Obsidian (DEV)

Discussion of anything and everything relating to chess playing software and machines.

Moderator: Ras

Jouni
Posts: 3546
Joined: Wed Mar 08, 2006 8:15 pm
Full name: Jouni Uski

Re: Obsidian (DEV)

Post by Jouni »

In my tests 14.11 has been the strongest:

Code: Select all

Score of Obsidian14.23 vs Obsidian1411: 8 - 19 - 173 [0.472]
...      Obsidian14.23 playing White: 6 - 3 - 91  [0.515] 100
...      Obsidian14.23 playing Black: 2 - 16 - 82  [0.430] 100
...      White vs Black: 22 - 5 - 173  [0.542] 200
Elo difference: -19.1 +/- 17.5, LOS: 1.7 %, DrawRatio: 86.5 %
200 of 200 games finished.
60 + 0,6 HERT book
Jouni
gabe_obsidian
Posts: 13
Joined: Fri Jan 26, 2024 10:21 pm
Full name: Gabriele Lombardo

Re: Obsidian (DEV)

Post by gabe_obsidian »

Jouni wrote: Mon Dec 30, 2024 9:56 am In my tests 14.11 has been the strongest:

Code: Select all

Score of Obsidian14.23 vs Obsidian1411: 8 - 19 - 173 [0.472]
...      Obsidian14.23 playing White: 6 - 3 - 91  [0.515] 100
...      Obsidian14.23 playing Black: 2 - 16 - 82  [0.430] 100
...      White vs Black: 22 - 5 - 173  [0.542] 200
Elo difference: -19.1 +/- 17.5, LOS: 1.7 %, DrawRatio: 86.5 %
200 of 200 games finished.
60 + 0,6 HERT book
You're gonna get bogus results if you use a book with 86% drawrate, and on top of that play only 200 games.
14.23 is stronger than 14.11 because I only ever commit patches that gain Elo
AndrewGrant
Posts: 1938
Joined: Tue Apr 19, 2016 6:08 am
Location: U.S.A
Full name: Andrew Grant

Re: Obsidian (DEV)

Post by AndrewGrant »

gabe_obsidian wrote: Mon Dec 30, 2024 12:32 pm
Jouni wrote: Mon Dec 30, 2024 9:56 am In my tests 14.11 has been the strongest:

Code: Select all

Score of Obsidian14.23 vs Obsidian1411: 8 - 19 - 173 [0.472]
...      Obsidian14.23 playing White: 6 - 3 - 91  [0.515] 100
...      Obsidian14.23 playing Black: 2 - 16 - 82  [0.430] 100
...      White vs Black: 22 - 5 - 173  [0.542] 200
Elo difference: -19.1 +/- 17.5, LOS: 1.7 %, DrawRatio: 86.5 %
200 of 200 games finished.
60 + 0,6 HERT book
You're gonna get bogus results if you use a book with 86% drawrate, and on top of that play only 200 games.
14.23 is stronger than 14.11 because I only ever commit patches that gain Elo
I would bet $100,000 with anyone who will take it, that Obsidian 14.23 is stronger that Obsidian 14.11.
We can agree to some conditions, put the money in Escrow, and then have a 3rd party run the games.
jefk
Posts: 888
Joined: Sun Jul 25, 2010 10:07 pm
Location: the Netherlands
Full name: Jef Kaan

Re: Obsidian (DEV)

Post by jefk »

Obsidian 14.23 is stronger that Obsidian 14.11.
most likely, but it's still stalling a bit around ply 30 :cry: (with my 6 threads
xeon comp). Maybe when searching deeper, it should become more
selective (compare eg. with SF17, although that may be a bit too selective
imho, depending on the type of positions of course); although it may
be partly 'cosmetics' it's important in a real game because of time controls, :idea:
cutting off the search when there's not enough time left; if you see
what i mean (no need for some SF programmer'ss comments , tough :wink:
Jouni
Posts: 3546
Joined: Wed Mar 08, 2006 8:15 pm
Full name: Jouni Uski

Re: Obsidian (DEV)

Post by Jouni »

I dare not to bet anything. This is with UHO book:

Code: Select all

Score of Obsidian14.23 vs Obsidian1411: 40 - 48 - 112 [0.480]
...      Obsidian14.23 playing White: 39 - 0 - 61  [0.695] 100
...      Obsidian14.23 playing Black: 1 - 48 - 51  [0.265] 100
...      White vs Black: 87 - 1 - 112  [0.715] 200
Elo difference: -13.9 +/- 32.0, LOS: 19.7 %, DrawRatio: 56.0 %
200 of 200 games finished.
Jouni
gabe_obsidian
Posts: 13
Joined: Fri Jan 26, 2024 10:21 pm
Full name: Gabriele Lombardo

Re: Obsidian (DEV)

Post by gabe_obsidian »

Jouni wrote: Mon Dec 30, 2024 5:59 pm I dare not to bet anything. This is with UHO book:

Code: Select all

Score of Obsidian14.23 vs Obsidian1411: 40 - 48 - 112 [0.480]
...      Obsidian14.23 playing White: 39 - 0 - 61  [0.695] 100
...      Obsidian14.23 playing Black: 1 - 48 - 51  [0.265] 100
...      White vs Black: 87 - 1 - 112  [0.715] 200
Elo difference: -13.9 +/- 32.0, LOS: 19.7 %, DrawRatio: 56.0 %
200 of 200 games finished.
Now with the UHO book you're on the right track, but as should be evident -13.9 +/- 32.0 8 is not even close to meaningful
jefk
Posts: 888
Joined: Sun Jul 25, 2010 10:07 pm
Location: the Netherlands
Full name: Jef Kaan

Re: Obsidian (DEV)

Post by jefk »

a few days ago i wrote :
it's still stalling a bit around ply 30 :cry: (with my 6 threads
xeon comp)
this is in the Chessparter6.1 interface (*), but i now see it works ok
in the Fritz GUI (going on beyond 30, eg 34 ply, etc, up to eg 40 maybe
not as fast as SF which arrives quite fast at '50 ply'(?) but who cares).
(*) maybe it's an uci thing, dunno, for me problem solved.
Jouni
Posts: 3546
Joined: Wed Mar 08, 2006 8:15 pm
Full name: Jouni Uski

Re: Obsidian (DEV)

Post by Jouni »

I don't have recources to a lot games. But Ipman has :)

Code: Select all

  5 Obsidian 14.22 avx512          : 3581    3   3 15000    61.7 %   3498   72.6 %
  6 Obsidian 14.0 avx512-pext      : 3575    3   3 15000    60.8 %   3498   73.4 %
10 + 1 and Noomen book
Jouni
User avatar
Rebel
Posts: 7280
Joined: Thu Aug 18, 2011 12:04 pm
Full name: Ed Schröder

Re: Obsidian (DEV)

Post by Rebel »

https://rebel7775.wixsite.com/rebel/mrl

14.23 gets +23 more than 14.0
90% of coding is debugging, the other 10% is writing bugs.
Jouni
Posts: 3546
Joined: Wed Mar 08, 2006 8:15 pm
Full name: Jouni Uski

Re: Obsidian (DEV)

Post by Jouni »

No more updates 2025?
Jouni