Why You Shouldnt Play HUSNG with 20 Buyins

[daveT]

Forgot about Kelly Criterion. The problem I have with it, and I believe TNixon's theory closely resembled full Kelly. Is that you are not able to know if your numbers will be true. It is a different ball-game than Economics, where your expectations can be calculated with more accuracy.

I have several WRs: 2,5,6,10,11,20, my over-all for turbos, my over all for regulars, and my over-all. Even if I decided to use strictly my WR for regulars, because of my sample size, I am not able to determine what my real WR is. I know that my over-all does not reflect what should be happening. But most importantly, I cannot determine what will happen to the WR when I move up to the next level. I could do some number finagling and just assume that I am winning 53% to show the absolute max I need.. interesting.

[jay_shark]

This is why it's good to use half-kelly because many people tend to over-estimate their win rates . Tnixons approach closely resembles full Kelly bets but it can be risky if you over-estimate . In fact , your long term bankroll growth rate will plummet to zero in the long run if you bet higher than full kelly .

[TNixon]

I actually figured out Kelly Criterion to see how it compared to my bankroll simulator at one point, and it has you playing more aggressively than 10 buyins, right around 7 buyins for a 58% winrate.

Now that cwar's "play patient" thread has been almost fully derailed, here's a link back to the thread where I posted the first results from a bankroll simulator I wrote.

http://forumserver.twoplustwo.com/showfl...part=2&vc=1

One thing to be careful of is that the first iteration was confusing. Step up buyin amounts were for the next level, and step down amounts were for the previous level. Since then, I've refined it, and it seems that stepping up with 10 and back down with 7 (so when you lose 3 buyins at the higher level) is pretty close to optimal, while retaining a very low risk of ruin. Somewhere around 7/7 is truly optimal (and matches the Kelly Criterion for 58%, which is where I baselined the simulator), but the small amount of extra value you get there comes at the price of what I considered too high of a chance of busting. The original simulation was for a starting point of $20 as well. If you start with $100 instead, your odds of surviving 1000 games without going bust, playing 10/7, are 98.68%. It's very easy to end up with less than a 1% chance of going bust without losing too much value by slightly modifying that to 11/8 or 11/9.

I've made quite a few posts on this topic, including a modification to the bankroll simulator where I tried to take a learning period into account, starting your winrate for a particular level at less than 50%, and increasing it a small amount for each game played up to a specified maximum. A search should tag them all.

[TNixon]

[ QUOTE ]
Tnixons approach closely resembles full Kelly bets but it can be risky if you over-estimate .

[/ QUOTE ]
It's more like around 70% Kelly for a 58% winrate.

For a 56% winrate, I'd probably want to do something closer to 15/12, where full kelly would be more like 10.

For a 53% winrate, you can't get over a 95% survival rate over 1k games (and I was really looking for 98.5% or higher) unless you start with at least $200 *and* play 30/20. And the resulting bankroll after 1k games is pretty pitiful. You win about $40, and probably only very rarely make it out of the $20s.

[STA654]

I've been building a roll playing 4 person HUSNGS using a 25BI rule.

Gone from 290 (which I built up from $1 playing cash games) to 643 playing HUSNGS (Playing the $20s now).

Should I switch to a 15BI BR managment system?

[jay_shark]

I prefer to err on the conservative side when estimating my win rate .

So using full kelly , we would invest f=[p*(v+1)-1]/v
where v is your v-to-one odds received on the wager , v= 0.9523:1

So if we use p=58% , we would risk [.58*(1.9523)-1]/0.9523 = 13.896% of our bankroll on each individual gamble . So using half-kelly we would risk about 7% of our bankroll which still maintains a high bankroll growth rate without as much volatility .

[jay_shark]

[ QUOTE ]
I've been building a roll playing 4 person HUSNGS using a 25BI rule.



[/ QUOTE ]

Sounds good if you have a high ROI playing in them .

Your variance is higher in 4-mans and your bankroll should compensate for this .

So if your ROI for a 4-man tourney is ~ 35% , and a s.d of ~ 1.93 then B = -*1/2*s.d^2/ROI*lnr where r=1%

B= ~ 25 buy-ins .

So there is a 1% chance you will bust at some point . The most difficult part is determining your true win rate . This is why it's better you err on the conservative side and you may choose more than 25 buy-ins unless you've been doing this for a very long time .

[TNixon]

If I could just post results like this every time I played, the number of buyins I have would be such an irrelevant concern:

2 player $30: 82.35% (14/17) ROI: 56.86% Total profit: $304.50
2 player $50: 71.43% (5/7) ROI: 36.05% Total profit: $132.50

All games: 79.17% (19/24) ROI: 48.39 Profit: 437

[img]/images/graemlins/smile.gif[/img]

[cwar]

jay shark and TNixon you are both very logical people and I think what you are overlooking is that a lot of people cant think the way you do, myself included. It would be great if everyone could jump up and down in limits and keep playing winning poker but that most definitely is not reality, often times the psychological factors I discussed in this article can lead to not playing winning poker for a very extended period of time or even force players out of poker completely.

[TNixon]

[ QUOTE ]
It would be great if everyone could jump up and down in limits and keep playing winning poker but that most definitely is not reality, often times the psychological factors I discussed in this article can lead to not playing winning poker for a very extended period of time or even force players out of poker completely.

[/ QUOTE ]

While this is absolutely true, you should not be recommending that those people play with 20 buyins. 20 isn't nearly enough for somebody who doesn't have at least some sort of grasp on the psychological issues. (perfection is not required, but introspection definitely is)

What you should be recommending is that they avoid heads-up poker like the plague.

Er, wait a minute. I think you'd be chasing off every single person who I can actually beat heads-up if you did that.

Nevermind, ignore me. These are not the droids you're looking for.

Isn't the weather lovely today?

[img]/images/graemlins/smile.gif[/img]