Article: Variance over the increasing hand sample

    • wnbMG
      wnbMG
      Bronze
      Joined: 10.04.2011 Posts: 586
      What makes poker the most thrilling and painful at the same time? Variance. Everyone can be a winner. But for how long?
      A few months ago me and my poker buddies had the debate: ‘What happens with the variance over the increasing hand sample?’ The answer seems obvious to everybody, but surprisingly it was different. Everyone’s arguments, reasoning and logic seems legit. But statistics can be tricky and sometimes, just like in poker, your ‘feeling’ is not enough. So I, the nerdiest guy in the group, decided to show off.

      Basics
      “Half of the answer lies within a good constructed question,” wise man once said. If I’m first to say it, then I’m that wise man. So, let’s break it down. First, definition of variance: ‘Variance is a number which measures how far a set of numbers is spread out around mean.’ More useful term is standard deviation which is the square root of variance. Both represent how swingy your poker results are. But standard deviation can tell you something more. Let’s get right to the poker example. Our hero plays 0 EV (mean) game with standard deviation 100 big blinds per 100 hands1 (he will stay the same rake back grinder through the whole post). Because of this exact standard deviation, there is 68% chance2 his bankroll will change less than 100 big blinds after 100 hands.
      But what would happen if hero played a different number of hands?

      Variance of winnings
      Let’s start with the obvious one, variance of winnings. It tells you how much swings you can expect in big blinds or moneywise. But, does it get smaller or bigger with more hands played? Think for yourself but be careful. Your peasant logic can get outplayed by the lady variance. And this time you can’t blame her.
      Here are seven different simulated graphs of hero’s winnings trough 100k hand sample. To make those graphs more understandable you can think about them as graphs of players with the exact same skill level or it can be the same player3 who made 700k hands and those are parts4 of his graph.



      It’s kind of hard to believe, those are all simulated graphs of 0 EV player. From one, you would think he plays like BalugaWhale, from another, like beluga whale. As you can see, lines are getting more and more dispersed. That means bigger variance and also standard deviation. If you still have thoughts such as “luck will cancel out at the end of the day”, let me show you another graph.
      Let’s say hero plays 1k hands and at the end of a hand sample, he makes a dot of his winnings. Then he goes from scratch again and makes 2k hands. Dot again. Then 3k hands, 4k, 5k… So on, till the 1M hands. Here is what he gets:



      One more time, it gets dispersed.
      To make things clear, I even displayed a standard deviation line. It’s proportional to the square root of number of number of hands. Equation for standard deviation is:
      (1)
      std dev=K*sqrt(#hands)

      In our case, K is equal to5 10. For example, if your standard deviation is 85bbs/100hands, the value of K would be 8.5. You can actually calculate a few interesting and useful information from equation (1). You can see that any fixed hand sample has the same standard deviation no matter where on the graph it is located.
      Those results are somewhat hard to understand at first sight so let’s bring a few examples, which make ‘logical’ sense:

        • You can’t win more than one stack in a single hand. If you play more hands, there is possibility to win more. That tiny portions add to dispersion.
        • Let’s say you roll a ball in a straight line on the concrete which is not completely flat and has some sand on it. Ball will eventually go out of the line. The longer way it makes, there is a higher chance it will be farther away from the initial line. The last graph represents this situation from bird eye perspective.

      Before you start playing as little as possible to lower your variance, read on.

      Variance of a win rate
      This case turns around the whole picture. Win rate is defined as big blinds won divided by number of hands6 played. Considering this we get a bit different standard deviation’s equation:
      (2)
      std dev=100*K / sqrt(#hands)

      Actually, something dramatically changed. Standard deviation now gets smaller trough the increasing hand sample. Let’s just divide the winnings with a number of hands/100 and here is what we get from Graph 1:



      Win rates are getting closer to 0 bb/100hands, which is hero's genuie win rate.
      Same thing for Graph 2:



      Obviously, the more hands you play, the more accurate your win rate should become. Using the (2) equation we can calculate how many hands you need to lower your standard deviation to 1 bb/100hands. That means, if your win rate is 1bb/100hands, you can be 67% sure your win rate is somewhere between 0 and 2 bb/100hands. And for that, you need 1 million hands. So grind on…

      Conclusion
      Understanding variance is one of the key factors of becoming a good poker players. Especially because of your mental game. You should expect sick things to happen and be capable of facing them. Don’t get too much into those poker software lines. They aren’t as good predictor of your skill as you may think, even EV line. Just remember – never underestimate variance and overestimate your poker skill. I wish you all the best at the poker tables.

      About the author
      wnbMG is a poker player who blames the variance for not making it as a pro. Nevertheless, he luckily had enough skill to become Slovenian poker champion (team event).



      Notes:
      1 This standard deviation is somewhat standard for no limit Texas hold' em shorthanded cash games.
      2 Variables have to be normally distributed. In our case, they are.
      3 We assume player and his opponent’s skill stayed the same for all 700k hands.
      4 We will use that example later on.
      5 Where do we get this 10? Our standard deviation should be 100 per 100 hands. So just insert this in equation you get. 100 = K*sqrt(100). K should be 10. It also means standard deviation for 1 hand.
      6 We usually define win rate in big blinds won per 100 hands. That’s why number of hands should be divided by 100.
  • 4 replies
    • Maverick290784
      Maverick290784
      Bronze
      Joined: 17.03.2008 Posts: 88
      So how lucky are you when you're a winning player at live cash games? Who can play 1 million hands at a live cash game?
    • VorpalF2F
      VorpalF2F
      Super Moderator
      Super Moderator
      Joined: 02.09.2010 Posts: 8,914
      Good question, Maverick290784.
      In live cash games there are personal skills that help that don't exist online.
      Hiding ones emotions, and reading the emotions of the other players, for example.

      Cheers,
      VS
    • wnbMG
      wnbMG
      Bronze
      Joined: 10.04.2011 Posts: 586
      Originally posted by Maverick290784
      So how lucky are you when you're a winning player at live cash games? Who can play 1 million hands at a live cash game?
      You don't play poker to see your winrate, you play poker because you think you have an edge. But after a decent sample you can check it with specific statistical tools.
      Btw, edges/winrates are ussually much bigger in live game, so you don't need as many hands as online to be certain if you are winner or not.
    • tieppofer
      tieppofer
      Bronze
      Joined: 25.01.2009 Posts: 282
      Originally posted by VorpalF2F
      Good question, Maverick290784.
      In live cash games there are personal skills that help that don't exist online.
      Hiding ones emotions, and reading the emotions of the other players, for example.

      Cheers,
      VS
      Here in my region there is a huge poker boom and a lot of people are starting to play at clubs, most of them play only live poker. I do agree with that argument, there are skills that you can use in a live table which are not available online, but on the other hand online you can get what - in my honest opinion - are the best tells about a player, that is, the stats and how he plays his hands.

      Originally posted by wnbMG
      Originally posted by Maverick290784
      So how lucky are you when you're a winning player at live cash games? Who can play 1 million hands at a live cash game?
      You don't play poker to see your winrate, you play poker because you think you have an edge. But after a decent sample you can check it with specific statistical tools.
      Btw, edges/winrates are ussually much bigger in live game, so you don't need as many hands as online to be certain if you are winner or not.
      I agree in parts, but even though winrates are usually much bigger, the variance can hit harder, specially depending on the situation you play. I usually play 5/10 and 10/20 games live and - well - sometimes it's really hard to be ahead, specially when you get those kind of players who don't really care about the money.

      As for the article, I think it's a very good inside for those who don't really understand statistics, I really liked it :f_thumbsup: