Kevin Pacey wrote on Sat, May 27, 2017 04:31 PM UTC:
At the risk of getting side tracked Fergus, I can see how a fairly large body of tentative opening theory might quickly be developed for a given chess variant. Get two (or more) fairly strong and seperate fairy chess engines to play each other many, many games of the variant, at a brisk (but not quite blitz, perhaps) time control, and put the games into a large database over time. Initially have each engine give an evaluation on the position following each sequence of moves from the start of a given game, after a certain number of moves, but only permantly record the winning engine's evaluation (in case of a draw, use the average evaluation of the two engines). Then anyone might study a portion of the database results at leisure, and decide what to trust or perhaps even write books about as far as a variant's opening theory is concerned.
In the case of chess, I use a certain opening database program that includes an engine's evaluation of a position at more or less almost every point of a sequence of possible chess openings (at least the fairly major ones; many human evaluation symbols are thrown in for comparison too), so the idea I put forward in the paragraph above may not be that unrealistic conceptually anyway. A serious drawback that the chess opening database has which I use, as far as I'm concerned, is that it does not helpfully point out when a position is a forced draw, let alone if it's a drawish position that's hardly worth playing out for a chess master at least, though even chess books by humans often fail to point such things out.
At the risk of getting side tracked Fergus, I can see how a fairly large body of tentative opening theory might quickly be developed for a given chess variant. Get two (or more) fairly strong and seperate fairy chess engines to play each other many, many games of the variant, at a brisk (but not quite blitz, perhaps) time control, and put the games into a large database over time. Initially have each engine give an evaluation on the position following each sequence of moves from the start of a given game, after a certain number of moves, but only permantly record the winning engine's evaluation (in case of a draw, use the average evaluation of the two engines). Then anyone might study a portion of the database results at leisure, and decide what to trust or perhaps even write books about as far as a variant's opening theory is concerned.
In the case of chess, I use a certain opening database program that includes an engine's evaluation of a position at more or less almost every point of a sequence of possible chess openings (at least the fairly major ones; many human evaluation symbols are thrown in for comparison too), so the idea I put forward in the paragraph above may not be that unrealistic conceptually anyway. A serious drawback that the chess opening database has which I use, as far as I'm concerned, is that it does not helpfully point out when a position is a forced draw, let alone if it's a drawish position that's hardly worth playing out for a chess master at least, though even chess books by humans often fail to point such things out.