Page 3 of 7

Re: OliThink 5.4.0 has been published with an big leap in strength for only 3 lines of code

Posted: Thu Jun 25, 2020 8:14 pm
by OliverBr
Dann Corbit wrote:
Thu Jun 25, 2020 4:16 pm
OliverBr wrote:
Thu Jun 25, 2020 1:22 pm
Dann Corbit wrote:
Wed Jun 24, 2020 9:44 am
It would be nice if you put your home page link in your sig.
Then old guys like me could remember where to look. :D
Can you please navigate me to the section where I can write my signature?
PS: 5.4.8 is out :)
Step 1:
https://drive.google.com/file/d/1RTZm2h ... sp=sharing
Step 2:
https://drive.google.com/file/d/1yrPjUz ... sp=sharing
Ah, it's "User Control Panel": Thank you!

Re: OliThink 5.4.0 has been published with an big leap in strength for only 3 lines of code

Posted: Fri Jun 26, 2020 7:22 am
by OliverBr
OliThink 5.4.9 is released (http://brausch.org/home/chess).

When self-playing it's a big improvement. I haven't got the time to test against other engines yet.

Re: OliThink 5.4.0 has been published with an big leap in strength for only 3 lines of code

Posted: Sat Jun 27, 2020 6:56 am
by OliverBr
Question to the community:
Some engine resign when falling under a certain score. I couldn't find a command in the chess engine protocol to tell the engine to resign. Protocol adapters like polyglot support it, but what if no adapters is used?

Is a command line option useful and will it be used? Anyway, if an engine falls under -600 against another engine, it's very over very quick.,,

Re: OliThink 5.4.0 has been published with an big leap in strength for only 3 lines of code

Posted: Fri Jul 03, 2020 8:04 am
by OliverBr
OliThink 5.4.13 released.
The frist time I changed something with the very simple Hash(TT)Tables. By this way I doubled the memory usage from 48MByte to 96Mbyte 8-)

Re: OliThink 5.4.0 has been published with an big leap in strength for only 3 lines of code

Posted: Fri Jul 03, 2020 8:18 am
by Dann Corbit
I'm thinking shared hash is next.
I want to see Olithink mop the floor against the big boys with under 100K of code.
No pressure though.
:twisted:

Re: OliThink 5.4.0 has been published with an big leap in strength for only 3 lines of code

Posted: Fri Jul 03, 2020 2:17 pm
by OliverBr
Dann Corbit wrote:
Fri Jul 03, 2020 8:18 am
I'm thinking shared hash is next.
I want to see Olithink mop the floor against the big boys with under 100K of code.
No pressure though.
:twisted:
It's quite difficult to increase strength without expanding code. I omitted a couple of ideas just because they were "too expensive" or "too expansive" :)

Re: OliThink 5.4.0 has been published with an big leap in strength for only 3 lines of code

Posted: Fri Jul 03, 2020 6:19 pm
by tmokonen
It looks like the links for the Java version are out of date. The site says 5.4.11, but the links point to 5.4.6.

Re: OliThink 5.4.0 has been published with an big leap in strength for only 3 lines of code

Posted: Sat Jul 04, 2020 9:35 am
by OliverBr
tmokonen wrote:
Fri Jul 03, 2020 6:19 pm
It looks like the links for the Java version are out of date. The site says 5.4.11, but the links point to 5.4.6.
Thank you for the hint. I already corrected this before.

Re: OliThink 5.4.0 has been published with an big leap in strength for only 3 lines of code

Posted: Sun Jul 05, 2020 8:03 pm
by OliverBr
OliThink 5.5.0 has been released today.

Most notable change is the refactoring of time management and parts of the search function.

Re: OliThink 5.4.0 has been published with an big leap in strength for only 3 lines of code

Posted: Sun Jul 05, 2020 9:31 pm
by OliverBr
OliThink plays a decent midgame, but still the endgame is not good yet. In 100 games against Glaurung 2.1 (17.0 to 83.0) OliThink has not won a single endgame. More than half of the games are open until endgame.
Of the 9 times OliThink won, none of them were endgame wins. Example:



Analysis by Leela (until move 29 both engines played very well):
29...Qe8? - Very defensive, Leela proposes 29...Ra8!
35...Rc8? - Again, Leele prefers 35...Ra8!
38. Rb1? - This time, white blunders (first time after 38 moves), 38. Ra6 !! is more precise.
41. Ra1! - Dagger thrust. White wins.