To solve this, it is going to be done in evaluation, not in search. There is a class of such positions that the programs don't understand (yet). But that will change. I'm going to do fortress-type analysis one day...
To solve this, it is going to be done in evaluation, not in search. There is a class of such positions that the programs don't understand (yet). But that will change. I'm going to do fortress-type analysis one day...
Im curious, how would you recognise fortress like positions ?
By loooking at the PV and see if there is some kind of progress or doing some static eval ?
my idea was not saying that a position was a draw just lowering the score gradually if the pogram just shuffled the same pieces around, im not sure if that is the same as the one(Little Goliath?) you referred to.
To solve this, it is going to be done in evaluation, not in search. There is a class of such positions that the programs don't understand (yet). But that will change. I'm going to do fortress-type analysis one day...
Im curious, how would you recognise fortress like positions ?
By loooking at the PV and see if there is some kind of progress or doing some static eval ?
my idea was not saying that a position was a draw just lowering the score gradually if the pogram just shuffled the same pieces around, im not sure if that is the same as the one(Little Goliath?) you referred to.
With curious regards
By far the most common type I see has to do with a king and queen vs a king rook and pawn, where the pawn holds the rook, and the king is in the area protected by the rook so that the opposing king has no way to get in and trading the Q for the R+P is simply a draw. There are other types of fortresses, but they are all based on the same basic idea...
One can try the Slate-like "weariness factor" evaluation where no progress over many moves pulls the score toward a draw, but I am probably going to recognize the pattern first, and see how that works...
I am not sure it is a fortress draw after the moves you suggest. The way to find out is to play it out (you with the black pieces against crafty with white).
After you try to set up the fortress, white can play the king to h4 and then move g4, which looks promising.
José.
bob wrote:
One can try the Slate-like "weariness factor" evaluation where no progress over many moves pulls the score toward a draw, but I am probably going to recognize the pattern first, and see how that works...
This is what i suggested or very like it, i would just look at the pieces also when checking the PV
) blunder. After for example 4...Nc7 there isn't any zugzwang