perft for different components of the evaluation

[Uri Blass]

I wonder if there is a program that has commands like perftpass when the program simply counts number of passed pawns in the leaf positions of games of n plies.

for example in the opening position 1.a4 b5 2.b4 bxa4 should be count as 1 when you calculate perftpass(4) because black has one passed pawn(pawn a4).

It means that in the opening position
perftpass(4)>=1 inspite of the fact that it is sum of numbers that almost all of them are 0.

1.a4 b5 2.b4 bxa4 3.b5 a5 4.bxa6 is calculated as 2 in perft 7(white has a6 and black has a4 so there are 2 passed pawns in the board)

Uri

[Jaime Benito de Valle]

Your engine has a perft. Why don't you add this feature yourself? It should be adding only a few counters. I'm sure I could add this for my perft pretty fast if you have a problem with it.

Jaime

[Uri Blass]

My implementation is not efficient(I calculate perftpass(n) at the same time that I calculate perft(n+1) and I spend time by generating moves that I do not need for perft(n+1) but here are results in the opening position

perftpass(4)=6
perft pass(5)=575
perftpass(6)=38791

Do you get the same numbers?

[Uri Blass]

Note that the number for white passed pawns are
perfpasswhite(4)=2
perfpasswhite(5)=371
perfpasswhite(6)=14894

Uri

[Dan Honeycutt]

Hi Uri.

Just curious; why do you want to do this?

Best
Dan H.

[Uri Blass]

I want to verify that I have no bugs in functions to calculate data.

There is data like passed pawns that all agree about the definition of it
I think that perft numbers for data can be productive to verify no bugs.

Uri

[Dan Honeycutt]

OK, I see. I'm pretty confident in my program's ability to correctly find the passed pawns. Now, how to score them ....

Dan H.

[Uri Blass]

I believe that it also correct but I am 100% sure of nothing and even with 90% confidence that every term is ok by itself I can expect to find bugs in one data information for 10.

Uri