Single Comment

When you talk about 5 or 6 seconds per turn, um, what kind of architecture is Chess V being run on? I'd think that space requirements are much more important than time requirements are. I apologize, but I have browsed here and there, and didn't quite notice a definitive statement as to what the exact hardware requirements for Chess V were. I'm naturally referring to Chess V running Ultima/Baroque in an environment all of its own, no phone lines involved, no other programs or applications eating up the available cycles (or semicycles) of execution. Even if it only requires 64 bytes to represent an 8x8 board, it helps to have some way of storing the moves leading up to that board, if you can't store the entire board by itself. Then set aside space for pointers with each pointer taking up 4 bytes so you can refer to an entire 32 bit address space (which is suficient for serious analysis). In attempting to make an Ultima/Baroque program on a 9x9 board, and starting from a single diagram thereof, I use up about 80K of RAM for every few tiers (or plies) worth of scrutiny (the 9x9 version of the game has the apparently innocuous Pusher and Puller pieces in it), so that may not be much memory to you, but my own code is very tight and efficient. When it comes to pruning the mini-max tree, it helps to avoid looking at moves that are undesirably similar. Since some moves are very 'similar' (inasmuch as they do not change the board much from turn to turn), the searches that I subject to the most scrutiny are those where a piece attempts to move as far as it can. For instance, if a Leaper - whether it is a Long Leaper or just a Single Leaper - moves to a place where its overall mobility is maximized, that is going to be the move that deserves the deepest investigation. To prove the point, see what happens when you pit two programs against each other, one of them preferring to examine the moves where the pieces consistently short-sheet themselves, opting for short moves when they have a choice, and those where the pieces usually move as far as they can. For instance, two entirely different personalities are at work when they come up with moves like 1. Pa2-a3, Pa7-a6, 2. Pb2-b3, Pb7-b6, 3. Pc2-c3, Pc7-c6 as opposed to a game like 1. Pa2-a6, Ph7-h3, 2. Pb2-b6, Pg7-g3, 3. Pc2-c6, Pf7-f3. The player that willingly short-sheets himself may be playing it safe, but he certainly isn't playing it adventurously. As for phobias and tropisms, it is a judgment call when a King must willingly go after an Immobilizer (for a coordination attack, as is usual, or a pinch, which is less usual), or an Immobilizer must flee from an Imitator (Chamaeleon). Is 25 percent really the proper amount of diminution to be suffered in calculating out a material score? How do we know that 20% or 33% isn't a better percentage? As I think someone else already pointed out, the piece values tend to change as the board grows emptier and emptier, so a sliding percentage might be better than an absolute percentage. Myself, I don't even use the 'percentage' idea, I just use an absolute plus or minus factor added to the positional score, and attempt to scope out another level or ply, even if it is incomplete, and hope that that is going to be enough.