statistics question

    • farbwenz
      farbwenz
      Bronze
      Joined: 01.01.2010 Posts: 359
      Hi, I just switched from SNGs to Cash(at least for some time^^), and I would like to know what kind of statistics I need to do in order to know whether I'm winning or not, and if possible, how much.
      I played only 2500 hands so far(NL4- is it called NL4 at 0.02/0.04 Blinds?), or rather, watched 2500 hands, and I'm up by a few $. I'm playing at Cake and Elephant doesn't like the hand historys, so I want to do the whole stuff with excel.
      Since now I only wrote down the amount of money I gained /lost and the number of hands I've seen in one session(playing 4 tables). So I could do a graph with hand nr vs bankroll, but each session would be one value only, no matter how long the session or how many tables played. Though the values would be weighted by distance on the x-axis. I got:

      bankroll hands
      160,51 500
      160,35 823
      153,93 1024
      153,91 1162
      156,98 1418
      151,5 1707
      153,56 2027
      166,51 2427

      (seems like I'm improving:D In the last one I 6-tabled and I liked it better)
      My graph would consist of only 8 values yet. Over a large sample size I should get a linear digression with y = mx+n with n being my initial bankroll and m my ROI. Is it correct so far? can I use R² to show correlation? or how can I find out how much Variance affects my graph? Do I need to get different variables? I mean, apart from hands/h, I'll just get a few samples later and average over them to have a rough idea...
  • 4 replies
    • wasy8
      wasy8
      Black
      Joined: 29.01.2009 Posts: 1,507
      It's impossible to extrapolate results from such a small sample for a few reasons:

      1) Variance. Poker graphs are not linear - even the really pretty winner's graphs have huge fluctuations over 100k+ hand samples within like a 2.5 million hand graph.
      2) Changes in your play. If you're a robot then your play doesn't change, but players will adapt so even then you can't extrapolate over a small sample. But you're human so you will either get better (or you should be if you're studying) or develop leaks and bad habits if you don't study and your graph gets worse.
      3) Difficulty in the games. Related to (2), but the reality is that the winners are getting better at the game and even the losing players are being surrounded by information that is so easily available and it's pretty obvious that the fish are not as fishy as they used to be. It's impossible to know how that will change, but it would be a pretty reasonable assumption that a winning style of play now might not be a winning style of play 5 years from now, just like a marginally winning style 5 years ago is definitely a losing style today.

      Poker is dynamic - don't attempt to take a small sample and use it to predict future results. For example you can't take an hour of play where you made $200 and say that your hourly is $200/hour without averaging every other hour you play. Focus less on the statistics of really small sample and focus more on playing your hands.

      This means constantly asking questions about why you're doing what you do.
      -Why am I calling here instead of raising or folding?
      -Is raising more profitable than folding or calling here?
      -How much equity do I need to properly call here (based on direct and/or implied odds?
      -etc, etc.

      Hope this helps.
    • farbwenz
      farbwenz
      Bronze
      Joined: 01.01.2010 Posts: 359
      Oh sorry I think you got me wrong: I wanted to know how to do the statistics, so that I can start collecting the right data. for example, is the #of hands played while I was sitting at the table a good indicator?
      I wasn't going to extrapolate on such a small sample size, of course. I just would like to know whether my way of collecting data is correct, and how to interpret them (i.e. probability, that my ROI is really >0%;5%;10%). Also, I expect to get a slightly(hopefully more than slightly;) )upward curve in my graph because of learning. Than my linear assumption would always underestimate my real current winrate, which is fine for me ^^
    • wasy8
      wasy8
      Black
      Joined: 29.01.2009 Posts: 1,507
      Well first of all, your winrate in cash games is not measured in terms of ROI. It is measured in terms of bb/100 (big blinds per 100 hands). The number of hands you played while at the table is basically worthless as an indicator. 50k hands is a reasonable sample size to start making assumptions, but even then its likely to be way off too.

      What I'm trying to say is that collecting this type of data is totally unnecessary and potentially harmful in your development as a player - since people usually feel compelled to come to conclusions with the data they have when the data has no worth in what they're trying to find anyway.
    • farbwenz
      farbwenz
      Bronze
      Joined: 01.01.2010 Posts: 359
      thanks, I'll call it bb/100 from now on:-) talking about hands, do you mean "hands you played yourself" or hands seen",i.e. "Big blinds posted divided by 9(#of players at the table)"? in other words: are hands I folded preflop also counted?

      The number of hands you played while at the table is basically worthless as an indicator

      here I was ambigous, I meant the variable "hand number" not it's value (2500).

      people usually feel compelled to come to conclusions with the data they have when the data has no worth in what they're trying to find anyway.


      that's just why I'm trying to collect data in a meaningful way. After playing my first 100K hands I just don't wanna be told "you should have collected data on....try again", so I want to start collecting the right data. I completely understand that my 8 values are far to few, just want to know whether the way of collecting them is correct/how I should collect them. As soon as I have evidence that I'm a winning player I want to move up a limit. suppose I had 100K hands played, and would like to know whether winning or not. what would be your questions to me?