Henk wrote:In my chess program I tried out several "good ideas" that did not improve performance:
1) Mate distance pruning
2) Futility pruning in Quiescence search
3) LMR
4) Hash table ( This I have to try out again)
Are there more good ideas that may not work ?
I like to have warnings in advance.
5) History heuristic (not killer moves)
6) lazy evaluation when normal evaluation is fast enough
7) alfa beta instead of pvs
8) MTD(f)
9) Aspiration windows
Wow, if none of these improved performance your program really must have some serious bugs.
You should throw it all out and carefully start from scratch - there are certain things that should work without question and focus on them.
Here is what you should focus on:
1. Hash table
2. PVS
3. LMR
4. Null move pruning.
Start with those and get them right - then call us in the morning.
Repeating messages over and over:
With 7) I mean that PVS works better than alfa beta in search. So alfa beta compared to PVS does not improve performance.
8) I gave up MTD(f) for it gives problems with pruning and accurate evaluation . But if pruning and accurate evaluation doesn't help much MTD(f) may still be a good candidate.
Don wrote:
Start with those and get them right - then call us in the morning.
If I would agree on this I have to predict future. And I cannot. I don't know state. The past is already a problem. OK I have to limit posting messages.
Henk wrote:In my chess program I tried out several "good ideas" that did not improve performance:
1) Mate distance pruning
2) Futility pruning in Quiescence search
3) LMR
4) Hash table ( This I have to try out again)
Are there more good ideas that may not work ?
I like to have warnings in advance.
I do not see any reason why all of those would not work. The first is a TINY gain, tiny enough it is likely worthless as an improvement in actual games. But the others? Futility pruning (I assume you mean ignore losing captures AND captures that have no chance of getting the score above alpha even though the capture wins material (win a pawn but you are a rook behind.)
Your statement suggests you need more time to develop and test as all of those are known good ideas.