|
JCapper Message Board General Discussion
|
| Home |
Register
Log In |
| By | Data Base - MatchUp Question | |
| jeff 11/7/2008 2:26:12 AM | --quote: "Jeff, this question was posed on another BBS. Is this possible using your software with it's DB capabilities? Thanks for any information.-----Mike.--end quote Received the above question from someone who posts at Binder's Pace and Cap (Sartin) Board via email on Thurs. Thought I would take a stab at it. Here's the answer I sent back: Hi Mike, The "gotcha" that you threw in is the phrase "Race projection other than early." What I have to say might ruffle a few feathers because it goes against the grain of current discussions on Sartin match up concepts... But my own research very clearly indicates the most accurate means of projecting a race as either "early" or "other than early" has far more to do with the surface being either speed friendly or speed tiring than the amount of pace pressure within a race or the way that the horses within the race match up. On speed tiring surfaces early horses tire early and spit the bit. On speed favoring surfaces even cheap speed has the ability to wire almost any field. My research very clearly indicates two long term patterns as the public bets the races: 1. Early speed is consistently ignored while late speed is consistently overbet. 2. Form or readiness is a measurable commodity and the public has very little idea of how to go about doing that. Workouts can be a good way to identify readiness. Horses with 5f or longer bullet drills are frequently ignored by the public. Yet if the same "move" were made in the afternoon during a race the public would be all over the horse. Ok. That said, back to your question: "How often does the early horse win?" I'll try to run a few data samples and shoot them your way (time allowing) later. -jp .
| |
| jeff 11/7/2008 3:36:54 AM | Before I start posting actual data, a few explanations about the data itself might be in order. First, all data used in this study was compiled using Bris DRF Single Format Data Files. All races used in the study were run during the first 3 calendar quarters of 2008. Second, in the results presented below, I didn't make any attempt to pre-project races as being "early" or "other than early." My philosophy has always been to basically let the data point the way. Second, the fields in my database do not include raw fractional time. There are fields for E1 and E2 pace figures (Bris) which are based on raw fractional time and variant. For purposes of this study I used a number in JCapper known as CompoundE1. CompoundE1 is generated by an algorithm I wrote designed to pick relevant pace lines found in a horse's record. Understand that pace lines picked by the algorithm may well be different from lines that individual handicappers might pick. The algorithm is 100 pct mechanical. And like most algorithms it won't ever get flustered by a rash of recent bad results. The algorithm itself is the result of several hundred hours of R&D time spent at the Data Window by yours truly during the latter part of 2007. CompoundE1 is (I think) the best number in JCapper that can be used to approximate "1/5 of a second." Constraint: Two or three early horses in a race with first call times within 1/5 second of each other. I purposely ignored the run style aspect of this constraint. I decided I wanted to see results for any and all horses within "1/5 of a second" of each other no matter what their run style might be. I promise I'll break out the results by run style later. To satisfy the "1/5 of a second" requirement I merely grabbed those races where two or more horses that had previously run existed where CompoundE1 was within 2 points of each other. The phrase "previously run" means that the study does not include data for races full of first time starters. Why 2 points? I found the following explanation in the Library section of the Bris site: "the BRIS Pace Ratings use a fixed scale of 2-points-per-length for all pace calls." Going on the assumption that 1 length equals 1/5 of a second I used 2 points of CompoundE1 as my dividing line. Continued... -jp . ~Edited by: jeff on: 11/7/2008 at: 3:36:54 AM~
| |
| jeff 11/7/2008 5:33:51 AM | As promised here is the data itself. First, to give perspective to the numbers that I'll post shortly, here are all starters during the time period of the study:
Now, here are all horses meeting the constraint of being within "1/5 of a second" based on there being at least 2 horses in a race where CompoundE1 gap for those horses is within 2 points:
The first thing that catches the eye is that the above sample (restricted to horses that are within 1/5 of a second or grouped within 2 points of CompoundE1) outperforms all horses that ran within the time period of the sample. Not just for win rate but roi-wise as well. This is to be expected. Almost every large sample based on early speed that I've ever looked at does this. Simply put: The public almost always underestimates early speed. Be aware that the "breakout" numbers I post from this point forward in this thread are all based on the above table for "1/5 of a second" horses. Here is data for for "1/5 of a second" horses broken out by CompoundE1 Rank:
What should catch the eye here is that even though the horses in the sample are all within 1/5 of a second of each other, the data shows a nice orderly distribution. The top ranked horse wins more often than the second ranked horse and so on right down the line. That should tell you that being within 1/5 of a second of each other does not make them equal. Keep this in mind as you start to ask yourself "Which of these is the most likely winner?" Specifically take note of the top 3 vs ranks 4 and lower. jp . Continued... ~Edited by: jeff on: 11/7/2008 at: 5:33:51 AM~
| |
| jeff 11/7/2008 4:59:37 AM | Now, here is the data broken out by run style:
Note that these are Bris run styles which may be quite different than either Sartin run styles or run styles assigned by an individual handicapper. Despite that, note the difference in the numbers for E, EP, & P runners vs S and NA runners. The original question was sent to me by a Sartin handicapper looking for hard data to be used to help him match them up and decide which (if any) might be worth betting. I'd be remiss if I didn't point out that JCapper has a factor called CompoundPaceFit which is a number produced by an algorithm that does just that. Here is the data broken out by JCapper CompoundPaceFit:
Note the win rate and overall roi of the top 3 vs ranks 4 and lower. In my initial answer I replied that Form was an important separator. Here is the data broken out by JCapper Form Rating:
Note the win rate and roi for the top 2 in Form vs ranks 3 and lower. Also note that near break even play appears possible with these horses when Form Rating is >= 90. Here is the data broken out by JCapper BasicFitness:
Note the win rate and roi when BF is >= 20 vs the same for those below 20. -jp . To be continued... ~Edited by: jeff on: 11/7/2008 at: 4:59:37 AM~
| |
| jeff 11/7/2008 5:21:19 AM | In my initial reply I said that track surface plays a big part in projecting "early" vs "other than early." Off the top of my head tracks that ran noticeably other than early this year were TPX, SAX-OSA, and SAR. Here is the data from Dirt races (in the case of SAX-OSA dirt can include bits of rubber and carpet fiber) at TPX-SAX-OSA-SAR:
Note the reduced win rate and roi for these horses on speed tiring surfaces vs win rate and roi for the overall sample. -jp . To be continued...
| |
| jeff 11/7/2008 1:23:47 PM | Some more Form based numbers... Here is the data broken out by Workout Brilliance:
Notice the win rate and roi for those rated at 54 or higher vs those below 54. Here is the data broken out by Race Brilliance:
Note the win rate and roi for those rated at 57 or higher vs those below. -jp . To be continued...
| |
| jeff 11/7/2008 1:28:32 PM | What about trainer intent? Here is the data broken out by CLEAN/DIRTY and Route Debut:
Note the win rate and roi for CLEAN horses vs DIRTY horses. Also note what happens when "1/5 of a second" horses make their route debut. -jp . To be continued
| |
| jeff 11/7/2008 1:48:47 PM | Let's have a look at a few other JCapper factors. Here is the data broken out by PMI, CMI, and Optimization Points:
Note the win rate and roi of the top ranked horses for these factors vs win rate and roi for the lower ranked horses. Keep in mind that PMI emphasizes early speed and current form... CMI emphasizes class and current form... while Opt Points emphasizes both. -jp . To be continued...
| |
| jeff 11/7/2008 2:01:26 PM | Ok. That's some basic data. To recap, the study looks at multiple horses grouped within 1/5 of a second of each other using JCapper CompoundE1. From the above I'm going to attempt to draw a few conclusions: 1. Early speed is still underestimated by the betting public. 2. Being within 1/5 of a second of each other on paper means these horses are very close in ability. However (my opinion) very close in ability is not quite the same thing as equal in ability. Data in this study suggests there are ways of separating horses that are very close in ability. 3. Form appears to be one way of separating horses that are very close in ability. 4. Trainer intent also appears to be a way separating horses that are very close in ability. 5. Surface (if you are able to identify it ahead of time) can have a profound effect on the ability of early speed horses to finish their races. Looking at the data I see several areas that would appear to offer solid UDM potential. For those who want to pursue things further and create a UDM or two - have at it. -jp .
| |
| JustRalph 11/7/2008 4:14:54 PM | after reading this thread, I realize one thing............ MY UDM's Suck!! now wonder my horses keep stopping in the lane Back to the drawing board...........LOL!@ ~Edited by: JustRalph on: 11/7/2008 at: 4:14:54 PM~
|
| Reply |
|
Copyright © 2018 JCapper Software back to the JCapper Message Board www.JCapper.com |