Duckfest Digest 03 Poker Perspective

Introduction

As mentioned in my first article, I’ve returned to the game of chess after a long absence. It’s absolutely astonishing how much has changed. I described earlier how I enjoyed the content freely available on YouTube. The features offered online by chess.com and other websites are also amazing, like lessons, online courses and I really enjoyed the puzzles. But the game analysis tool is an absolute gamechanger. For people who grew up in the digital age it may seem obvious, but the availability of these chess engines (at zero costs) is mind blowing. I was in highschool when Deep Blue beat Kasparov, a project IBM spent $ 100M on. An engine that absolutely annihilates Deep Blue is available to me to analyze every game I play and tell me what I did wrong. 

I immediately loved game analysis. Still do. However, there is something I missed.

Poker game analysis

Fifteen years ago, when poker had it’s huge boom in popularity, I played poker online.  Poker, specifically No Limit Holdem, is said to take a minute to learn and a lifetime to master. It’s a very difficult game to master, for entirely different reasons than chess is. Chess is a game of perfect information, while Poker is a game of imperfect information. In chess a best move is a best move, in poker a best move means that it is statistically the best move. Poker has an extreme amount of variance, where chess has none. In theory at least. In practice, there is variance in chess, which is a key reason I’m writing this article. 

While poker is characterized by unpredictability, in some ways, it is remarkably predictable. You’ll get pocket aces (AA as starting hand) every 1 in 221 hands. At first glance this looks like an infrequent occurrence, but it’s not. When you are playing 12 tables simultaneously (which sounds absurd but is fairly feasible) at 60-80 hands per hour for 4 hours straight, you are dealt 3,360 hands in an evening. You will play pocket aces 15 times per session (or roughly every 15 minutes). 

In poker you determine a move not on the outcome of the hand, but on the long term outcome. What the outcome would be if you would repeat the same position 100s if not 1000s of times. Common tools to use were Holdem Manager and /or Poker Tracker. These allowed you to analyze all hands you had played. It was easy to select which hands gave you a good winrate and which hands didn’t. You could select and deselect other factors like the position you were in or all actions up to that point. 

Of course, you needed a large enough sample size (analysing poker games is a pure statistical analysis) and still needed to look at specific hands played to come to a reliable assessment. But in general, you could easily pinpoint which areas of your game needed improvement. How to improve, that was something you needed to find out for yourself.  

In chess the challenge is the other way around. Every chess engine can give you the best move in any position, but few tools are available to show you where you need to improve on an aggregate level. Sure, it’s easy to compare how you perform against e4 vs against d4 and maybe a bit further. But chess is not approached from a statistical perspective. In chess two almost identical positions are treated as completely different positions. They are of course different positions with different evaluations and different best moves. That is not my issue. My issue is, It's so difficult to compare positions that are roughly equal. In poker, AJs on the button vs AJs under the gun are different hands, but you can easily compare how well you do in both.

A practical example

The London system allows some flexibility regarding castling. You can castle both kingside and queenside. Or not.  I’ve had some of my most spectacular wins without castling at all. Do I perform better castling kingside or queenside or not at all? And what determines which way to go?  An easier question might be, how often do I castle kingside when I play the London? I was surprised by how difficult this was to answer. 

A better phrasing of the difference might be

In poker analysis  it’s easy to find the patterns, it’s not easy to find the answers. In chess it’s easy to find (specific) answers, it’s not easy to find the patterns. 

This is an oversimplification.  But you get the idea

Identifying your weaknesses

The way poker and poker analysis tools work encourage players to focus on specific parts of their game play. On poker forums players ask 

• My win rate with AJs and ATs is too low, how can I improve?
• My win rate with high pocket pairs is great (AA, KK and QQ), but with pocket Jacks (JJ) I lose money, especially out of position. Should I raise more preflop or should I fold more when facing opposition?
• Against an aggressive 3-better on the button, what should my 4 bet range be?

The way people asks question in chess the chess community are

too specific:

• Can you please comment on this game I just played?
• What’s the best move here?
• Why is this a blunder?

or way too broad:

• how can I stop blundering?
• why do I suck? 
• how do I suck less?
• can I still become a GM?

In chess I rarely see people refer to specific game areas. It’s not just that people posted higher quality questions 15 years ago and communities did more to enforce high quality questions. It’s because it is fundamentally difficult to discover patterns in your game play.

Bonus observation: In poker “what is a good winrate?” is a perfectly reasonable question, while in chess “what is a good rating?” is a completely irrelevant question (though it remains a popular question). 

Poker perspective is a probability perspective

Against a lower rated opponent, with a large enough gap in skills, the better chess player will win every game. If Daniel Negreanu and Magnus Carlsen play a game of chess, he will not win a single game. Ever. On the other hand, if Magnus and Daniel play poker for an evening there is a chance Magnus will win. Given enough games Negreanu will beat Carlsen, of course. But he will not win every hand. Maybe he will not even win more hands than his opponent. Fact is, in poker you are not supposed to win every hand. A good poker player wins more money (or chips) than he loses in the long run. 

For me it feels intuitive to look at my chess play from this long term, more statistical standpoint. I am not supposed to win every game. I want to win more in my next series of games than in my previous series of games. In principle this means that if I play against similarly rated opponents, my expected winrate will be 50%. If I play 100 games, I will on average win 50 games and I will lose 50 games.

That's why I feel an individual game doesn't have that much importance. Not all moves are equally important and not all blunders are created equally. Even though losing a game due to a massive blunder makes you feel like the worst player in the world, that’s not always where you need to improve. A small but fundamental misunderstanding of opening principles, that affects all of your games is much worse than blundering away a certain win every now and then. That’s why I disagree with the common idea that at lower ratings, the game is won by the person who makes less blunders. (As a side note: In the Gotham Chess Guess the Elo rating, I noticed that when Levy looks at a game, his assessment is based on all moves. In some cases it even happened that a Queen blunder had no real impact on his guess.). Sure, if I want to get better I should blunder less. But if I really want to improve, I need to focus on parts of the game that have a bigger impact.

So, the question for me is

How to play better in my next 100 games than I did in my previous 100 games?