@labobo Yeah, I agree. The four-block pieces (tetrominos) are really the optimum for a fun game. 3 poses no challenge whatsoever, and 5 gets too frustrating after a while, even though an AI can work with it for surprisingly long time. I'm not sure about the field's width and height though. 10 is the standard for the width, but it is rather arbitrary.
@labobo The line length actually makes the game easier, not more difficult. Too narrow rows means a very soon game-over. However, a narrower board allows a better AI because the search space is smaller. I tried out 20x50 at COM3 (this video has 64x50 at COM1); it managed to clear one row before gameover. See bisqwit.iki.fi / kala/tetre8b.png Note: Increasing the ply level obeys the law of diminishing returns, so you wouldn't see anything particularly exciting with e.g. COM60 compared to COM5.
Wouldn't be a very fun game for a human, i think
labobo 1 month ago
@labobo Yeah, I agree. The four-block pieces (tetrominos) are really the optimum for a fun game. 3 poses no challenge whatsoever, and 5 gets too frustrating after a while, even though an AI can work with it for surprisingly long time. I'm not sure about the field's width and height though. 10 is the standard for the width, but it is rather arbitrary.
Bisqwit 1 month ago
Due to the many different pieces of blocks and the line length makes the probability of making a complete line too low.
labobo 1 month ago
@labobo The line length actually makes the game easier, not more difficult. Too narrow rows means a very soon game-over. However, a narrower board allows a better AI because the search space is smaller. I tried out 20x50 at COM3 (this video has 64x50 at COM1); it managed to clear one row before gameover. See bisqwit.iki.fi / kala/tetre8b.png Note: Increasing the ply level obeys the law of diminishing returns, so you wouldn't see anything particularly exciting with e.g. COM60 compared to COM5.
Bisqwit 1 month ago