ren
User
Newbie
| Posts: 4 |   |
Karma: 0
|
Using Transposition Table with Wildly Varying Score - 2007/01/25 21:25
This isn't strictly a chess question, but relates to something you guys ought to be familiar with. Say the game you're conducting the alpha-beta search for is one in which the raw material consistently changes radically from one ply to the next. Say we're using a naive static eval, which is just our material value minus their materal value. Then, we'd have something like this:
Imagine player 1 is judging the value of a move:
at depth move A looks worth
1 5 points
2 -4 points
3 6 points
4 -5 points ...etc.
If you're using a transposition table, would you still want your iterative deepening loop to increment depth by one ply each iteration, even though the score is flip-flopping? Or would you want to increment depth by two ply, so the score "curve" remains "smoother?"
Also, say the computer is searching to depth eight, and using a transposition table. The computer does its thing, moves, and then the human moves. Is it correct that the transposition table should be being used exclusively for the first six ply of the computer's next search?
That's what's happening in my program--the first "n minus two" ply of the following search happen instantaneously, then the last two ply require a regular search. I mean, it's pretty neat if that's what's supposed to happen, but it seems too good to be true!.
---------
The days are too short even for love; how can there be enough time for quarreling? - Margaret Gatty
Popular posts by ren
Choosing Static Evaluation Features...
Kasparovchess Software
chess ebooks
|
post new topic