Probability percentage from a rating

[cj]

I looked over my code and there are a couple other tweaks I made over the years. If anyone is interested say the word and I'll post. None of the change the basic idea.

[Magister Ludi]

#~~~a~~~~~b~~~~~c~~~~~d~~~~~e~~~~~f~~~~~g~~~~~h
9~~105.5~~.111~~~.272~~~-.159~~~1~~~~.405~~~.135~~~.294
3~~104.3~~.110~~~.169~~~-.060~~~4.1~~~.159~~-.011~~~.049
2~~101.1~~.107~~~.129~~~-.023~~~6.9~~~.103~~-.027~~~.004
5~~~89.9~~.095~~~.110~~~-.015~~10.9~~~.068~~-.042~~~.027
4~~~87.72~~.092~~~.087~~~.005~~13.6~~~.055~~-.032~~~.037
8~~~79.96~~.084~~~.064~~~.020~~14~~~~.054~~-.010~~~.030
10~~77.95~~.082~~~.052~~~.030~~19.1~~~.040~~-.012~~~.042
7~~~75.89~~.080~~~.041~~~.039~~19.8~~~.039~~-.002~~~.041
11~~73.84~~.078~~~.031~~~~.047~~21.9~~~.035~~.004~~~ .042
1~~~72.65~~.077~~~.025~~~.052~~~47.4~~~.017~~-.008~~.060
12~~69.82~~.074~~~.013~~~.061~~~51.3~~~.015~~~.002~ ~.058
6~~~10.2~~.011~~~.007~~~.004~~~89.6~~~.009~~~.002~ ~.002

a your ratings
b your ratings normalized to total 1
c track-dependant historical probabilities
d b - c information added by your ratings to historical probabilities
e tote odds
f tote probabilities normalized to total 1
g f - c information added by tote probabilities to historical probabilities
h absolute value of d - g. difference in information added by your ratings vs information added by tote odds

The probabilities in column c are historical probabilities which are developed for a specific track. What can be done with this data?

Examine the differences in column d. Do your ratings make sense in light of historical probabilities?

Compare d, g, and h. Example: can you really justify a .294 difference between your ratings and "the public" for #9?

Ala Benter, average b with c or f.

Where b > f, overlays. Conversely, where b < f, underlays.

I'm sure that if you try this method, you'll discover more useful work to do with this information.

[classhandicapper]

I love the idea of using a confidence interval.

Most of the line making models I come across assume that an "x" point advantage on your figures will always equal the same percentage advantage. But in real life sometimes I feel confident I know how good the horse is and how he's likely to run and at other times his record may be a mixture of races on different surfaces, off tracks, distances, layoffs, equipment changes etc.. that gives you a very muddy view of how good he is, let alone how he's going to run today. IMO those two extremes and everything in between should be built into your thinking about fair value. The greater the confidence, the less margin of safety you need above the actual odds to make the bet.

[whodoyoulike]

Originally Posted by classhandicapper

... I just found that I rarely go to the window unless I am fairly sure I actually have value. At that point, I really don't need an odds line. It kind of screams at you because the horse will generally be misranked by the public (in other words, the horse I am making the most likely winner, is the 3rd choice in the betting, the horse I am making 2nd most likely is the 5th choice etc..) So putting the rating into the spreadsheet was costing me time for very little gain.

This is also what I usually try to do.

And Raybo, we all have or will probably face your dilemma, for me I came up with an alternate solution which seems to work for me. But, I don't have a black box instead I make an evaluation of my program's data results.

[raybo]

Lots of things to read and mess with. Thanks guys!

[Dave Schwartz]

Originally Posted by cj

I looked over my code and there are a couple other tweaks I made over the years. If anyone is interested say the word and I'll post. None of the change the basic idea.

I am certainly all ears!

(Screen shots at the ready.)

[Capper Al]

It's all about having fun. Enjoy yourself with your odds line. Good luck!

[098poi]

Good thread. I remember years ago trying to convert some ratings I made straight across to an odds line and had the same result as Raybo. Evenly spread out and bearing little relation to an actual odds line where the first 3 or 4 favs dominate the lowest odds.

[cj]

Originally Posted by Dave Schwartz

I am certainly all ears!

(Screen shots at the ready.)

One is that the base minimum of 2 I up depending on the number of contenders.

Minimum

>5 contenders = 2
4 contenders = 3
3 contenders = 4
2 contenders = 5

The reason was if you keep the minimum too low with a small number of contenders it tends to overrate the highest rated horse. Anyone using this really needs to figure out how much of a spread between horses is meaningful with their own ratings. A minimum of 2 for a contender may work just fine for some ratings, maybe higher is good for others, and variable can work too.

The other thing wasn't really important here. It has to do with how I comprise the rating for each horse but it really doesn't matter to the method I outlined.

[Dave Schwartz]

CJ,

That entire concept is just excellent.

I am assuming several things:

1. This was designed to work with speed ratings (or numbers that approximate speed ratings).

2. Your experience has determined that 16-points-from-top is the threshold at which a hit rate graph would drop precipitously.

3. The 2-point (or 3,4,5) is derived from experience and designed as a margin of error.


Thanks for explaining this. Raises the bar nicely in this thread.


Regards,
Dave Schwartz

[cj]

Originally Posted by Dave Schwartz

CJ,

That entire concept is just excellent.

I am assuming several things:

1. This was designed to work with speed ratings (or numbers that approximate speed ratings).

2. Your experience has determined that 16-points-from-top is the threshold at which a hit rate graph would drop precipitously.

3. The 2-point (or 3,4,5) is derived from experience and designed as a margin of error.


Thanks for explaining this. Raises the bar nicely in this thread.


Regards,
Dave Schwartz

1) It was originally, but I don't always use speed figures. This is true for first time starters of course, but also trainer changes (especially to super trainers or from super trainers) and layoff types and a few other scenarios.

2) The 16 points was just used with raybo's data to include four horses and make a few extra points. I use about half of that, though it varies a little by class.

3) Exactly. You obviously can't use zero, so I worked up from there.

[Capper Al]

Building an odds line should only be understood as playing with numbers. What would be the predictive assertion be worth when our top choice is wrong at least two-thirds of the time? Odds line are more self serving as our own tribute to our systems and logic. In practice, they are useless. I too suffer from this ailment from loving my numbers. That's why we are handicappers.

[headhawg]

Originally Posted by Capper Al

Building an odds line should only be understood as playing with numbers. What would be the predictive assertion be worth when our top choice is wrong at least two-thirds of the time? Odds line are more self serving as our own tribute to our systems and logic. In practice, they are useless. I too suffer from this ailment from loving my numbers. That's why we are handicappers.

If you believe this then you must not believe in betting for value or you're still focused on predicting the winner. Betting the top choice (lowest odds) is only of value when you can do it at a rate and at a price that leads to a positive ROI. So if the chalk is even money, but the odds line is 3-5 that's a profitable situation. If it were possible to create a 100% accurate odds line then betting only overlays would lead to profits assuming the bettor understood how to make the bet(s).

[cj]

Originally Posted by Capper Al

Building an odds line should only be understood as playing with numbers. What would be the predictive assertion be worth when our top choice is wrong at least two-thirds of the time? Odds line are more self serving as our own tribute to our systems and logic. In practice, they are useless. I too suffer from this ailment from loving my numbers. That's why we are handicappers.

Hogwash. It has nothing to do with if your percentages mirror the actual win percentages. Here is the simple test of an odds line. If your odds line leads you to bet on enough horses that you have as overlays to make you an overall winner, it is a fabulous tool. End of story.

[Capper Al]

I know that some believe that their odds lines reflect true value. And odds line are fun to make. Still, an odds line is only as good as the predictive powers of the odds maker. When it's time to wager, if our goal is profit, one must be honest and realize that they are in the dark about which horse will win and what these horses are really worth. There is only one absolute truth in handicapping: Good handicapping defeats random odds. Anything else is a guess. And that's no hogwash.