oh i meant to say 3bb winrate in my post above, i ran it with oinks data to try and prove he was a luckbox

Megster

Your sample size is really small. But it looks like you are rather tight, most will say that you should steal and defend more.

The stars 5/10 is a tough game to beat. Several have posted screenshots of datamined databases showing rather few long term winners.

If you have the roll keep playing that game but you need to realize that you have to be pretty solid to compete. I dont think very many can beat that game playing an oldschool TAG style

just ran another sim

100k simulations of a 1.5bb winrate 20bb std dev over 10k hands

7.3% of them had 300bb downswings

gonna do same thing but with 20k hands


btw im not really sure how accurate this stuff is gonna be. anyone with insight (looking at you leader) plz tell me

How the hell are u runnning these sims? Its very interesting by the way. Why dont you run 1bb/100 winners over that span. Oink is the only person winning at 3bb or higher at mid stakes over a large sample given he has provided evidence in this thread lol...

Do you know where one could see these 5/10 databases you speak of? id love to see a 200k 300k or 500k sample. I think this info is much more protected than you say it is.....

I made an excel sheet that can simulate potential winnings for a given winrate, std dev, and # of hands. my buddy made a macro that looks for downswings of a specified size in a specified number of trials

I'll do some stuff for 1BB and post it in this thread. it'll take a while though, these sims take hours to run

btw some of the graphs of simulated 1BB winrate results are disgustingly sick. looking at periods of 400k hands, one of the trials had a 1500BB downswing, and the next one had a 200k hand breakeven stretch

ehh its been posted here several times

Considering this I wonder if a 1.5kBB BRM is a good idea for pro high stakes players. I mean +30/60

This shouldn't take hours. I'm going to try and put something together in Matlab. I'll post some results assuming a standard deviation of 21BB/100.

doughnutz talks about a 1k BB downswing. if all he had was 1k BB for his br, then he's broke. so, i guess you have the simulation as well as the real world example arguing for 1.5k BB.

yea matlab should be able to do it in a matter of minutes. excel is just really slow at generating the random numbers

1BB/100, Standard deviation of 21BB/100, 1 million hands:



200K break even stretch, multiple 1000BB+ downswings and a 200K hand stretch running at 2BB/100.

0.5BB/100, Standard deviation of 21BB/100, 1 million hands:



Same stats:



So it can vary quite a bit.

0BB/100, Standard deviation of 21BB/100, 1 million hands:



The life of a rackback pro.

3BB/100, Standard deviation of 21BB/100, 1 million hands:



It's good to be Oink

why did the 0 BB/100 winner not approach 0 BB/100 as # of hands approached infinity? what am i not getting?

Thanks Abso! Do u have a rope for me?

I still hope my graph looks soon more like the 2nd or 3rd graph again. The first and last graph are just so far away currently.

It's only a million hands.

this was also my conclusion

I Have PT3 now. PT2 would give me stdev bb/100. I cant get pt3 to give me that stat anyone know how i could go about calculating it?

Those graphs have some exciting implications though isnt this kind of relative to ever changing table/game conditions and a changing personal skill level.

Its pretty hard to define your winrate if the game is so volatile.

I have a higher standard deviation as a huhu player but a much smaller number of hands over time. I think bryce said something a long the line that 20-25 k hands of huhu would equal 100.000 k of 6 max hands because you play a much higher number of hands.

What would your thoughts about this be?

Depends on the exact meaning of the claim.

Just by counting the number of decisions made in both of those samples would perhaps put them on the same ballbark, though I'd still guess it's more like HUHU has 2-3x more decisions per 100. What this implies that you can attain a decent expectation, because your opponent has so many opportunities to make mistakes.

Edit: And of course because you can play with the fish alone instead of with 4 other decent players.

If the claim is that the winrates converge faster in HUHU, then it's definitely not true. In fact, the opposite is the case. The fact that the std dev / 100 hands is higher in HUHU is an indication that it has higher variance in a give sample and thus the confidence interval is wider. It's not like it means something different for HUHU and for 6-max.

Of course, if one's winrate is what is considered "typical" winrates for good players in both games, then it's also true that you're much more likely to end up winning a given sample of hands (provided that it's large enough that the winrate has an effect).

But a 4BB/100 winner in HUHU can end up winning only like 1BB/100 in a decently sized sample, because the variance is so high. But the corresponding player in 6-max would probably be losing over a similar sample, because he is so much closer to break-even expectation.

So, if you're a decent winner in HUHU games, you can expect to be up more likely in a given sample, but your winrate doesn't converge faster.

I'm sure Leader or anyone else competent will correct me if I'm wrong. :-)

I'm curious about this also, and aggression stats as well.

Wookie's stats post shows a flop aggression range of 2.5-3.5 and river aggression of 1.5-2.5. Are these still considered good numbers? I ask because these don't seem to match up with some numbers ITT that people seem to feel are solid. For example, in his stats Oink's flop aggression is 2.1, and river is 1.48.

Postflop aggression factor depends on your preflop strategy. Wookie's preflop recommendations back then were really nitty compared to today's standards, which in turn means that the postflop AF recommendations are off too.

There are few reasons for this. When you start stealing at a much higher percentage from late positions, other good players will start to adapt to it by 3-betting you a ton more. In these spots, you can't play fit or fold, because you would have to give up way too often. First, you need to peel more liberally in these situations.

But you also need to get to showdown pretty lightly, with hands that have pretty marginal showdown value. And since the overall strength of your postflop distributions are so much weaker (less hands that can see showdown UI), your postflop aggression has to decrease somewhat. This is reflected in went to showdown (when saw flop) stat, which is significantly higher by today's standards than it was in Wookie's stat post.

My AF is 1.8 I think. 2.1 on flop 1.5 on river.

I thought about it the other day and I kinda reached the conclusion that anyone with an AF above 2 is having leaks in some part of his game unless he is playing in some very very passive games.

So maybe rephrase that to anyone with an AF above 2 in reasonably aggro games, say 3/6 and above, is doing something wrong

My af is 1.97 which is weird because I seem to have a more passive style than most.

Also its interesting how my af decreases at higher limits.

2.04 at 10/20
1.98 at 15/30
1.86 at 30/60.

Maybe 2 is too low. But if its above 2.5 you are definetely doing something wrong. I also highly doubt having an AF above 2.2 is correct

Thanks for the aggression feedback. Thoughts on this?