lichess.org
Donate

What is the centi pawn loss cap? Lets see...

#1
en.lichess.org/JYRSYbO2

Was seeing just how high I could get the centipawn loss. Set up both sides to have mate in 1 on each other... Then left the mate in 1 vs mate in 1 situation in stasis doing waiting moves for both sides for a great amount of time... 102 blunders in the game.

Can anyone beat 1740?
#3
Well, most people know how to deal with it. And it works pretty well on average in its limitations. Sure one can create exceptions where it fails miserably. I knew this before, does anyone feel better or worse now?
#5
i got 1921 for white and 1940 for black after 51 moves:
en.lichess.org/ozkOVg9g#6

I'm thinking the cap is 2000. I think it sets the eval limit at +10 and -10, so that's an eval change of 20 pawns a.k.a. 2000 centipawns per move.
#6
i made this game back in the day (1970 white acp and 1980 black acp) which still is the world record, as far as i know en.lichess.org/GZ1EKrJR
i'm pretty sure i applied all the optimizations possible to exploit the small lichess quirks (evaluation caps at +10 and -10, only the first 100 moves are counted).
here's my proof that this is the highest acp possible:

- instead of going for fastest mutual mate in 1, i just play fool's mate for white and leave black's queen hanging as soon as possible. that's enough to swing the eval to over -10 and +10 as soon as move 2
- according to the engine, 1 g4 is worse than 1 f3 or 1 f4, and 1..e6 is worse than 1..e5 (i have to move the e-pawn to set up the fool's mate)
- the only other reasonable method of giving black over -10 right on move 2 would be by hanging the white queen, but then black can't make a 2000 centipawn blunder on move 2

Q.E.D.
it's up to you to prove me wrong :o
#7
wow fate that's really smart, I think you have it figured out. I didn't know that only the first 100 moves were evaluated, otherwise you could get it to 1999 or 2000 I'd imagine.

This topic has been archived and can no longer be replied to.