Quote:
Originally Posted by robert_utk
According to Kelly, poker players are massive nits pushing tiny slivers of bankroll back and forth on a poker table.
You have QQ, and your opponent has AK. How much does Kelly recommend to bet? More than you brought to the table.
If you are playing with a very small percentage of your bankroll, it might. A brief look at the QQvAK situation says about 12% of your bankroll.
Quote:
Originally Posted by heehaww
I'm a big believer in the Kelly Criterion but I think it would be rather impossible to use it for Poker. Even if you had a good estimate of the probability distribution of outcomes per hand, it's not like that would scale up. If you move to a higher stakes, the distribution will be different because the competition is better.
I think the reason there might be a way to calculate a number here is that the bet is the buy in and all of the choices within that buy in are compacted together. So it seems that hand to hand analysis is just internal mechanics of the ROI function. I have seen a few of these attempts to find ROI return things like 7bb/100 (+/- 15%), totally useless in the context of MTTs.
getmeofcompletely linked the an mtt attempt, it seems to be close to what I am trying to get at.
Quote:
Originally Posted by getmeoffcompletely
https://forumserver.twoplustwo.com/1...tsngs-1553950/
It turns out the common BR advice you hear is pretty close to what Kelly says too. The main thing to realize is how drastically field size influences how many BI you need. Frequently playing a high buyin small field event will be preferable to a huge field small buyin event. Also be conservative with what ROI you use for this, if you're a winning player it's more likely you've ran above expectation in your career.
That link is great, thank you. Fish on a heeter confirmed.
My anecdotal experience is that I have a huge edge in $10 events and near zero if any at $50+. This has always led to playing lower, but also why I want to investigate Kelly to so I can play a reasonable balance of buy ins and field sizes.
I have two ideas that might make sense. First is considering two probabilities, P(make final table) and P(2nd or better). The first would be highly sensitive to field size while the second a constant with most impact on ROI.
The second is to assume that the sample is played at N kelly and solve for ROI for various N. This is to say, If I was playing half kelly, here is my ROI, and if i was playing at 2 kelly here is my ROI.