DeepButi
October 1st, 2004, 08:38 AM
As a small part of a learning module for a bot playing a trick tacking card game, I use a minimax.
Move ordering, forward prunning and transposition tables are all coded and working fine.
Preliminary tests show MTD(f) performs better than NegaScout, but this is not yet definitive.
Now, I'm trying to get the Principal Variation and it's getting me mad ...
To make things easy, I use simple NegaMax. It works WITHOUT Transposition Tables, but I'm unable to find the PV with TT.
Since it's a fixed sum game with a known depth (number of tricks) I always examine to the end of game (or a cut-off) and I can store the PV on the TT and reuse it according to a lower/upper/real flag. It does work for the value and best move ... but somehow it does not work for the whole PV.
Google has all answers ... but I'm unable to find them. Almost nothing on getting the PV and nothing at all on PV when TT are involved.
Can someone point me in the right direction?
Thanks in advance.
Move ordering, forward prunning and transposition tables are all coded and working fine.
Preliminary tests show MTD(f) performs better than NegaScout, but this is not yet definitive.
Now, I'm trying to get the Principal Variation and it's getting me mad ...
To make things easy, I use simple NegaMax. It works WITHOUT Transposition Tables, but I'm unable to find the PV with TT.
Since it's a fixed sum game with a known depth (number of tricks) I always examine to the end of game (or a cut-off) and I can store the PV on the TT and reuse it according to a lower/upper/real flag. It does work for the value and best move ... but somehow it does not work for the whole PV.
Google has all answers ... but I'm unable to find them. Almost nothing on getting the PV and nothing at all on PV when TT are involved.
Can someone point me in the right direction?
Thanks in advance.