brtzsnr wrote:Do you know how much ELO KPK and KBKP give? My current Elo density is about 0.5 Elo / line (CCRL, only engine, no uci, no tests) and it's something I consider everytime I add significant chunks of code.
Was long ago that I implemented kpk bitbase from H.G. Muller:
I don't know how much elo added, but was a net win.
And I have not implemented this kbpk big thing that I have shown here. I have a modest more simple way of dealing with it. Maybe at some time I will try.
As usual such patterns are not so much about Elo as they are about having a more complete program.
It's embarrassing to show +2 something when every beginner in chess knows when a kpk position is a draw and it hurts to see your engine convert a significantly better position to such an elementary draw.
But these are relatively common. If you run a tournament in the background on your computer for a day odds are good you'll see this ending especially against engines that don't know about it.
Wrong bishop and edge pawn is even more common. Think I faced those twice in OTB computerchess tournaments.
Problem comes when the bitbase comes later, when the position is a draw.
So some complementary code may cause the motor to see it before,
and not simplify immediately.
