- The previous PV of iterative deepening is feed back into the search. This is been implemented for a while, but I fixed a bug here which caused the engine to hang. A slight difference is that the entire PV line is used, even if a better line is found the previous line is still investigated first. I was quite amazed by the speedup gained from this.
- Quiescence search now uses a Static Exchange Evaluation (SEE), which turned out far easier than I thought and to my surprise, caused a fair amount of speedup. The SEE I use also uses an early exit technique. Say for example, I have a pawn and queen attacking a bishop, while my opponent has two pawns protecting the bishop. The full SEE evaluation will result in a negative value: 3-1+1-9+1=-5 (I gain a bishop and a pawn for a queen and a pawn). With early exit I only realize that I can stop after capturing the knight, resulting in +2. A great move!
I am now concentrating my efforts on the ordering of normals moves as part of alpha-beta (in conjunction with the PV-value). So far the results are very promising, but it still needs some work before being released.