Quote:
Originally Posted by DollarWill
"If someones win rate was absurdly high (like 20bb/100), the variance would be pretty much non-existent at that point." fwiw I also find this statement absurd.
Variance in absolute terms is going to be fairly constant regardless of your win rate (ignoring a bit that differences in style can cause changes in variance). But variance in respect to your win rate is a more interesting or important number. It's important enough to have it's own term: "coefficient of variation" which is standard deviation divided by the mean.
It can seem absurd to talk about playing with a small number of buyins, such as 10, because it's tempting to think "you could have a bad run right at the beginning and lose 10 buyins." It's true that a downward streak of 10BI could happen at any time, to anyone. But, whenver you don't start with a downswing, if your win rate is very high then you start accumulating more bankroll quickly.
There is a basic "risk of ruin" formula that you might find useful. The forumla is
ROR = e^(-2WB / (S ^ 2))
W is winrate, S is standard deviation, B is bankroll and e is the numerical constant e
We can reorganize this to solve for the size of a bankroll required for a given risk of ruin. Let's use "R" as risk of ruin.
R = e^(-2WB / (S ^ 2))
ln(R) = -2WB/(S^2)
S^2 * ln(R) = -2WB
B = -S^2 * ln(R) / 2W
So let's consider some numbers. Let's say your standard deviation was 100BB/100 hands. Win rate ridiculously high, like 20BB/100. And let's say we're comfortable with a 5% risk of ruin - that is, a 5% chance that if we never change stakes, we'll go bust.
B = -100*100*ln(.05)/(2*20) = 748 BB
That is to say, that player could get by with a 7.5 buyin bankroll. If you're happier with a 1% risk of ruin, you'd need 11.5 buyins. Note that if you can move down, your risk of ruin becomes lower, but for a lot of people, moving down isn't practical.
But that's a really obscene example. You can play around with other numbers.
This calculator is really interesting and useful also. It calculates risk of ruin and stuff like that, but it also visually shows you what a sample of players with those stats would "look" like over time. If you plug in W=20, S=100, R=0.05 you'll see what I mean
http://pokerdope.com/poker-variance-calculator/