I am testing various rating systems (Elo rating and some alternatives). I need a"world" with computer players participating in championships. Computer players are characterized by two parameters:
(1.1) Depth of search (probability to search yet another ply deeper)
(1.2) Probability to remember the outcome in their previous games. (Relatively high)
(1.3) Probability to learn from "published games". (Relatively low)
I tried tic-tac-toe, but there are two problems:
(2.1) Players learn too fast. (Decreasing learning rate turns it in a toss-a-coin)
(2.2) Ply count is granular. The difference between "can/cannot see a fork" or "can/cannot see opponent's fork/one move win is huge.
A battleship game doesn't have complete information. A game like chess or five in a raw would require prohibitively large amount of memory to remember all possible combinations.