Sartin Question

[Big Bill]

In the current "Questions for Handicapping Magic Users" thread, a post by Amazin included the following:

" I would be more interested to hear HM's theory on selecting contenders......"

After reading that I recalled that in Brohamer's MPH there was something stated regarding contender selection. So, for the umpteenth time, I pulled my copy and read on page 120 that "PIRCO instructors have developed mechanical procedures for picing contenders."

Not being a Sartin client I was wondering if any on the board could answer these questions:

1. Were the mechanical procedures part of the Sarting software programs?

2. If not, were the procedures made available to clients in print form?

3. Either way, were they effective in contender selection?

4. If the procedures were in print form, would anyone having them wish to share them? E-mail me at bwb900@cox.net to discuss further.

Big Bill

[Tom]

I stopped buying software at Thoromation/KGen, but up until them the guidelines evolved.
The most formal included doing ESP analysis on every horse, then doing APV and Class ratings. Then Form cycle amalysis - +,0,and "plus within an oh." Plus = a good race, Minus = a bad race, and plus within an oh was a race in which the horse showed a move of some kind. Up close at the stretch call was added by Brohamer, and the tandem was also considered a reason to go back, same for high APV was a reason, "plus within an oh" was a reason, the matchup - too fast a pace last time-was a reason. All in all, the whole thing was an exercise in class/form/pace analysis that amounted to a checklist of important things not to overlook. To this, day, I do pretty much that same thing, but with different rating (Ie, HTR class has replaced APV).

Eventuially, the process evolved in to "use the best of last three, similar distance and sufrace." Sartin wrote an article for ATM about this procedure in conjunction with Talbot's Pace Calculator, which is still on their website for free.

Say the last races for a horse were:

6f Dirt 80-13 = 93
7f Dirt 76-23 = 99
8f Turf
9f Turf
6f Dirt 85-16 = 101

If today's race were a dirt sprint, you would choose the best of the last two, or even go back to the 5th back. Best was explained as highest DRF SR+TV number (think this won't work today at any track? Try it frst and see before you say no.) Here, it was either the 2nd or 5th back. You might even go back further is the last race was a plus or plus within an oh race.

If today's race were a grass route, you would use either of the two turf races showing.

[LOU M.]

It's been a while since anyone has given credibility to the DRF SR-TV couldyou expand on the subject, maybe how or if you use the numbers. Thanks,Lou.

[Tom]

Look at Aqu R2 yesterday.
Add the SR+TV together for each of the last three races at a similar distance and surface:

So Social 98 79 Beyers
Mrscopol 104 3rd 76
Silver Ma 95 56
Our Bread 89 57
Pro Motio 100 73
Heavenly 100 4th 66
Lifebythe 103 1st 77
Seattle As 100 2nd 62

OK-you want the top 5 for your contenders.
You have the winner($14.60), exacta($84.50),Trifecta($270.50), and Quinnela($51.00).

NOW...I posted here that the winners was a surprise to me and I didn't use it at all. I bet a non-contender by this method, So Social.
Before everyone postes why the SR+TV is worthless and cannot work, sufice it to say that races like this are plentiful-you get horses that are bad bets a lot of time, but you also get horses that win at good enough odds to make it worthwhile. A lot of horse picked by this have bad Beyers and so the odds are better.
It is easy to do and the real power comes from using pace analysis to seperate the horses.
I'll post later how I use the DRF TV to do that.

[LOU M.]

Thanks for the info.The obvious question is how to go about filtering this info to get a better idea to get on the winners. Any suggestions on how to model this?

[Rexdale You]

Lou,

This system can produce good results.

I pick the top six candidates and try and catch a tri

with it. I do not use a 5F race as the SR & Beyer are

deceiving. It looks like Tom goes back as much as 5 races.

In Toms sample the top #s 180 landed win/show while the

bottom # 162 placed, So it produced contenders.

You should monitor it a few times before using real cash.


Regards

Rex You

[Secretariat]

Doc advocated the SR-TV ratings for a variety of reasons.

1. Everyone knows how they're figured. No guesswork.

2. With Beyers, the DRF-SR-TV would be underplayed more and more.

3. Jim Bayle of Sports Stat did a study which showed the DRF-SR-TV out performing other Speed figures including Bris and Beyer in his tests.

4. Sartin did do a slight adjustment for variants above a certain range both high and low.

Personally, I'd like to see some more udpated work on Speed figure comparisions with just the last race on all ratings to create a level playing field. Same races as well.

DRF+TV
Bris
Beyer
Cramer
Trackmaster
Adjusted Quirin
And any other reputable figures.

Will post some of my own results on Bris and DRF

[cj]

You have to define "outperform" to get an answer. In terms of performance on the track, no way the DRF+TV is better than the others. In terms of value, maybe.

[LOU M.]

I recall someone using an average of the TV as a starting point and creating a variant from that. It was in the track pars I bought from cynthia publishing. I'm sure Dave Schwartz could answer this.

[Dave Schwartz]

Lou,

Actually, I have experimented with the SR+TV method as well as ParSpdRtg+TV method.

No matter how I try, the SR+TV has turned out poorly at all but the very dry tracks (i.e. SoCal, Texas, etc.).

In using the Avg Daily Variant approach, I compared that to using the raw times without any variant at all (just comparing to par). The raw vs par comes out better UNLESS one does a lot of adjusting for day-of-week averages at each track.

So, the raw times is better than the raw times + TV. The downside? It is better in the summer and worse in the winter.

If you think about that, you will realize how logical that is. Speed is more accurate in the summer because it does not take as much effort to deal with off tracks - there are less of them.

Hope this helps.

BTW, in the next issue of the HorseStreet Par Times I will be demonstrating a new way of making a daily variant from the TV. It is SOME work, but not insurmountable. Very logical. (Available from a HorseStreet website near you in January. <G>)


Dave Schwartz

[Tom]

I don't advocate betting off the raw numbers like in my example.
I use that method a lot to get contenders.
I then use the TV to adjust the pacelines and then run velocity numbers off from the adjusted pacelines, AP,EP,SP,FX,LP,etc.
From these, and either a track model or pace analysis, I make my selections.
I alos do not always use the TV the same way. I look at a specific race and decide if I want ot use the full value of the variant or half of it, depending on the type of race, the spread in the raw times, etc. It is just an inherent gut feeling as to when to zig znd when to zag.
I do find it works better at track like FL, Pen, Mnr, etc, where there are a lot of races at a few basic distances.
I don't use this method as much as I used to becasue I am paying $119 a month for data and tend to want to use it. LOL
But like I said before, I like to enjoy handicapping, and this is a lot of fun, I can do it with a form in my car at lunch, or traveling, in airports, where I tend to spend countless hours a year :(

[Secretariat]

Originally posted by cjmilkowski
You have to define "outperform" to get an answer. In terms of performance on the track, no way the DRF+TV is better than the others. In terms of value, maybe.

By "outpeform" Doc referenced the Jim Bayle Sports Stat Report of 1994 (emphasizing ROI- because he often said its not about picking winners, but making money) which ran a test of different services from a month in 1994 - (May 25, 1994 to June 25, 1994 at Arlington (246 races), Belmont (232 races), Calder (207 races), and Hollywood (199 races)) broken down as follows:

Best Last Race Figure and Best of Last Three (which Sartin recommended with the twist Tom said, going back five races max for comparable class, distance, and surface)

DRF SR+TV -
Best Last Act Win% was 21.97%
Best Last Act ROI was -0.17

Best of Last Three Win% was 21.31%
Best of Last Three ROI was -0.04
(Playing 4/1 up horses was +0.13 ROI)

Beyer speed
Best Last Act Win% was 25.24%
Best Last Act ROI was -0.17

Best of Last Three Win% was 24.43%
Best of Last Three ROI was -0.17
(Playing 4/1 up horses was -0.10 ROI)

BRIS
Best Last Act Win% was 24.22%
Best Last Act ROI was -0.17

Best of Last Three Win% was 24.42%
Best of Last Three ROI was -0.16
(Playing 4/1 up horses was -0.13 ROI)

Kuck Ratings
Best Last Act Win% was 23.95%
Best Last Act ROI was -0.18

Best of Last Three Win% was 24.94%
Best of Last Three ROI was -0.13
(Playing 4/1 up horses was -0.06 ROI)

Ragozin Performance Ratings
Best Last Act Win% was 24.48%
Best Last Act ROI was -0.20

Best of Last Three Win% was 27.16%
Best of Last Three ROI was -0.05
(Playing 4/1 up horses was +0.11 ROI)

In terms of DRF-SR-TV, Doc used this in The Follow Up magazine to support his position that (a) the DRF-SR-TV were cheaper than other services, understandable how they are figured, and (c) make as much or more money than other services.

I agree this was a one month study in 1994 which needs reviewed, but it does represent the statistical logic behind Doc's claims, and beleif in using the DRF SR-TV. I through Ragozin in because even though it is more of performace rating for reference. The DRF-SR+TV is unadjusted for class, or any other "power" factors.

I'll be looking forward to seeing Dave's January piece on variants.

[BMeadow]

From the August 2003 issue of Meadow's Racing Monthly:

An Internet poster using the name pmd62ndst decided to check out several months of races in his own database to try to come up with some stats about this question.

He looked at more than 18,000 races and compared the Beyers with the SR + TV. He looked at win percentage and ROI.

For easy comparison, the Beyer figures are listed first, with the SR + TV figures listed in parentheses:

Best
Win % ROI
Career 22% (17%) 0.89 (0.83)
Year 25% (19%) 0.92 (0.86)
3 Months 27% (21%) 0.91 (0.88)
C & Y 27% (19%) 0.93 (0.84)
Y & 3M 32% (23%) 0.95 (0.90)
C & Y & 3M 33% (23%) 0.96 (0.88)

Average
Career 24% (22%) 0.87 (0.88)
Year 26% (22%) 0.87 (0.84)
3 Months 28% (23%) 0.89 (0.85)
C & Y 29% (24%) 0.87 (0.86)
Y & 3M 34% (26%) 0.88 (0.84)
C & Y & 3M 36% (28%) 0.89 (0.84)

Clearly, the Beyers thoroughly outperformed the Speed Rating + Track Variant.

[shanta]

JUST CURIOUS. WAS THE "BEST" THAT WAS LISTED IN YOUR POST ACTUALLY THE "BEST OF LAST 3 RACES PERIOD?" REGARDLESS OF SURFACE, DISTANCE STRUCTURE,ETC. THAT IS THE CRITERIA DOC SET YEARS AGO WHEN SPORT STAT DID THEIR RESEARCH.AT THAT TIME SPEED RATING +TRACK VARIANT OUTPERFORMED OTHERS.
HOWEVER MAYBE TIMES HAVE CHANGED.
RICHIE

[BMeadow]

JUST CURIOUS. WAS THE "BEST" THAT WAS LISTED IN YOUR POST ACTUALLY THE "BEST OF LAST 3 RACES PERIOD?" REGARDLESS OF SURFACE, DISTANCE STRUCTURE,ETC.

Yes, this was "best of" without consideration of surface, distance, etc. The Beyers, in all categories, outperformed the SR + TV in this very large survey.

[takeout]

I'm just asking (because I don't use them) but don't they sometimes change Beyer numbers after the fact depending upon what happens in subsequent races? If so, that would make them look good in retrospect when they perhaps lead to losing bets the first time around. Also, wouldn't that skew a large database study?

[cj]

I don't think they do that anymore, I can't find any cases of it happening.

[sjk]

This is not related to the thread, but if they do not change the Beyer numbers based on what happens when the horses run again, I would view that as a big weakness.

[cj]

Also, if I remember correctly, Beyer didn't adjust a few horses, he adjusted ALL horses. He said he had noticed in making variants that he tended to underestimate the improvement of the winners, so his figures would gradually get smaller. So there was no selectivity on his part, just a method of keeping his figures from "drifting."

[Shacopate]

I would be interested in a study that compared the Beyers to this: DRF SR + TV -17 for sprints and DRF SR + TV -23 for routes.

[Bodacious]

The best of the last 3 was qualified to Distance, Track and Surface and sometimes class. Horses running way above their class levels could post very slow SR. But that was the early nineties when he developed this. With VALIDATOR programs we used the adjusted trackmaster speed ratings. They aren't the the speed ratings you get in the downloads. It's proprietary. Even then you could get away with using the best of the last 3 but commonsense tells one to look for comparable distance and surface.

What Mr. Meadows doesn't mention was that this process was for use in Doc's programs, it wasn't designed for general handicapping. Using the SR+TV was very effective for selecting a paceline for his programs. It wasn't designed to be used as a single handicapping method. (Not that Meadows says it was.)

A much earlier guideline for contender selection was very simple.
1) take horses dropping in class
2) take horses rising in class
3) include any horse that ran 1st, 2nd or 3rd or finished within 4.5 lengths.
4) Showed some moves in it's last race.

These guidelines worked really well with Phase 3, Modern Pace Handicappping. You could also use the +,0,(+) notation but that was more of training exercise. It's way to tedious for regular use.

The Validator programs with the trackmaster downloads really simplified the process a lot. As a methodology contender selection changed over time to match the programs.

[Tom]

When I bought K Gen, the guidelines were to use any paceline with a DRF SR of 80 or better. This worked surprisingly well, especially if the 80 were earned in an ugly race. The stregth was in the programs use of POR as a key component. Tracks like FL, Penn, Del, Tam were ones where this technique worked well.
The other suggestion was to run several pacelines for each horse and visually pick one that graphically looked "normal" for the horse. Then, use the three best early and three best late as your contenders. You had 3-6 horse that way. In fact, this idea works very well in HTR-the top 3 E and top 3 L horses generally contain the winner-or, stated differently, I thorw out any horses not ranked at least a 3 in E or L. Very potent factor.
I really liked the graphs and wish newer prgrams would offer this feature-I see clearer thorugh graphs and charts than raw data.

[hracingplyr]

interesting concept tom i am a new member at htr how do u decide on your final choice using your top 3e horses and top 3 late horses?

bob

[Tom]

Originally posted by hracingplyr
interesting concept tom i am a new member at htr how do u decide on your final choice using your top 3e horses and top 3 late horses?

bob


I go to PL0-manula paceline mode and de-selct lines for all non-contenders. Then I go to the velocity screen and all the show up now are the contenders.
From there, I analyze the velocity read-outs-is there a horse likely to steal the race? Any horse with a fps advantage of .50 or greater gets looked ate real close.

If no standout in F1 is there, I just looka t who the early horse are and how they mathc up in the first two fractions-usually I can throw one or two out here. Does it look like the early speed has nothing left late? Can the closers get there?

To do this, I have 3rd fraction parameters-the base is a fps rating of 50.00 minimum, but that goes lower on lower class races. I look at %Early for guidelines, but am not really comfortable using that figure since it can vary so much race to race on the same horse. But when the odds are high, I tend to use it more. I recently caaught a 18-1 shot at Calder in a cheap race where only two horse met the profile energy demands for the distance, which was around 51.26%-everything else was way high or way low. At 18-1, I was happy to throw out the 3-5 favorite (53.89%)
dutcht the 18-1 and 5-1 shots (51.33 and 51.66).
Had the 18-1 been 2-1, I would have not bet it based on energy.

When nothing stands out on the match up, I use the optimal pace theory from Randy Guiles, modified like this:

F orE horses - must be 1 or 2 early and at least 5 late
EP horses must be either at least 2-4 or 4-2
P horse must be either at leat 2-4 or 4-2
S or L horses must be 1 or 2 late

E, EP, of P horses ranked 3-3 can be played.
S or L horse are never prefered over qualified earlier horse, unless the odds warrant a two horse bet. They are always in my exacta boxes, though.

I am experimenting using the 4 Qs of handicapping odds templates to make a line on my top three or four horse based on velocity, but that is pretty new and I don't have anyhting to offer in that vein yet.