Random SOO

by zenquaker

Okay, moving forward on Chess. I have decided not only to shift my work to my correspondence Chess, but also to run an experiment with it as part of my greater data collection.

One thing new players are often interested in is opening theory. Chess has been around for hundreds of years, and all that time the best players have been thinking about what are the best moves to make early in the game. With the advent of computers this study has become more intense. Not only can computers play out hundreds of games from a given opening position, but they can do statistical analysis on large data sets of games played by human masters.

The better players have what is called an opening repertoire: a set of opening moves that lead to positions they’ve studied, feel comfortable with, and fit their style of play. They have to adapt to what their opponent does, but even there they can steer the game in directions they prefer.

Newer players are discouraged from worrying too much about opening repertoires. There are thousands of ways to play out the opening that lead to positions that are so even that newer players lack the skill to see who has the advantage, or how to exploit it. I have been playing for a long time, but my skill level is such that I still fall into this category.

But in correspondence Chess you can use databases of opening moves. You can see all the moves made by masters from a given position, and get statistics on them depending on the database program. Some think of this as cheating. Others think it’s silly for weaker players to do this, because they will find themselves in positions they don’t understand.

So here I am with the resources to make any reasonable opening out there, but I’m too weak to understand what positions are good for my style and level of player, much less understand what positions I won’t understand. That’s where the experiment comes in. I am going to use a Chess opening database to make random moves. Each move in the opening I will use the database to find all the moves that are made at least 9.5% of the time and has at least 64 games in the database.

Why 9.5% of the games? There’s two kinds of openings: classical and hypermodern. Using 9.5% just barely allows for one hypermodern opening by white in the database I am using. Why at least 64 games? As a statistician, I don’t trust small sample sizes. Once there are no moves with at least 64 games I’m leaving the database behind.

Of course there is still the issue of not understanding the positions I am in. However, I have a Chess book that attempts to explain what is going on in the openings. As I play randomly I am going to read along in that book and try to understand what is going on in the game, making notes of my dim understanding in my record of the game.

I will try to play as many games as I can using this methodology. Then I will analyze them and find out which opening moves lead to better results. Those moves will form the basis for my opening repertoire next year. How deep I will be able to analyze this will depend on how many games I manage to play, and I may not get enough games to perceive a statistically valid difference. And between now and then I will have to decide what constitutes a practical difference, and what to do if there isn’t one of those. While there are difficulties, if this does work out I can iterate it over many years. I don’t expect to be able to analyze more than two or three moves into the game. But once I have found the best moves (for me) in the first two or three moves, I can start always playing those and do the experiment on the next two or three moves.

I have lots of plans on how I’m going to play the rest of the game. But I’ll bore you with those later.