Kent D.

[The Indomitable DrugS]

Originally Posted by Kasept

However you view the sample size, it's 50% of the universe you're going to end up with...

It's less than a hundred races ... which, for a jockey, is like a squirt of piss in a very large bucket.

[Dunbar]

Originally Posted by The Indomitable DrugS

It's less than a hundred races ... which, for a jockey, is like a squirt of piss in a very large bucket.

There's a way to get more statistical significance out of a smaller number of races...

A bettor who has an edge should be betting less on longshots than on favorites.* So, instead of looking at ROI for flat $2 bets, it makes more sense to look at ROI for proportional bets. And it turns out that the end result is much more statistically meaningful. (the standard deviation is much smaller when compared to the ROI fig.)

Here's a link to something I posted on another forum that gives an example of the difference:
http://sharpsportsbetting.com/forums....cgi/read/9760

If you or anyone posts a list of Desormeaux and Lezcano's races (showing odds and if they won), I'll calculate the proportional bet ROI and show how I did it.

--Dunbar


* Ed Thorp showed years ago that the optimal horseracing bet for bankroll growth is Edge*Bankroll/Odds. If you have a 10% edge and a $1000 bankroll, you should bet 10%*1000/2 = $50 on a 2-1 shot, but 10%*1000/10 = $10 on a 10-1 shot.

Some may argue that you do not have the same "edge" on every bet, and that you in fact have a bigger edge on longshots. But given how difficult it is to estimate our edge at all, it's probably safest to assume our edge is roughly the same on every bet.

[Bobby Fischer]

Originally Posted by Dunbar

* Ed Thorp showed years ago that the optimal horseracing bet for bankroll growth is Edge*Bankroll/Odds. If you have a 10% edge and a $1000 bankroll, you should bet 10%*1000/2 = $50 on a 2-1 shot, but 10%*1000/10 = $10 on a 10-1 shot.

i'll look it up, i'm just burn't out right now in need of a power nap... but do you have a link to Ed Thorp's papers?
The formula that I came up with is similar but more reliant on the individual than the public Bankroll*0.05*(lowestimateHit%squared*lowestimate$ 2payout)/2
You plug it all in a spreadsheet with conditions that say if ($2price*hit%)/2 isn't more than 1.3 or1.2 whatever "value" demand, then zero is listed in betsize.
It's not a bad type of thing do work out with bets for a while and develop an intuitive feel of the type of bets you should be making and why, but it's pretty boring.


the effect that Thorp gets from using the odds has the same basis as why I choose to square the low estimate of hit%. Hit percentage is very important in keeping "randomness" out.

[chucklestheclown]

Originally Posted by Bobby Fischer

i'll look it up, i'm just burn't out right now in need of a power nap... but do you have a link to Ed Thorp's papers?
The formula that I came up with is similar but more reliant on the individual than the public Bankroll*0.05*(lowestimateHit%squared*lowestimate$ 2payout)/2
You plug it all in a spreadsheet with conditions that say if ($2price*hit%)/2 isn't more than 1.3 or1.2 whatever "value" demand, then zero is listed in betsize.
It's not a bad type of thing do work out with bets for a while and develop an intuitive feel of the type of bets you should be making and why, but it's pretty boring.

the effect that Thorp gets from using the odds has the same basis as why I choose to square the low estimate of hit%. Hit percentage is very important in keeping "randomness" out.

Hence my losing streak.

[Dunbar]

Originally Posted by Bobby Fischer

i'll look it up, i'm just burn't out right now in need of a power nap... but do you have a link to Ed Thorp's papers?

The idea of betting Edge*Bankroll/Odds that I referred to above is the Kelly Criterion. Wikipedia has a decent description of it:
http://en.wikipedia.org/wiki/Kelly_criterion

Can't find a Thorp link that's ideal for horseracing. Here's one paper he wrote that covers a lot of Kelly betting, but it's pretty heavy math.
http://www.bjmath.com/bjmath/thorp/paper.htm

--Dunbar

[jms62]

Originally Posted by Dunbar

There's a way to get more statistical significance out of a smaller number of races...

A bettor who has an edge should be betting less on longshots than on favorites.* So, instead of looking at ROI for flat $2 bets, it makes more sense to look at ROI for proportional bets. And it turns out that the end result is much more statistically meaningful. (the standard deviation is much smaller when compared to the ROI fig.)

Here's a link to something I posted on another forum that gives an example of the difference:
http://sharpsportsbetting.com/forums....cgi/read/9760

If you or anyone posts a list of Desormeaux and Lezcano's races (showing odds and if they won), I'll calculate the proportional bet ROI and show how I did it.

--Dunbar


* Ed Thorp showed years ago that the optimal horseracing bet for bankroll growth is Edge*Bankroll/Odds. If you have a 10% edge and a $1000 bankroll, you should bet 10%*1000/2 = $50 on a 2-1 shot, but 10%*1000/10 = $10 on a 10-1 shot.

Some may argue that you do not have the same "edge" on every bet, and that you in fact have a bigger edge on longshots. But given how difficult it is to estimate our edge at all, it's probably safest to assume our edge is roughly the same on every bet.

That little factor called Edge and its phantom nature has cut through a lot of bankrolls.