Quote:
Originally Posted by kwedaras
where does variance 1 and 2 numbers come from ? what number should i use for super turbos ?
Variance is the square of the standard deviation which is derived from
this formula. fwiw it's easy to point out that the variance of a 2man is 1 buyin (obv you always go 1 BI up or 1 BI down), while it's 4 buyins for 4mans (the math is more complicated but easily doable using the formula). Thus you should use 1 BI as your SD in superturbos (B=1/ROI)
Quote:
Originally Posted by kaby
also, while it's definitely interesting to know what's mathematicly optimal, optimal BRM for almost anyone is going to be 'as agressive as you can be without going insane'. for me as a pro that's probably 60-100bi for the highest stake i play depending on my game selection
yes, that's true. I dunno how conservative you have to be to profit in MSNL but you should work on your mental game imo, esp. if that highest stake of yours is 4-6 times the stake where you play most of your hands
Quote:
Originally Posted by alex23
I have a case where I think this formula doesn't lead to maximal bankroll growth. Say there are only 3 stakes, $100, $200, and $400. Let's assume:
-At the $100, our ROI is 10%
-At the $200, our ROI is 1%
-The $400 is incredibly tough and we lack the talent to ever win there
-Our bankroll is $1 million
Won't this formula tell us that we can play $200 games when in fact we should be playing $100 games for maximal expected bankroll growth?
Assuming you can play X of each game in one hour, your hourly rate for $100s is $10*X while your hourly rate for $200s is $2*X, so altho we're way more than sufficiently rolled for both games, it doesn't change the fact that $100s give us a better hourly. However I'd still mix in $200s and the occasional $400 to work on your game and see if you can beat the learning curve. imo thankfully poker is not football and the degree of "talent" in poker is fairly low, so unless you're mentally ******ed or sth, the reason why you may not be playing $200s right now is something fixable: either you haven't put enough volume, or you haven't worked enough on your game, or you have a mental crutch of some kind etc.
Quote:
Originally Posted by quinn132
But surely the kelly criterion is used as a bankroll management system to help players move up to their regular level. that of which would yield them the most profit and they should be playing even with an infinate bankroll.
Once players start to hit their ceiling stakes for their current skill level it should not be so much about bankroll managment as you will quickly become over rolled. Then moving up becomes a product of skill and not bankroll.
imo your overrolled when you're playing too low stakes while you could play higher stakes and yield a higher profit. As I said to alex23, the most profitable level to play is not necessarily the highest, but considering grinding e.g. $22s in the long term is quite ridic as you will def attain a skill level at some point that allows you to have a higher hourly at $33s at the very least. After a while of not withdrawing too much you'll end up w a huge roll w ability to profitably play any game, but that doesn't mean the mechanics of BRM become wrong, they just can't aim us at a level anymore, so you can focus solely on hourly rate
Quote:
Originally Posted by Emus
This question will probably be meh for alot of people but I do not care.
Do you want to try to explain me, how they come to
E(G) = (1 + (O-1) * X)^p * (1 - X)^(1-p) - 1
in the first paper.
http://www.sbrforum.com/betting+arti...criterion.aspx
I do not fully understand how they come to this formula ...
AFAIK the expected growth is the expectancy of the increase of log(bankroll), which solves the problem of losing your roll (meaning you're banned from playing 4ever) and doubling it up (meaning you just move up a notch in the stakes) having the same absolute value. If you lose your whole roll, log(roll) becomes negative infinity. The formula in the article is a way of adapting EV to work w log(roll) instead of roll itself, so instead of using
EV(roll) = r(x1)*p(x1)+r(x2)*p(x2)+...+r(xn)*p(xn)
where x1...xn are outcomes, r(x) is the roll after outcome x and p(x) is the probability of outcome x; you just raise it up one level
EV(log(roll)) = r(x1)^p(x1)*r(x2)^p(x2)*...*r(xn)^p(xn)