WyldChess 1.4 released

[mkchan]

Changes since 1.3:
-> Fixed time control bugs where the increment was greater than the total time left
-> Reversed futility pruning(now in the move loop)
-> Fixed check extension bug
-> Improved(and simplified) passed pawn evaluation
-> Refactored move ordering
-> Reverted to old LMR formula(log-based formula missing crucial moves occasionally)
This version is probably ~50 elo stronger and easier to tinker with.
If someone can provide a MacOS compile, I'll add it to the release
Find it here: https://github.com/Mk-Chan/WyldChess/releases/tag/v1.4

[Graham Banks]

Sounds like a nice improvement.

[mkchan]

Hopefully it won't be a fiasco like 1.2
but i did test it in self play and a few other engines

[pkumar]

Time management problem in Arena-3.0 gauntlet match at 40 moves in 1 or 5 mins repeating - consumes allotted time too early.

[mkchan]

Time management problem in Arena-3.0 gauntlet match at 40 moves in 1 or 5 mins repeating - consumes allotted time too early.

I tried a few games with your mentioned settings and it doesn't seem to be losing on time(not sure if it is for you). If it isn't losing on time, it shouldn't be a problem

[pkumar]

Time management problem in Arena-3.0 gauntlet match at 40 moves in 1 or 5 mins repeating - consumes allotted time too early.

I tried a few games with your mentioned settings and it doesn't seem to be losing on time(not sure if it is for you). If it isn't losing on time, it shouldn't be a problem

I shall elaborate.

I need a collection of stable engines in the Elo range 2500-2600. WyldChess 1.4 would be in that range with the expected Elo increase of ~50. So I set up an Arena 3.0 gauntlet match with Trace 1.37a (CCRL Elo 2505). Time control was 40 moves in 1 minute repeating - my usual setting. Neutral starting positions from Dann Corbit were used. In all the games WyldChess used up nearly all the allotted time by 15 moves and made the remaining moves after very shallow searches. Naturally WyldChess lost - often from equal or leading positions - with the following result:

Wyldchess1.4_popcnt - Trace : 0.5/32 0-31-1 (00000000000000000000000=00000000) 2% -676

I am perplexed because WyldChess 1.0 64-bit has been tested in CCRL for Elo 2476.

[mkchan]

Time management problem in Arena-3.0 gauntlet match at 40 moves in 1 or 5 mins repeating - consumes allotted time too early.

I tried a few games with your mentioned settings and it doesn't seem to be losing on time(not sure if it is for you). If it isn't losing on time, it shouldn't be a problem

I shall elaborate.

I need a collection of stable engines in the Elo range 2500-2600. WyldChess 1.4 would be in that range with the expected Elo increase of ~50. So I set up an Arena 3.0 gauntlet match with Trace 1.37a (CCRL Elo 2505). Time control was 40 moves in 1 minute repeating - my usual setting. Neutral starting positions from Dann Corbit were used. In all the games WyldChess used up nearly all the allotted time by 15 moves and made the remaining moves after very shallow searches. Naturally WyldChess lost - often from equal or leading positions - with the following result:

Wyldchess1.4_popcnt - Trace : 0.5/32 0-31-1 (00000000000000000000000=00000000) 2% -676

I am perplexed because WyldChess 1.0 64-bit has been tested in CCRL for Elo 2476.

I am unable to reproduce the problem you seem to be having. I setup the following in Arena 3.5.1 and Arena 3.0:
Wyldchess1.4_popcnt - Trace
Tournament 40moves/1min repeating
Gauntlet
It's running perfectly fine for me. If possible could you please show me the logs when you start the tournament (the setting up of the engines)

[brianr]

WyldChess crashes at the end of most games when exiting (the non-popcnt Windows .exe)

This happens when run in a plain cmd console window with just the basic
uci
isready
go infinite
stop
quit
command sequence.

Some tournament managers may not be happy out the engine not exiting cleanly.

[mkchan]

WyldChess crashes at the end of most games when exiting (the non-popcnt Windows .exe)

This happens when run in a plain cmd console window with just the basic
uci
isready
go infinite
stop
quit
command sequence.

Some tournament managers may not be happy out the engine not exiting cleanly.

Thanks for informing me Brian, I fixed the bug and reuploaded the binaries.