Page 1 of 1
Xiphos 0.3
Posted: Mon Jun 25, 2018 7:05 am
by mt
I have released Xiphos 0.3:
https://github.com/milostatarevic/xiphos/releases
This version is about 80 elo stronger in self-play than 0.2.2 (measured at tc=40/20s). Most of the strength came from the new TM and improved evaluation.
I would like to thank all the testers, TCEC crew, and everyone interested in this project.
Re: Xiphos 0.3
Posted: Mon Jun 25, 2018 7:10 am
by Graham Banks
Nice progress, Milos.
Re: Xiphos 0.3
Posted: Mon Jun 25, 2018 9:14 am
by Damir
Re: Xiphos 0.3
Posted: Mon Jun 25, 2018 10:00 am
by LocutusOfPenguin
Milos,
thats great! I like your engine (but see below, why i cant include it)
But i hope y *ALL* (engines devs) take a look for
complete uci implementation after some time (needed to make the engine run).
Please see my issue at github. Missing uci features is more a problem as a powerful engine
Surely i can understand its abit boring to implement uci stuff in compare to squeeze out some elos
Jürgen
Re: Xiphos 0.3
Posted: Mon Jun 25, 2018 10:56 am
by Dokterchen
Great news. Thanks Milos.
Re: Xiphos 0.3
Posted: Mon Jun 25, 2018 4:25 pm
by jstanback
Hi Milos, I just downloaded and did a quick depth=20 search with the bmi2 and sse versions and got different node counts. Is this normal?
Thanks, John
Re: Xiphos 0.3
Posted: Mon Jun 25, 2018 5:22 pm
by tomitank
jstanback wrote: ↑Mon Jun 25, 2018 4:25 pm
Hi Milos, I just downloaded and did a quick depth=20 search with the bmi2 and sse versions and got different node counts. Is this normal?
Thanks, John
If you have (initialize) different hash keys, then yes.
Re: Xiphos 0.3
Posted: Mon Jun 25, 2018 6:31 pm
by mt
tomitank wrote: ↑Mon Jun 25, 2018 5:22 pm
jstanback wrote: ↑Mon Jun 25, 2018 4:25 pm
Hi Milos, I just downloaded and did a quick depth=20 search with the bmi2 and sse versions and got different node counts. Is this normal?
Thanks, John
If you have (initialize) different hash keys, then yes.
yes. i'm using the time as a seed for the random generator, so some differences are expected each run.
Re: Xiphos 0.3
Posted: Mon Jun 25, 2018 6:32 pm
by mt
LocutusOfPenguin wrote: ↑Mon Jun 25, 2018 10:00 am
Milos,
thats great! I like your engine (but see below, why i cant include it)
But i hope y *ALL* (engines devs) take a look for
complete uci implementation after some time (needed to make the engine run).
Please see my issue at github. Missing uci features is more a problem as a powerful engine
Surely i can understand its abit boring to implement uci stuff in compare to squeeze out some elos
Jürgen
thank you for the interest, will get there
Re: Xiphos 0.3
Posted: Wed Jun 27, 2018 4:16 am
by Fuddur
Xiphos gain great elo in few months.
I am a big fan of xiphos.
