My rating process

[jeebus1083]

One of the things that I enjoy about spreadsheet handicapping is that you can take numbers and crunch them. For the harness spreadsheet, I employed a modified version of the Dick Mitchell "Power Rating" methodology.

Basically, you take a factor and break it down in this fashion:

Weighted factor = 60 + 40 ((raw factor - lowest raw factor)/range of factors)*%weight

Example: if you did a rating that was 100% weighed on driver UDRS, the highest UDRS would score a 100, while the lowest in the race would score a 60.

If you are rating a group of factors, you have to give each component of the group an assigned weight. The sum of those weighed components would be your group score.

In my spreadsheet, I weigh 5 different groups of factors: 1) Race fractions, 2) Compound ratings derived from the fractions of the race, 3) the front, middle, and back halves of a race, 4) final time, and a compounded rating incorporating the final quarter, the 3/4 mile time and final time, and 5) the driver/trainer factor.

To break each component down:

Race Fractions (Feet per second: 1320 feet/fraction time)
1Fr=10%
2Fr=20%
3Fr=30%
4Fr=40%
TOTAL: 100%

Since harness races are basically late-fraction heavy, I wanted to weigh the later portions of the race heavier as the race goes longer. In Mitchell's book, each fraction was weighed based on distance (furlongs of call/total distance). For a standard 1 mile race, each fraction would be worth 25%. However, as one poster pointed out to me in an earlier thread, the tempo of thoroughbred racing differs greatly from that of harness racing. I felt that the 10-20-30-40 progression was a fitting compromise.

Compounded Figures (Sartin-style ratings)
EP (front (1Fr+2Fr+3Fr) 3/4 mile time)=25%
SP (average of final 1/4 and front 3/4 mile time)=25%
AP (average of EP and SP ratings)=25%
LP (back (2Fr+3Fr+4Fr) 3/4 mile time)=25%
TOTAL: 100%

Front/Middle/Back Halves (2640 feet/fraction)
Front Half (FH)=1/3rd of 100%
Middle Half (MH)=1/3rd of 100%
Back Half (BH)=1/3rd of 100%
TOTAL: 100%

Final Time Factors
Final Time (5280 feet/final time)=50%
Compounded Final Rating (average of front 3/4, final quarter and final time)=50%
TOTAL: 100%

Driver/Trainer Factors
Driver UDRS: 75%
Trainer UTRS: 25%
TOTAL: 100%

Perhaps I weigh the driver factor too heavily, but their tactics more often than not influence the outcome of a race. There is a reason why the good drivers have high UDRs and why the not-so-good drivers have low UDRs, and it is often because the better drivers make the better decisions.

Now that you have your five factors, now the scores get broken down further. By default, I weighed each factor 20% of the final score to give it balance. If a horse scores 100 on all factors (equates to 20 points per factor), his overall rating would equal 100.

Once the overall "power rating" is arrived at, we go a step further. We now have to break each power rating down into an "ability rating":

Ability = .0245 x Power Rating - 1.46

Basically, a horse with a perfect 100 would get a .9900 rating, and a horse with a 60 would get a .01 rating. It is a simulation of the best horse running against the worst. Each horse is assigned an ability rating based on its Power scores.

Once the ability ratings are arrived at, they would then be converted to %probabilities. Simply put, divide the horse's ability rating by the total sum of the ability ratings. That gets turned into a percentage, which then is converted into odds.

Since not all horses in the race are contenders, you will eliminate horses. In my spreadsheet, as I explained in the "Need Help" thread, by typing in "YES" or "NO" in the specific area, you are able to choose whether or not a horse contends in this race. By selecing "NO", this will eliminate their ability score from the tally, thus changing the calculations for the remaining horses.

So there you have it. This is how I arrive at my lines.

Any suggestions?

[Sinner369]

You fail to mention how well (accurate) does your method..........works?

[jeebus1083]

To be fair, it takes a lot of practice and good paceline selection to achieve a respectable level of accuracy. I have been tinkering with my spreadsheet for weeks, and where I was going wrong was not manually selecting my contenders and having the computer set so that horses who met a cutoff qualified. This resulted in runners who really were not contenders being listed as such. Basically, it takes time, skill and practice to hone the line to something that resembles accuracy.

With that said, I may add an "override" function so that the user can plug in his own % probability for the race. Sometimes, you may disagree with the computer, so you can formulate your own sense of value.

[6furlongs]

Sounds like you're on the right track. I do something similar with the thoroughbreds.

[markgoldie]

Look. Please take this in the spirit of helpfulness in which it is written. You seem like a highly-intelligent guy who is or is about to spend a great deal of time and energy on a spreadsheet project. But it won't work. Why?

The short answer is that success in this game is far beyond the arena of assessing power ratings for horses. So slicing and dicing all the relevant factors will lead to a dead end. The harness-betting public is smart. Very smart. And that's because they have undergone what might be termed a Darwinian culling that leaves the best and brightest standing while the lesser-lights fall by the wayside.

The betting public knows almost without fail who the best drivers and horses are in a given race. So getting an advantage over them is extremely difficult. This means that any chance of long-term success is dependent on situational analysis of a race. For want of a better term, it involves the development of specific angles and using them in the proper race scenarios.

Power ratings won't accomplish this. In particular, there are problems with your stated approach, but even so, I say this on a generalized basis. There is one North American track where they have a fighting chance and that is Balmoral Park, but even there it is difficult. At other tracks, trip and positioning will be more important than the straight power of the animal.

As for specific problems with what you have stated. First, back-fraction analysis is non-instructive. S types are almost always disadvantaged, particularly from outer post positions, and it doesn't matter how fast they can travel last quarters, halves, or even back 3 panels. First-quarter analysis is instructive because the vast majority of winners will come from on or near the front. Most of all, never consider back speed when it is accomplished with benefit of cover. This is a trap which should be exploited on the reverse side of the coin. Fast last quarters with cover are over-bet if anything.

Driver-strength analysis is fruitless because the public knows who the best drivers are and they bet accordingly. If you want to something serious with driver ability, study driver changes only as well as situational performance of inferior drivers. You'll make far more money studying when and how you can use bad drivers simply because they are under-bet in almost all cases. Same with low-percentage trainers.

Analyze change as a key to potential profitability: Changes in driver, trainer, track, equipment, and medication.

Success is tough and the only way is through situational analysis combined with situational use of profitable angles. In sum, don't waste your intelligence and energy on a dead-end quest. Look for the "game within the game" and you'll be ahead of the game.

Take this for what it's worth and good luck.

[LottaKash]

Originally Posted by markgoldie

Look. Please take this in the spirit of helpfulness in which it is written.
For want of a better term, it involves the development of specific angles and using them in the proper race scenarios.

Driver-strength analysis is fruitless because the public knows who the best drivers are and they bet accordingly. If you want to something serious with driver ability, study driver changes only as well as situational performance of inferior drivers. You'll make far more money studying when and how you can use bad drivers simply because they are under-bet in almost all cases. Same with low-percentage trainers.

Analyze change as a key to potential profitability: Changes in driver, trainer, track, equipment, and medication.

Success is tough and the only way is through situational analysis combined with situational use of profitable angles. In sum, don't waste your intelligence and energy on a dead-end quest. Look for the "game within the game" and you'll be ahead of the game.

Take this for what it's worth and good luck.

MG, couldn't have said it any better.... ...You're such a smarty pants......I like that, and your willingness to share it...

best,

[jeebus1083]

I respectfully disagree that the use of Power Ratings is a fruitless exercise in futility. If anything, it's not a lone handicapping tool, but part of a process. I know that there is more to handicapping than crunching the numbers in a computer. I know that you have to examine many facets of how races are run. What the Power Ratings are supposed to do, once your contenders are narrowed down, is produce a line for those contenders. It is not a method of contender selection, but rather a method of applying "fair" value to the contenders.

As with anything, record-keeping is key. If the records suggest that I'm over-valuing or under-valuing horses, I have to make adjustments. Nothing is ever perfect.

Am I listening to what you have to say? Yes.

However, if I'm going to pay tuition for my harness racing education, I'm going to do it on my own device.

[LottaKash]

Originally Posted by jeebus1083

I respectfully disagree that the use of Power Ratings is a fruitless exercise in futility.

Am I listening to what you have to say? Yes.

However, if I'm going to pay tuition for my harness racing education, I'm going to do it on my own device.

Hey Jeebus, I don't think that that was exactyly what MG was trying to say....Hey, we all use numbers....I know I do, and they are my first filter in my handicapping....Still, the "angles" and "situations" are where the real-juice is....

In my way of going, the numbers lead me to the ones to look at firstly, but I don't just use the numbers without some additional "edge"....As I see it, if you want to succeed in the long haul, you must find an edge....Numbers all by themselves are ovebet and usually don't offer much of an edge, for as MG stated, there are some very serious players all using the same things...We must look elsewhere for our edge....

My philosophy is: "No Edge, No Bet", "naked numbers" just don't make it for me...I don't get to play as much, but I still remain strong in this game just doing my thing, my way....

I applaud you for your passion in creating and tweaking your spread, I hope you find the "key to the mint" by your efforts...

[harness2008]

Trackmaster as a third party provider enables harness fans to purchase power ratings which have been created by various sources such as Compubet and Trackmaster Greg. These I'm thinking are along the sames lines as your spreadsheet method in that various handicapping factors are given a certain weight and voila, ratings are provided.

Check this out though. While some tracks provide a flat bet profit in some areas, there are others that do not and that was echoed by MG. The Compubet ratings themselves are ok I guess but I have been always concerned why some tracks at times throw out a flat bet profit on say the top rated horse while other tracks do not show a positive ROI. I mean if you have an automated computer utilizing the same handicapping factors for each track the results should generally be the same.

I would think that if a model such as Compubet which analyzes the same handicapping factors across the board for all tracks, would generally show similiar results for all tracks. But the answer is they don't, so there has to be something else determining the winners at tracks besides just the number crunching.

I myself love the numbers and am not disapproving of your spreadsheet method at all but we're just suggesting that perhaps there needs to be something else that should be included beyond the number crunching and I think that MG and Kash hit it right on the head.

[jeebus1083]

Originally Posted by harness2008

Trackmaster as a third party provider enables harness fans to purchase power ratings which have been created by various sources such as Compubet and Trackmaster Greg. These I'm thinking are along the sames lines as your spreadsheet method in that various handicapping factors are given a certain weight and voila, ratings are provided.

Check this out though. While some tracks provide a flat bet profit in some areas, there are others that do not and that was echoed by MG. The Compubet ratings themselves are ok I guess but I have been always concerned why some tracks at times throw out a flat bet profit on say the top rated horse while other tracks do not show a positive ROI. I mean if you have an automated computer utilizing the same handicapping factors for each track the results should generally be the same.

I would think that if a model such as Compubet which analyzes the same handicapping factors across the board for all tracks, would generally show similiar results for all tracks. But the answer is they don't, so there has to be something else determining the winners at tracks besides just the number crunching.

I myself love the numbers and am not disapproving of your spreadsheet method at all but we're just suggesting that perhaps there needs to be something else that should be included beyond the number crunching and I think that MG and Kash hit it right on the head.

I think people are mis-understanding me. I am going beyond number-crunching with the rest of the handicapping. This is merely a TOOL to build an odds line with, once I've narrowed through the minutia of the PPs and have narrowed a field down to 1, 2, 3 or 4 runners. Of course I realize that I may love a horse that looks overpowering, yet he comes up 3rd or 4th when I apply my contenders to the spreadsheet. Having said that, I'm going to have to re-examine a way to make subjective changes to my odds line when I disagree with what the computer is spitting out.

[xfile]

Does anyone mess with "Turn-Time" in harness racing?

[jeebus1083]

Originally Posted by xfile

Does anyone mess with "Turn-Time" in harness racing?

IMO, "turn-time" is very important in harness racing, especially when the horse doing the moving makes the brush uncovered. You're not always going to see the rail horses go indian-file for 7/8ths of a mile, and see the pocket rocket take advantage of his perfect 2-hole trip and scream down the open stretch. Horses making their moves from the 1/2 to 3/4-pole uncovered merit attention. They are generally running their fastest fraction of the race, and are doing so without being sucked in by live cover. Of course, horses with cover may have more strength from the 3/4s to the finish, but that's another perfect trip scenario. Many horses cannot sustain that uncovered move from the 1/2 to the 3/4s, as it's just too taxing, so a horse who has demonstrated strength with that move, AND has a strong 3rd quarter time and/or 3/4 mile time is showing some good ability.

Of course, put a loaded gun in the wrong hands and you could have disaster. Some drivers just can't get the most out of a horse 1st or 2nd over without cover in the 3rd quarter. The key is TIMING. Give me a capable (not necessarily a leading) driver who has good reflexes, and has a sense of pace, and he'll guide a good horse to a strong 3rd quarter. Hopefully at that point, he's pulling away.

[markgoldie]

Jeebus:

What you are trying to do is logical. In fact, very logical. And I understand the impetus and background that has led you to this effort. In short, you are looking to analyze the combinations of factors which will allow you to create a fair value line on the race, which will then (presumably) be exploited to take advantage of overlay situations. Such overlays might be exploited in the win pool. However, this is highly problematic because the late-money odds' fluctuations at harness tracks are virtually insurmountable, with the possible exception of The Meadowlands. Still, you might consider exacta-pool anomalies and this is possible if you have a wagering nimbleness, by which I mean the ability to assess value and make late-minute wagers.

In your quest for spreadsheet accuracy, you will look at long-term results with an eye to tweak the weighing factors which comprise it. The idea here is that you want the program to better reflect the realities of actual outcomes. Only the empirical data will be important, because as a scientific observer, you will allow the results to dictate the reality of what's actually happening on the racetrack. That is, preconceived notions will not cause you to decelerate the learning curve.

Okay. Think I've got the scenario pretty well?

So what am I saying to you? Well, first off, I have 45 years of serious wagering of harness horses on which to base my opinion. Frankly, the sport has changed and so the processes I was using many years ago no longer have relevance. On the other hand, just seeing these changes and adapting to them has something to be said for it. Also, I have been blessed (or maybe cursed) with an analytical mind. By that I mean I always look at results with an eye toward understanding the underlying principles of what I witness. Hopefully, these observations have led me to some understanding of pattern recognition- that is, the ability to relate a current problem (race) to the general pattern of race scenarios into which it might fit.

Fine. So what have I learned? Enough to know that, as I said, the "line" at most harness venues is pretty tight. Not a lot of yokels in bib overalls making the odds. And so, maybe 95% of all the tinkering that you will do with factor-dicing, even after months and possibly years of concerted work will be incorporated in the wagering odds. (And I'm being generous in assuming that you will eventiually derive a 5% loophole, so to speak).

And that's why I say what you are attempting is relatively futile. Will you learn some things? Sure. You bet. But I would respectfully recommend that if you want to learn faster, you would be better off not trying to out-perform the existing line, but rather analyze the line as it is and try to understand where and why it may fail. Ask yourself: What are the scenarios in which the existing line is likely to be wrong? Why did that 3-5 favorite fail? Why did that 15-1 shot win? Is there a pattern by which some of these failures and success can be predicted beforehand?

Another way of putting it might be: Don't try to build a mousetrap from the ground up because it takes a lot of time and energy. And then when you're done, you'll probably find that the ones already invented are better than yours. Instead, study the best mousetrap available. Find out where and how it may sometimes fail. Look for complex and hidden scenarios.

So basically, I'm saying that I have been relentlessly studying the existing mousetrap for years. I've learned some things and I've "been there, done that" on many variables. You may refer to my earlier post to save yourself some time in ruling out and ruling in certain areas of inquiry which may prove fruitful.

That's all. Just trying to save you some time. Good luck.

[MiJan]

you would be better off not trying to out-perform the existing line, but rather analyze the line as it is and try to understand where and why it may fail.

While I totally agree with the above statement I believe that what Jeebus is attempting is to first build his own line to compare against the oddsmaker's (track or other thrid party) line. Where his line does not agree with the oddsmakers, he then looks for the reason why.

If this is so, I don't see this as a waste of time -- especially if he has it computerized. I have written software that creates a line and compares it to the oddsmakers line. It has worked well for me over a good number of years. However, tweaking this software was necessary over those years for many of the reasons you point out in your message (change).

The quote (above) from your message has more worth than I have seen in print in a good long time. However, I still believe that creating ones own line to compare is a worthy effort.

[Ray2000]

Originally Posted by jeebus1083

So there you have it. This is how I arrive at my lines.

Any suggestions?

Since you asked for suggestions.

FWIW

Almost all computer programs, spreadsheets, and paper and pen methods end up by assigning some numerical rating to each horse. Expressing that number as an Odds Line can be tricky.

Using the individual rating divided by sum of ratings is dependent on the size and variance of that sum and subtracting some constant is guesswork.

The best method I've found (maybe invented) is to subtract the average rating from each individual rating and then divide by the standard deviation of all the ratings. Call this answer "R" and use these formulas:

A * e^(B * R)

For 10 horse field A = 8.8761; B = -.700
For 9 horse field A = 7.7386; B = -.754
For 8 horse field A = 6.9319; B = -.772
For 7 horse field A = 5.7476; B = -.728
For 6 horse field A = 4.9607; B = -.770

e = natural log or 2.7183

The attached spreadsheet shows the odds for any ratings you might have. The sample ratings used are taken from Trackmaster Gregs picks yesterday for Chester R12. (Free sample post)


5 5 IT'S COAST TIME 5/1 Lachance, Mike - 10 533.16
4 4 BLACK TIE 6/1 Callahan, Corey - 10 527.73
7 7 LORD BURGHLEY 10/1 Gingras, Yannick - 18 517.60
6 6 COUNSEL 8/1 Morgan, Anthony - 9 517.10
1 1 LUKAS ROSSI 9/2 Napolitano, Geo. Jr - 23 513.26
8 8 KATIE'S KISS 12/1 Tetrick, Tim - 19 500.22
3 3 POWERFUL COMMITTEE 4/1 Miller, David - 17 478.15
2 2 ZERO BOUNDARIES 3/1 Pierce, Ron - 13 453.76


I don't play any race with Field sizes less than 6 or more than 10 so I've never worked out the constants for them. You can estimate those constants from the trend. Hope this helps some....