W. Benter’s paper, “Computer Based Horse Race Handicapping and Wagering Systems:
 

W. Benter’s paper, “Computer Based Horse Race Handicapping and Wagering Systems: A Report”
Back in 1995 a Paceadvantage member recommended the book “Efficiency Of Racetrack Betting Markets” to me. At that time, the book cost me $65. Today, I looked on Amazon.com and that very same book, the hard cover copy, is selling for over $350.
Using the compound interest tables, that calculated to About 6% / year. I’m already ahead of the game!!
There are about 60 papers presented in the book, by authors like Harvillel, Hausch, Bolton,, Chapman and W. Benter. Not presented is my favorite author on this subject, though referenced several times, is Burton Fabricand, Ph.D.

Now to the point.

Benter in tables 1 through 7, using data from several thousand horses, presents various scenarios for the public’s incremental win odd range estimate’s win estimates compared to his various models and conditions, in a statistical analysis format. All well and good.
I will point out here that proving a statistical significance in horse race betting does not support a profitable method of betting. Substantial significance, overcoming the booking percentage, is mandated.
In table 8 he presents 1 race of 14 horses, demonstrating as he says “After computing the combined and therefore unbiased probability estimates as described above, one can make accurate estimations of the advantage of any particular”

I need more.

Fabricand, in his book “Horse Sense” published in 1965 cost me $4.95. Today I find it on E-Bay for $100. That’s a compound interest rate of about 5.25%. I continue to be ahead of the game, and I have 3 “used” copies!!

Again, to the point.

Fabricant presents the “more”.
In his case, he compares “all favorites” to “system favorites” in a statistical format, clearing stating the incremental odds percent gain or loss . For instance, in the odds range 2-1 through 2.4-1, all favorites won 437 of 1456 races. 30% winners, for a loss of 3%, compared to his “theory of maximum confusion” system, 164 of 386 races. 42% winners for a gain of 37% and a 1% level of significance, 2.5 standard units.
Am I missing The “more” in Benter’s paper.

What I'd like to know is are we talking about a backfit with Fabricand's dataset? Show a second profit on a large out of sample dataset and you're going a very long way toward real-world significance in my book. Or you can keep adding to your dataset, backfitting and then trying to convince ourselves that it's significant despite the ongoing backfit.

Per Fabricand's book, Chapter 8 "Results".
" We shall now attempt the neat trick of plucking from a random sample of races only those races offering favorable bets on favorites, using the 15 rules of the proceeding chapter"

Given that, I say there is no backfitting..

Hmm I don't know about that, I think he has a 10,000 race sample (period). He finds a subset of favorites within that 10,000 race sample which showed a profit. That's a backfit. Nothing wrong with that but he proceeds to try to convince us that it's unlikely to be profitable by chance. Yes unlikely things do happen. That's a whole lot different from a second independent 5000-10000 completely separate sample. No neat tricks are necessary then. If he did that and it was profitable then I suspect there wouldn't have been a book in the first place.

I don't know if he ever placed a bet!!
Google Burton Fabricand.
I have always been impressed by his background.
He goes way beyond horse racing analysis.

But as per the original intent of my post, I ask where is Benter's "more"

Isn't the more that Benter actually made money betting on horses? I mean the paper is the paper, maybe I'm wrong but I don't think it was meant to be anything more than a high level overview.

Why are you asking this question? Are you trying to downgrade Benter's accomplishments?

I don't get it.

in table 7, he indicates the number of horses, the mean expected probability, and the actual number of winners.

in addition to that,I need to know the number of system horse's, and the number of system winners. The table should also indicate the track vig, which somewhere in the paper he may have indicated it was 19%.
From that I can calculate the min required wins for 2 and 2.5 z scores and the $1 returns.

ATTACHED IS TABLE 8 FROM BENTER’S PAPER AND MY UNDERSTANDING OF SAME

COLUMNS F,G,H,I AND K ARE HIS INPUTS

I MARKED COLUMN K IN GREY BECAUSE I DON’T UNDEERSTAND HOW HE ARRIVES AT THE DATA.

COLUMNS G AND H ARE SELF EXPLANITORY. I SUMMED THE COLUMNS TO OBTAIN THE BOOKING PERCENTAGE. THIS SUM INDICATES TO ME THAT PROBABILITY ESTIMATES ARE THE TRUE ESTIMATES.

IN COLUMN I, BENTER IDICATES THE $1 EXPECTED RETURN. ANYTHING GREATER THAT 1 IS A POSITIVE RETURN, THEY ARE HORSE S 2,5,8 AND 9

I ADDED COLUMN L WHICH, BASED ON HIS PROBABILITES IN COL G INDICATES THOSE HORSES RETURNING GREATER THAN A $1 RETURN. I SUMMED THOSE PROBABILITIES TO 0.345.

BASED ON TRUE ODD THOES HORSES WIN 34.5 PERCENT OF THE TIME, BUT OFCOURSE THE TRACK WILL NOT PAY YOU ODDS BASED ON THOSE PROBABILITIES.

I ADDED COLUMN M INDICATING THE BOOKING PERCENTAGE FOR A 19 PERCENT VIG.

MY COLUMN N INDICATES THOSE EXPECTED WIN RETURNS SHOWN IN COLUMN I. EXPECTED RETURNS THAT EXCEED THE BOOKING PERCENTAGE IS LIMITED TO HORSES 5 AND 8.

$1 RETURNS MUST EXCEED THE BOOKING PERCENTAGE INORDER TO PRODUCE A PROFIT

BASED ON COLUMN G THOSE TWO HORSES COMBINE FOR A WIN PERCENTAGE PROBABILITS OF 21 PERCENT.


RESPECTFUL COMMENTS OR QUESTIONS ARE ALWAYS APPRECIATED

Column k is odd as displayed in the tote
 

It says in the paper “div's are the win dividends (as a payoff for a S1 bet) displayed on the tote board.”

THANKS. THAT WORKS OUT TO COLUMN M (BOOKING PERCENTAGE/COLUMN L TRUE PROBABILITY).

BENTER'S $1 EXPECTED $1 RETURN (COLUMN J) IS BASE ON TRUE ODDS WITH NO ALLOWANCE FOR TRACK BOOKING % (MY COLUMN N).
IN THIS SINGULAR RACEHORSE'S ONLY 1 AND 8 QUALIFY. HE WOULD WIN THAT PLAY 21% + OF THE TIME.

I mean the as respectfully as I can, but you clearly have absolutely no idea what you're talking about.


As you say F, G, H, I and K are his values. They need no fiddling with or additional calculations.



I is the expected return of his probablilites using with the public market (win dividends). This already includes any takeout.


On his probabilities there were four profitable plays in the race.

[QUOTE=iamt;2880483]I mean the as respectfully as I can, but you clearly have absolutely no idea what you're talking about.


As you say F, G, H, I and K are his values. They need no fiddling with or additional calculations.



I is the expected return of his probablilites using with the public market (win dividends). This already includes any takeout.


On his probabilities there were four profitable plays in the race.[/QUOTE]

yes based on a track with no take-out. the sum of his and the public's probabilities sum to 1.0, indicating true probabilities.

after adjusting for take-out he still had 2 remaining probabilities. that's not a bad thing, just less probability of winning at a profit.

No. That is not right, not even close.



Column I looks to be his expect return based on the dividends in Column K.


Column K is the actual betting market at the time with takeout incorporated.


For example look at Runner #2. He makes the horse 7/1 (12.5% win prob) and the market has a price of 8.3/1. This is a profitable play into the actual market.


There is no need for anything you have done. It is pointless busywork. But if you felt the need to do it Column L should have referenced Column J, not Column I as it (column J) is a comparison to the market probabilities with no takeout applied.

i added column O that sums to the same booking % I used 1.23
column P,Q and r are self-explanatory
column R confirms that only two horses exceed the booking %, a requirement for a profit.

The intent of column J is so, as intended.