Thursday, March 8, 2012

Skill level in Poker and Reliability in Trading Strategies

When playing poker, a crucial skill that most players lack is being able to determine an opponent's skill level. The large amount of variance makes it impossible to tie a player's skill level based on how much they won or lost at the end of the night. The only way to know whether or not a player was good was if that player kept honest records of their winnings over the course of at least 50 thousand hands. This was easy for online players, since there was software like PokerTracker.

This was the only method available to me because I was not a truly good player. I was good in terms of being profitable, but I was not good in that I could not adapt as fast. My mechanical way of playing only worked well against bad players and knowing I was outmatched, I stayed away from higher limit tables where the average skill level was higher.

Eventually, I did adapt and get a bit better. I found out there was a different way to determine skill level. I had not seen it all along because I was not skilled enough to understand. Many good players look like bad players to other bad players. It's only other good players that can see this and know to stay away. Being marginally skilled, I had incorrectly classified many of these players.

Here's how it works. 99% of the time, there's only 1 right play in poker, given a particular set of information. If you show poker scenarios to a group of good players, they will all agree on the same answer. In scenarios where the decisions are marginal (mathematical expectations for multiple decisions are very close) they will acknowledge that as well. If you show the same scenarios to a group of bad players, they will give different answers to all different scenarios. The likelihood of bad players making wrong decisions is actually pretty high. In terms of position, flop, turn, river, and all the betting sizes along the way, bad players have plenty of opportunities to make bad decisions. As a result, you only needed to see the cards of someone who played 1 or 2 hands to see if they've made any mistakes.

This is quite wonderful for many reasons. Most people who claim that they are profitable are not actually good. They merely remember the times that they win and casually write off their losses as entertainment. Most people don't keep detailed records of their play. I now had a way to access a player's skill without relying on what they tell me verbally and without requiring scarce historical play data.

This can be applied when researching stocks. Many analysts offer strategies to help promote the stock that they are recommending. Their strategy, on which they will stake their reputation, (how meaningful is that anyway? how bout refunding our losses?), is something that they may claim has been backtested. It only takes a few minutes of looking at financials of the recommended stock to discredit the strategy. This is like when a bad player makes the one mistake and you can instantly classify him as a bad player. Many companies are so terrible or marginal that any strategy that spits out a recommendation should be disregarded.

3 comments:

  1. A trading strategy, if based on quantifiable specifications, can be analyzed on historical data to project the future performance of the strategy.

    ReplyDelete
    Replies
    1. This is probably true. However, I don't know of any trading strategies which work with purely quantifiable specifications, except for those that attempt arbitrage. While more probably exist, there are probably limitations on how much money can be deployed to the strategy. I don't think there are any algorithms to read and interpret annual reports.

      Delete
  2. The very basic thing to learn how to play a poker is to be very good observer on your own game. By this, it will help you became a good poker player just like some experts on this field. This is right that trading strategy may be an effective way in order to become a skillful poker player.

    ReplyDelete