations/tests told you?

The T90 for the track has been taken from a very reliable set of track pars, but as Dave Schultz points out determining Track Pars is not an exact science.
In the case i am currently most concerned with the track surface at NP was rebuilt for the 2003 season, and the pars don't seem to reflect that. And i get substantially different pars for NP and STP. I have trouble figuring how several hundred horses lose 10 speed points by travelling 180 miles north near the end of June.

Yes, the DTV is usually not 0, but i was referring to the average DTV which i believe should be near zero... in the long run.

Calculation of DTV is done by averaging race times vs par times for routes/sprints, Turf/dirt... in the standard way for ClassPar based DTVs. Actually i take Win, place and show speeds vs Pars for each race to get the DTV, in an attempt to minimize the effects of run-away winners etc

Thank you for your response

Fleww

[Macdiarmadillo]

Schultz is correct when he says track pars are not an exact science. Maybe you want to go all the way back and calculate pars rather than using someone else's (Agh! The torture of data entry!) I don't know any other way to filter out possible discrepencies and plain mistakes.

However, I make the case that there is a linear relationship of pars vs. levels. There may be some odd points when you plot these things, but they may be transient. I did this for Bay Meadows many years ago and found $4k claimers had a higher par (yet the horses had a hard time stepping up against $5 claimers). But if you stripped out the older claimers at $4k that were once much better (as coincidentally happened one season), the 4k par fell into line with the rest. There were also some levels that were rarely run whose pars were way off that linear relationship. This was forcibly corrected, i.e. the point on the linear graph was used for par.

For the comment about substantially different pars for the 2 tracks, then we are looking for a universal par chart than track-by-track pars. Try a correlation by purse value instead of claiming value.

I don't have any Stampede charts, but I took a look at Northlandz and see several conditioned claiming races (nw3L, nw2 6m). These should never be included in calculations of pars except as strict categories themselves. These are often specifically written to suit specific horses (say, eliminating competition from potentially better shippers), just real oddballs. You can lump them into straight claimers (no conditions) for handicapping realizing they are normally weak races for the level. Just don't use them for calculating pars.

Average DTV won't be zero, I feel. And I wouldn't try to use this as a check on pars. Sometimes tracks just "evolve" and the average will hover around some other number. Get a large rainstorm, and after drying out, the DTV for that period may perturb around some other number entirely. Ever see this? All other things being equal, particularly purse structure and composition of races (same track secretary helps!), pars can be used for many years, I feel. Others will disagree, but adjustments can be handled by proper calculation of DTVs. Yes, track renovation can screw things up, but compensation for this would normally be taken care of in DTVs, using old pars.

If you have it all automated you may want to take DTV calculation through lower placings. Some races break down where the placings fall to those still able to finish, some horse makes a massive improve and destroys the field by 10 (it happens often enough), a front runner slows the pace so that even he finishes much slower than par. So sometimes it is the perennial loser who runs 46,48,46,44 that is the key to getting the proper variant for that race.

[Brian Flewwelling]

just two of interested in this and i appreciate your thot-full points

I had, still have, a system for Pars and Speed Figs that i initially calibated against BRIS speeds (at 90)... i had a lot of results for the tracks i followed, and used them to estimate my T90s. Where i had lots of results, the sets of Pars and the speeds derived from them were quite fine.

But i started to run into trouble for tracks where i did not have much in the way of Bris data, but by making 'educated' guesses for Pars, and making them fit thru lots of itterations, several came into line and again were fine. My system took a couple days to run the basic analysis and cycles for each track.

But i want a quicker way to line up the times at various tracks... all of 'em. So i bought Dave's pars... where i was familiar with the track, like WO, HST, California ... they were pretty much in line with my old T90s ...so i assumed that with the good work done by Dave, that would save me days of machine time. And that has proven to be the case.

i can loop through all the tracks for the past 3.5 years in about 3-4 min per cycle.

Few changes occur after 10 cycles.

Now my biggest problem is NP, which is just the second half of the STP season. the SAME horse run at each track. During the week including the 21st of June they all travel from Calgary to Edmonton, and i assume it's by trailer. Maybe that's it, they herd them and they are so tired they run 10 points slower:)

since the pars were lower at NP i assumed that the times were wrong so made some changes. i set the T90s higher, so the speeds would seem higher, but that didn't help very much... so is the system i described convergent?? and Why NOT?

Brian

PS the differences are in the open claimers at 10k, 13 k etc

[Macdiarmadillo]

I like the herding theory! Works for me.

Is it just individual horse figs not matching STP or are they actually beating you? And is it across the board or isolated incidences? If some horses started their season before STP, they may be worn-out doggies being kept going until the serious try, which might even be after the NP meet. There may be others saved for the NP meet, which would throw more confusion into what seems to be illogical differences in performances.

Maybe NP has a bad timer, human or electronic. Certainly has been true of GP turf races historically, and I believe it happened this year for a spell also. And those were supposedly checked against a stopwatch.