|
JCapper Message Board General Discussion
|
| Home |
Register
Log In |
| By | Track profile theory | |||||||
| bill uminn 6/19/2008 10:04:34 PM | I have really enjoyed playing with the new addition to jcapper, and am anticipating the update including the 999 days back, in part because I struggle with the concept of focusing on recent bias profiles. So, heres my thought/question. Is it possible to prove (at least to me) that looking at a 5-7 day track profile (or even a 30-60 day) to determine an immediate bias to incorporate into one's wagering is a sound approach? I have never understood the validity of this form of analysis, but would really like to. I have always failed to see the line that exists between a so called "bias" and a simple statistical variance that happens with every factor we have. Heres my thought pattern: if a particular factor is running hot at a certain track for a specified of time, and that indicates that factor will win in the immediate future...then...when you bet that factor in the immediate future it will do one of two things, win or lose. If it wins, then it becomes the factor that is "hot" for the new set of 7 days, and the cycle would seemingly be endless if in fact looking at a 7 day cycle accurately predicted the current and immediate future track bias. If it loses, then one is left uncertain to whether this approachis valid. Not sure if that is clear. I really want to advance in my track specific handicapping, but am struggling in a theoretical approach to doing it that seems vaild. Thoughts?
| |||||||
| jeff 6/20/2008 2:11:48 PM | Some thoughts on theory first... The first order of business is knowing what a bias is and what it isn't. To my way of thinking there are two types of bias: 1. Early vs Late 2. Path Each of these two types can further be categorized as: 1. Long Term 2. Short Term Long Term vs Short Term Whenever I spot a bias, the first thing I want to know is whether it's long term or long term. I label a bias as long term if it shows up in the numbers from a 90 day or longer snapshot. I label a bias as short term if it shows up in the numbers on a track profile report covering a shorter time period. Spotting a Bias Some players can look at the same 90 day report I have in front of me and not see the same bias that I see. The best way to spot a bias, IMHO, is to compare the numbers on a single surface single distance track profile report for a specific time period to the numbers on a track profile for all horses everywhere for a much longer time period - or the ALL button on the Data Window. It's the degree of difference that defines a bias. If the numbers are essentially the same as what you find with the ALL button then there isn't a bias. But when the numbers are different - then you have a bias. Let's look at some numbers:
My read: This is very similar to what you might expect to see in a large data sample... about 7300 races are summarized here.
My read: Early horses enjoyed a strong advantage - different than the larger ALL button pattern. Note the pct winners distribution among run styles. Also note that CompoundAP123 is slightly stronger than CompoundSP123.
My read: Early horses were at a severe disadvantage. Note the pct winners distribution among run styles and CompoundAP vs CompoundSP. I've (hopefully) been able to shed some light on identifying an early vs late bias. Like anything else, doing that gets easier with practice. Once you've spotted an early or late bias how do you go about exploiting it? To be continued... -jp . ~Edited by: jeff on: 6/20/2008 at: 2:11:48 PM~
| |||||||
| jeff 6/20/2008 3:14:08 PM | Let's go after the early bias on the AQU inner first. The first 10 days of 2008 produced a Track Profile for 6f that looked exactly like this: http://www.JCapper.com/HelpDocs/AQU_6fd_10days.html Look at the pct distribution for run styles and CompoundAP vs CompoundSP. My read has to be that 6f on the inner dirt surface is running early. I can create a simple UDM to take advantage of an early track profile... CPace rank=123, JPR rank=1, running style = E-EP. And running that UDM against the first 10 days of 2008 we get the following:
Not bad right? But the REAL question becomes: Will it hold up going forward? Track profile theory says it should - provided it is based on causal factors (a long term pattern) and not just variance or statistical noise. Here's that same UDM from 1-11-2008 through the end of the quarter:
Obviously there's some very real things at work that are causing this... depth/consistency of the surface, etc. Watching the races run there this past winter told me (visually) that horses on or near the lead weren't getting tired... and the data pretty much bears that out. IMHO this is about the furthest thing possible from sample noise or statistical variance. -jp . ~Edited by: jeff on: 6/20/2008 at: 3:14:08 PM~
| |||||||
| bill uminn 6/20/2008 10:27:24 PM | Thanks Jeff, I appreciate your time and thoughts - i look forward to progressing in this area.
| |||||||
| ryesteve 6/21/2008 10:54:28 AM | If I may, let me describe where I run into trouble with this stuff: In Jeff's example, the forward results looked very good because the Aqueduct inner really does have a speed bias. The problem is that when you see promising results such as this, the natural next step is to look for 10 day stretches at ANY track that look similar to that initial 10 day stretch on the Aqueduct inner. When you do, you will encounter some tracks with a legit speed bias, but a lot of the time what you'll be tagging are, as Bill said, the results of short-term natural variation, or perhaps an actual short-term bias that will not persist. So, if you were to look at forward results in ALL these situations, the results would tend to be poor, because you don't have enough instances of true biases that will persist... and given they ways I've looked at this up until now, the losses incurred by these "pseudo" biases eat up all the profits attained by the real biases.
| |||||||
| jeff 6/21/2008 4:18:06 PM | Steve, that's a very valid point. I don't know if I can quite put it into words but I'll try. The trick, IMHO, is to go beyond the numbers. There are visual clues you can pick up as you watch the horses run - how they win races on any given day. During post parades and warmups, on a speed biased surface, horses hooves tend to not sink in very far. Sometimes it almost looks like they are walking on a sidewalk. On a speed tiring surface, they tend to sink in deeper as they walk. Think of a plowed field. When horses run on a speed biased surface you can see the leader(s) still full of "run" in the stretch. When horses run on a speed tiring surface, you can visibly see the leader(s) shorten stride in the stretch. If you photograph the field from a helicopter hovering 500 ft above the finish line at the moment the lead horse hits the wire you will often see two very distinct patterns in the positions of the field:
The line _______ represents the rail and the | character represents the finish line. Each > represents the position of a horse at the moment the winner hits the wire. The top diagram is what you generally see on a speed favoring surface. The lower diagram is what you generally see on a speed tiring surface. Also, keep in mind that wind speed and direction can create a bias. Sometimes you might THINK it's the surface. But when it's the wind and the wind direction is different the next day: look out! More than once I've cashed in on a bias simply by looking for a strong head or tail wind by watching for the wind "grabbing" horse's tails and manes during post parades. And believe it or not on really windy days I've actually gone so far as to use Google to get satellite images of tracks that I play and correlate the track layout to wind speed from that day's weather report and match that to the track profile for that day. I've had some incredible success in the springtime doing that at AQU, HAW, and TUP. I'll admit that much of this is art. When a horse or two wires a field and I don't see any of the visual clues I've related then I have to consider the possibility that the horse was just better than the others and the rider sent him to the front. But if you go after it hard enough you can (eventually) get to the point to where you can be confident about knowing a speed bias when you see one. -jp . ~Edited by: jeff on: 6/21/2008 at: 4:18:06 PM~
| |||||||
| ryesteve 6/22/2008 7:38:18 AM | --quote: "The trick, IMHO, is to go beyond the numbers"--end quote That's exactly right. Thinking back at what I originally wrote, "It works at Aqu because there really IS a speed bias there", led me to ask myself, "Well, how do I KNOW?". And like you said, it's not just the numbers, it's the intangibles. The best answer I could give myself to that question was, "You just know!" :) Obviously not helpful if the goal is to build an objective methodology, the feasibility of which I've wondered aloud about in an earlier post. My goal has always been to take myself out of the equation to as great an extent as possible, but this is one area where I suspect it'd be very, very difficuly.
| |||||||
| bettheoverlay 6/22/2008 11:11:46 AM | I've been working on a slightly different approach to track profiling. I don't know how much long range validity it has but I thought I'd pass on an example. Instead of looking for what is winning I'm looking for what isn't winning. Trying to establish minimum numbers in pace factors. In the following examples, UserFactor2 represents a form modified CompoundE2, UserFactor3 represents CompoundSP, and UserFactor5 represents CompoundLate. I further modified sprints by CompPaceFit rankings and routes by CompoundAP rankings. Here are WOX routes from the beginning of May
Here are the last 11 calender days
Late speed has been more crucial lately, only 4/54 at User5 60-69. So if I modify User5 to a minimum of 70.
|
| Reply |
|
Copyright © 2018 JCapper Software back to the JCapper Message Board www.JCapper.com |