Is this Kelly?

[Gerald]

Let B(0) = original bank
B(new) = bank after a bet (and result)
y = decimal odds a la Betfair
s = normal odds = y-1
p = probability of winning
q = probability of losing
So, p+q = 1
x = proportion of bank bet

x = (sp+p-1)/s = [p(s+1)-1]/s = (py-1)/(y-1)[/color:37jrxsyf]

Okay, I was able to find it on the Internet, once I knew that I was looking for Kelly Criterion, rather than typing-in Kelly and seeing what cropped up; Kelly Brook, nice.

The good news is that my maths is right. It is normally expressed slightly differently though.

In my notation it is x = (ps-q)/s[/color:37jrxsyf]

(Where I write x = (sp+p-1)/s , this is (sp-q)/s, as 1=p+q)

There seems to be a lot of disquiet about proportional betting, compared to level stake betting. There also seems to be a bit of mischief being caused, by people comparing the "average-sized" kelly bet with the level-stake bet.
A long losing-run will hit both proportional betting and level-stake betting, but proportional betting can’t go down to zero, while level stakes can run out. But when the long winning-run comes along, proportional betting is in a position to take advantage.

[carvillshill]

I’m familiar with the concept of the Kelly Criterion from the writings of Nick Mordin amongst others. The difficulty is that the percentage probability of winning is completely subjective and has been said, if one large mistake is made a big chunk of the notional bank goes west. Personally I’ve found the biggest improver of my profits in recent years has been forcing myself to have as near as dammit level stakes on almost all bets regardless of price. That means I’m having as much on a 50-1 shot as on a 2-1 poke. Why? Because if I review profits year on year it’s the big price winners that make the profits, the shorties just about hold their own. It’s a sort of reverse of the favourite longshot bias in my hands where my edge if I consider a 20-1 shot overpriced is far bigger than if I’m punting at 9/4. The only exception I’ll make is having 2,3 or 5 times normal on if I think a price is just plain wrong, this tends to happen more ante-post. When I fancy a price NRNB before Cheltenham within reason I’ll have as much on as I can as I know the horse can’t start that price on the day. Hope this helps a bit Gerald!

[Gingertipster]

I used to have the same staking plan Carv.
But found it mattered very little if the short priced horses won or lost, mattering a good deal how many long shots won. Of course if you are always better at finding big priced winners tghis suits you.

But I changed and instead backed everything to win the same amount. However, then it did not matter whether the outsiders won or not.

So now I bet acoording to the table of odds and chances. If I believe something has a 40% chance I’d put 40 points on it at say 7/4 winning 70 points. If I thought it had a 13% chance I’d put 13 points on at say 15/2 winning 97.5 points.

Although something at a short price has a theoretical better chance of winning, so there should (imo) be more money on them compared to outsiders. But this being the case also means more will be lost if they lose. Therefore, (imo) short prices should have more money on them, but should not win as much as outsiders. If that makes sense.

Mark

Value Is Everything

[Gerald]

From the days that I used to buy the RP every day, and used them for various pieces of system testing, I am in agreement with Carvills here. Unless you are very successful with your short-priced horses, it doesn’t make too much difference if you back them to smaller or larger stakes, compared to your other selections. What makes a difference is if you can get 10 long-priced winners up out of 50, rather than 3 long-priced winners up out of 50. Of course, one can’t concentrate solely on these horses, for both psychological and financial reasons (long losing runs are what I am referrring to in both instances.)

[Author]

Posts