Search found 900 matches

by AlvaroBegue
Sun Jul 18, 2010 8:54 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Random number generating CPU
Replies: 5
Views: 952

Re: Random number generating CPU

Pft ;) not even a new feature on x86 CPUs by a long shot. VIA Padlock has been available for what in computer terms is geological ages. Its a good idea to have it on die. Capable of creating random numbers at rates of between 800K to 1600K bits per second, the VIA PadLock RNG addresses the needs of...
by AlvaroBegue
Thu Jul 15, 2010 12:50 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Transposition table random numbers
Replies: 45
Views: 4929

Re: Transposition table random numbers

Any situation where there is a pawn on the 5th rank flanked by two opponent's pawns is ambiguous. Given how easy to fix this is, why risk it? I don't think anyone can seriously argue that en-passant info shouldn't be hashed properly. Having a separate entry in a table for each possible pawn costs yo...
by AlvaroBegue
Wed Jul 14, 2010 10:42 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Transposition table random numbers
Replies: 45
Views: 4929

Re: Transposition table random numbers

bob wrote:While I agree it is possible to have ambiguities, how often would you find the _exact_ same position, but with ep captures possible on different squares?
1. e4, d5 2. e5, f5

1. e4, f5 2. e5, d5

That wasn't so hard to imagine...
by AlvaroBegue
Wed Jul 14, 2010 3:30 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Transposition table random numbers
Replies: 45
Views: 4929

Re: Transposition table random numbers

What is 769? It seems to be 64*12+1. I guess it's just the 12 piece types (although pawns on ranks 1 and 8 shouldn't matter much) plus one for whose turn it is? If that's the case, you are missing en-passant information, which can be important (Ruy-López lost a game against Diep in WCCC'99 partly be...
by AlvaroBegue
Wed Jul 14, 2010 12:50 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Transposition table random numbers
Replies: 45
Views: 4929

Re: Transposition table random numbers

If anyone is concerned about the randomness of the numbers in the Zobrist table (although I agree with Bob that it doesn't matter), you can come up with several tables using different methods and then XOR them together. The result will be at least as random as the most random "ingredient" you put in...
by AlvaroBegue
Thu Jul 08, 2010 3:58 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: root move ordering
Replies: 79
Views: 13240

Re: root move ordering

We also observed the program during games, and it was clear that the very last move searched turned out to be chosen surprisingly often. Are you sure it wasn't just the last move to be *finished* searching, as opposed to the last move in the array? I ask because I've been seeing that when splitting...
by AlvaroBegue
Thu Jul 08, 2010 3:19 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: root move ordering
Replies: 79
Views: 13240

Re: root move ordering

We also observed the program during games, and it was clear that the very last move searched turned out to be chosen surprisingly often. Are you sure it wasn't just the last move to be *finished* searching, as opposed to the last move in the array? I ask because I've been seeing that when splitting...
by AlvaroBegue
Thu Jul 01, 2010 3:38 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: root move ordering
Replies: 79
Views: 13240

Re: root move ordering

Here's the question. When you say "solved the problem" how do you measure that? My measurement tool is to play 30,000 games with the change and compare that to the previous version. That is a real objective measure that is hard to ignore. We used a small battery of test positions that we understood...
by AlvaroBegue
Thu Jul 01, 2010 3:14 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: root move ordering
Replies: 79
Views: 13240

Re: root move ordering

I've been doing the q-search root ordering for years in Crafty, and it works well for a first cut. But once I start searching, I always force the best move to the top of the list, and the rest get ordered by the size of their trees. I tried keeping old best moves at the front, but it was not as goo...
by AlvaroBegue
Thu Jul 01, 2010 2:24 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: root move ordering
Replies: 79
Views: 13240

Re: root move ordering

I implemented a hybrid approach that worked really well for me in Ruy-López. I first ran quiescence search with full window on every move and sorted based on the resulting score. Then whenever a move is selected as the new best-so-far, I move it to the beginning of the list. I also remember how many...