January 04, 2011
I thought, I'd spend this post clearing up some stuff about my last 2 posts and responding to comments I got in the thread, but also from other sources.
First of all, let me say that I was very surprised by one result in the tourney: that if you are the best player in the tourney, in fact have a large edge over the rest of the field, THEN is the time to take marginal EV spots. That playing nitty then will not yield you the most money, while when you are barely a winner, you actually stand to gain to see others go down in flames while tightening up or just playing slow - it seems the EV is just too low to really make a difference to the fast increasing blinds that is a turbo tourney.
A few notes:
a) This is a Rush tourney. Your opponents constantly change. Hence you have an edge "over the field" which is a very wide concept and we can't really know that beyond of what mistakes we see at our table. In a normal tourney, where you are at your table for a long time, it is my feeling that we can increase our theoretical edge over the table by accumulating good reads and that the rest of the field is fairly unimportant to us and our standing in the tourney.
b) I am somewhat worried about the smoothness of my simulation. In the end, poker results are fairly discreet and I am afraid there are some problems at the edge, i.e. when we bust from the tourney. Since it is a limiting case, I don't quite know how to deal with that mathematically.
c) I am not a tourney expert. There are probably miriads of better insights written before this and I know for many people, that I'm reinventing the wheel. Still, I find it interesting and again, I found the turning point quite high as mentioned above, so it was very shocking to me.
The whole series of post was inspired by Jesse, his poker lunch and his post on 2p2. Thanks Jesse! http://forumserver.twoplustwo.com/17/small-stakes-limit/no-woman-no-cry-925446/
That post also brought up some questions from thesilverbail, which I will answer here:
" 1. I'm not sure I understand your methodology in part 2. You populate an LHE tourney with players with different winrates. 4 different categories (10, 5, 2, 0, -2). Do these winrates balance out ? Otherwise your conclusions might not be accurate."
No, I play one tourney, where I have a xBB/100 edge over the rest of the field throughout the tourney. Obviously that is a flawed assumption, since your edge over the field doesn't stay the same over the whole tourney. In fact, in the next post I will run some sims mostly with decreasing edge over the field throughout the tourney, to see how this affects especially later stage risk approach.
"Do you mind sharing the code you wrote for the simulation?"
Oh boy, this is embarrassing, since I'm such a bad programmer (spaguetti code, here I come). I know there are much quicker ways of doing this, like just input a sheet of the tourney structure instead of playing each level out one by one by hand, but for now here it is:
"2. I think assigning each line to hands per hour and labeling the x axis in number of minutes per level is confusing. The difference is really number of hands per tournament level which is what allows the edge of the good players to stand out."
Yeah, it was just a starting point. Remember, the argument started with the discussion if you should play fast or slow in a game where you are constantly overbetting you bankroll and the blinds go up fast, but where you do have quite the edge over the field. After I posted this many people suggested I should start to vary the speeds, maybe proportional to the edge to get a sense of how this would affect it throughout the tourney, but I don't want to tweak to many things at the same time. For now I will do the varying edge thing, plus some slower structures, let the blinds increase less quickly.
"3. It's interesting to me that the avg #BB /person slopes very gently upwards in all the curves after level 10-13. why would this be so?"
It's just the function of the edge just finally catching up with the increasing blinds. It is dramatized or magnified due to the fact that I left constant edges over the field throughout the entire tourney, which of course doesn't happen
"4. I think you are going to argue based on these results that a player with > 3BB/100 winrate should widen his range to increase his "effective" hands/hr. The problem is that the hands you add at the bottom of your range won't have anywhere near ~3BB/100.
In fact this is almost an argument for tightening up in the tournament ! ( I think albacorela pointed this out). "
Besides the varying edge thing, the next thing I will try out is non-rush poker tourneys. At the end of the day, this is a turbo and an extremely fast structure. My feeling is that with a slower structure, you don't have to be so conservative, and can take more EV shots, but I still have no idea. It was quite eye opening to me that it took so large edges >3BB/100 for the EV to even make a dent and that perhaps playing nitty when having small edges in turbos and passing up some spots, does work, but only if the EV difference is minimal! In NL, this would be akin to passing up 22 vs AK (51:49), but not QQ vs AK (57:43).
On to copoka's comments:
"Bella, don't take it the wrong way, but "absolutely sure of your winrate" is border line silly. The only way to use "absolutely sure" and "winrate" in one sentence would be "one can be absolutely sure that he can never be absolutely sure about his winrate"
WR is nothing more than suggestive and convenient number and should not be used for any kind of "predictions" or forward looking analyses. Any simulation made based on assumed WR is basically useless since the assumption made is HUGE and never accurate."
I see this all the time, that one can never be sure of their winrate, that you have to play 17 million hands to know for sure, etc. Yes, you don't know your absolute winrate for a given limit or something like x number of hands. But you can fairly quickly assess if you are a winner in a certain lineup, estimate your edge and even pin down you approximate winrate based on that after a few hundred hands (loser, breakeven, marginal winner, winner, crusher).
Copoka, you and I have maybe played 20-50 hands total lifetime, yet at the time when you were playing a lot of HUHU, you constantly denied me action. Why did you do that if you didn't know what winrate you had against me and you could never be absolutely sure of it?
Notice how I am not deriving the winrate from past events. Take, for example the stockmarket, everybody will agree that forecasting the success of a company based on its past performance alone is very iffy and should be taken with a grain of salt. However, when you look at the fundamentals of a company, if it is based on a good product and not just hype or a bubble, you will win money. The more information you have about the company, the better you will be able to determine its likelihood to succeed and for you to make money in the long run.
If I see a game where I see constant mistakes and things I can exploit, I will know I can win in that game and I can fairly easy estimate my winrate in that situation. Otherwise you'd be truly lost, wouldn't you?
"So, although number of AA dealt, sets flopped or straights we make on the river can not effect difference in skill for a long time, there is another thing that can, is and will.
There are monies involved."
You are talking about poker being a game of people, that some of us tilt, that nobody can every play their best. Yes, those things all exist, but for the purpose of this calculation, they should not be important. If you sit down at a game, but are tired, you should factor that into your winrate estimation. Of course we can't be 1BB/100 (or whatever number) over long periods of times, there are fluctuations, even with the same lineup, sure. Then again, this is just a benchmark, something to hold on to. Otherwise let's just all throw our hands up in the air and proclaim that nothing can be predicted and it's all chance.
"First, those things are fairly rare comparing to the number of hands dealt so they converge much much slower in "calendar terms" so to speak. Nasty thing is that although they are rare, the effect on your earn is huge."
For every one of those you nasty ones, if you are a winner you will have an even better super-duper high winrate one statistically speaking. Variance goes both ways, not just negative!
"Second, even if we are lucky enough to see them balancing out, deck is still stacked against us due to measures we have to take to survive in this game."
Your argument makes no sense. If your theoretical winrate is the same it is exactly just as likely that you will have coolers at 10/20 as at 5/T. You are running into a huge gambler's fallacy here. Coolers don't even out, every hand is its own in the vacuum. It is just that heaters are a teeny bit more likely to happen than coolers if you are a good winning player.
If you lose money at the stockmarket and lose half your worth, of course you need to double your worth to be exactly at the starting point again, but if you have a positive expectation and bet proportional to your edge, in the long run you WILL win money and the bets you make will be bigger and bigger proportionally, provided your edge keeps existing. I don't understand what is so hard to graps about that. The only thing that will make you lose more pots at higher stakes is that you might not have the edge there.
"Third. There is so much human and/or social factors involved in giving and getting action in poker, that the argument can be made that bell curve simply cant govern this process for the same reasons it is useless in any other field where human nature runs wild, like stock market, for example.So what convergence are we talking about in a first place? Where would that come from?
You don't believe poker follows the Central Limit Theorem, you believe in fat tails or black swans. Fine. I have never seen any evidence that suggests black swans for poker, but who knows, I have not seen any evidence that denies it either. For a while I tried to prove this by gathering about 1.5 million hands from winning players and describing their deviations from the normal distribution, I didn't see any, but I could not prove that there weren't any either. I don't know 100% what happens at the extreme edges, it might be that your disastrous sessions are even more disastrous. Fat tails is not my territory and it is very disputed, for the discussions in Mathematics of Poker it is safe to assume that the Central Limit Theorem holds, as most results are determined by something that is really truly random and has no psychology - the deck of cards.
"I'd say you are looking at graphs of skilled poker players who where lucky enough positive cooler balance for a long time. The proportion of skills vs luck is unknown. Prove me wrong.
Oh, btw, there is a pretty good HU poker player named Matt Hawrilenko. Have you seen his graph lately?
There is another one. Very talented and very naive. Talented enough to break the game of HULNE down to the science. Naive enough to declare in one of his blogs that "making 300k playing poker is a child's play" His name is is Bryce Paradis. Take a look how much he made over past 12 month."
Both players are playing in games where they don't have clear edges anymore, their winrates, while positive, are not the 4-5BB/100 they were used to when playing in softer games in 2006-08.
AFAIK both Hoss and The_Bryce have been playing a lot of 6max lately and not that much HU. Bryce himself has been going through tough times with his mind being on his real estate losses and his depression. I hardly think that you can be a clear winner during those times.
Both said that they played HUHU LHE at a time where there was no clear theory on how to play it and they had just extreme winrates, just because the skill difference is so big. That is not the case anymore, any 3/6 HUHU reg is much better than the 2006 50/100 players, where these guys made their money off.
What I was mostly referring to, are the bumhunters that refuse action, but to the biggest of fish. They have smooth graphs, no matter how bad they run, because their winrate over the fish smooths any graph. I took those HUHU players as an example, because HUHU is almost the only place where such extreme skill difference is shown and it has a direct impact on your winrate. You might be right though in today's games, where the games are tougher and even HUHU edges are small, so my comparison was for bigger edges :)
Anyway, enough for now, even though I could talk for hours about this, I need to go to sleep.