New engine: Stash

[Modern Times]

Hi,

I can’t find any official grading for this Engine.

Originally I believe it was estimated around 1450 ELO, but on PicoChess we are seeing v11 holding it’s own against Engines in the 2000-2200 ELO range.


Thanks,

Al.

There are some games on the CCRL chess960 list. 1732 Elo on that list so far. I suspect on the standard chess list it will be around 1800-1850.

[Scally]

Thanks Guys,

The Blitz list shows ELO 1415, as for others still a lot lower than we think ....

Thanks very much ...

Al.

[Modern Times]

Thanks Guys,

The Blitz list shows ELO 1415, as for others still a lot lower than we think ....

Thanks very much ...

Al.

That is the older 9.01 rather than 11.0.1 that is on the chess960 list.

[tmokonen]

I created a 64 bit windows executable for Stash 11.0.1:

https://mega.nz/#!PgUX1K4K!myQDEiRtOdQa ... 9nsMgcwlqQ

[mhouppin]

Hi there, version 12.0 is out !

I added Null Move Pruning and Late Move Reductions to the code. I'm now going to work on a better move ordering, with a Static Exchange Evaluation and a move history, and maybe a better time management implementation.

I'm still running matches against past versions to mesure the Elo gain, but my bet is around 60-70 additional self-Elo.

[tmokonen]

I received a request to make a generic non-popcount 64 bit compile of Stash-bot v12, so here it is:

https://mega.nz/#!G1sD3CJQ!FZAsMQ97NjYc ... OkasfWX81o

[Scally]

Hi,

Your latest code no longer compiles on my Raspberry Pi:

sources/uci/uci_bench.c: In function ‘uci_bench’:
sources/uci/uci_bench.c:72:26: error: format ‘%lu’ expects argument of type ‘long unsigned int’, but argument 2 has type ‘uint64_t {aka long long unsigned int}’ [-Werror=format=]
printf("Total nodes: %lu\n", total_nodes);
^
sources/uci/uci_bench.c:73:26: error: format ‘%lu’ expects argument of type ‘long unsigned int’, but argument 2 has type ‘uint64_t {aka long long unsigned int}’ [-Werror=format=]
printf("Nodes/second: %lu\n", (total_nodes * 1000) / bench_time);
^
cc1: all warnings being treated as errors
Makefile:48: recipe for target 'objects/uci/uci_bench.o' failed
make: *** [objects/uci/uci_bench.o] Error 1
make: *** Waiting for unfinished jobs....
Al@PicoChess:~/stash/src $

Thanks,

Al.

[mhouppin]

Hi,

Your latest code no longer compiles on my Raspberry Pi:

sources/uci/uci_bench.c: In function ‘uci_bench’:
sources/uci/uci_bench.c:72:26: error: format ‘%lu’ expects argument of type ‘long unsigned int’, but argument 2 has type ‘uint64_t {aka long long unsigned int}’ [-Werror=format=]
printf("Total nodes: %lu\n", total_nodes);
^
sources/uci/uci_bench.c:73:26: error: format ‘%lu’ expects argument of type ‘long unsigned int’, but argument 2 has type ‘uint64_t {aka long long unsigned int}’ [-Werror=format=]
printf("Nodes/second: %lu\n", (total_nodes * 1000) / bench_time);
^
cc1: all warnings being treated as errors
Makefile:48: recipe for target 'objects/uci/uci_bench.o' failed
make: *** [objects/uci/uci_bench.o] Error 1
make: *** Waiting for unfinished jobs....
Al@PicoChess:~/stash/src $

Thanks,

Al.

Fixed, I made the same mistake I already made on pv display ^^
I also merged last work I did on Null Move Pruning (version is now 12.1, pruning more aggressively when eval is already way above beta), should be around +20 Elo.

[Scally]

Thanks very much,

Compiles and works a treat:

Al@PicoChess:~ $ ./stash121
uci
id name Stash v12.1
id author Morgan Houppin (@mhouppin)
option name Hash type spin default 16 min 1 max 131072
option name Clear Hash type button
option name MultiPV type spin default 1 min 1 max 16
option name Minimum Thinking Time type spin default 20 min 0 max 30000
option name Move Overhead type spin default 20 min 0 max 1000
option name UCI_Chess960 type check default false
uciok
position startpos
go infinite
info depth 1 seldepth 2 multipv 1 nodes 20 nps 20000 hashfull 0 time 1 score cp 40 pv e2e4
info depth 2 seldepth 4 multipv 1 nodes 85 nps 85000 hashfull 0 time 1 score cp 15 pv e2e4
info depth 3 seldepth 8 multipv 1 nodes 604 nps 302000 hashfull 0 time 2 score cp 25 pv e2e4

Al.

[mhouppin]

Version 12.2 is out: with two changes:

- TT entries do not get overwritten on the same position if the current search depth is lower than the TT depth (could occur in case of a partial bound).
- I finally implemented full-pv display (not only the bestmove).

I also completed release binaries for Windows and Linux for version 11 and 12.

Have a good evening !