Yet another game analysis method

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

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Ferdy
Posts: 4111
Joined: Sun Aug 10, 2008 1:15 pm
Location: Philippines

Yet another game analysis method

Post by Ferdy » Thu Mar 28, 2019 2:13 am

This method involves 2 analyzing engines, Lc0 and Sf10. In the opening and early middle phases Lc0 will take charge on the analysis. The rest will be done by Sf.

Code: Select all

Q=9, R=5, B=N=3, P=1
piece_value = 9*(wq+bq) + 5*(wr+br) + 3*(wb+wn+bb+bn)
pawn_value = wp+bp
if piece_value >= 50 and pawn_value >= 12 then use Lc0 othewise Sf.
Analysis features:
1. Light descriptive comments, i.e "better is" etc.
2. Variations are generated by actual game playout up to 5 plies.
3. 2 analysis engines

Pick one of the best games of Kramnik (retiring from chess).

Analysis settings:
1. 10s/pos
2. Threads = 1
3. Hash = 256
4. Intel i7 3.4 Ghz
5. Engine1: Lc0 v0.21.1, backend: blas, WeightsFile: 11258-120x9-se.pb.gz, SmartPruningFactor: 0
6. Engine2: Stockfish 10 64 POPCNT



Impressive Lc0's opening analysis even on this slow backend.

Another game is a bit tactical. A human vs machine.


User avatar
Kirk
Posts: 5630
Joined: Sat Mar 11, 2006 2:44 am

Re: Yet another game analysis method

Post by Kirk » Thu Mar 28, 2019 1:37 pm

Ferdy wrote:
Thu Mar 28, 2019 2:13 am
This method involves 2 analyzing engines, Lc0 and Sf10. In the opening and early middle phases Lc0 will take charge on the analysis. The rest will be done by Sf.

Code: Select all

Q=9, R=5, B=N=3, P=1
piece_value = 9*(wq+bq) + 5*(wr+br) + 3*(wb+wn+bb+bn)
pawn_value = wp+bp
if piece_value >= 50 and pawn_value >= 12 then use Lc0 othewise Sf.
Analysis features:
1. Light descriptive comments, i.e "better is" etc.
2. Variations are generated by actual game playout up to 5 plies.
3. 2 analysis engines

Pick one of the best games of Kramnik (retiring from chess).

Analysis settings:
1. 10s/pos
2. Threads = 1
3. Hash = 256
4. Intel i7 3.4 Ghz
5. Engine1: Lc0 v0.21.1, backend: blas, WeightsFile: 11258-120x9-se.pb.gz, SmartPruningFactor: 0
6. Engine2: Stockfish 10 64 POPCNT



Impressive Lc0's opening analysis even on this slow backend.

Another game is a bit tactical. A human vs machine.

Hi Ferdy!

Is this built an extension of the Nucleus analysis tool or Chess Artist ? :)
“He knew all the tricks, dramatic irony, metaphor, pathos, puns, parody, litotes and... satire. He was vicious”

Ferdy
Posts: 4111
Joined: Sun Aug 10, 2008 1:15 pm
Location: Philippines

Re: Yet another game analysis method

Post by Ferdy » Thu Mar 28, 2019 3:27 pm

Kirk wrote:
Thu Mar 28, 2019 1:37 pm
Hi Ferdy!

Is this built an extension of the Nucleus analysis tool or Chess Artist ? :)
I already have sf and lco combo where sf will handle the ending. It is not in Nucleus and not in Aiquiri either, not released. For chess artist, I might include this feature.

User avatar
Kirk
Posts: 5630
Joined: Sat Mar 11, 2006 2:44 am

Re: Yet another game analysis method

Post by Kirk » Fri Mar 29, 2019 2:52 pm

Ferdy wrote:
Thu Mar 28, 2019 3:27 pm
Kirk wrote:
Thu Mar 28, 2019 1:37 pm
Hi Ferdy!

Is this built an extension of the Nucleus analysis tool or Chess Artist ? :)
I already have sf and lco combo where sf will handle the ending. It is not in Nucleus and not in Aiquiri either, not released. For chess artist, I might include this feature.
Is there a Windows executable for Chess Artist? Thanks!
“He knew all the tricks, dramatic irony, metaphor, pathos, puns, parody, litotes and... satire. He was vicious”

Ferdy
Posts: 4111
Joined: Sun Aug 10, 2008 1:15 pm
Location: Philippines

Re: Yet another game analysis method

Post by Ferdy » Fri Mar 29, 2019 2:59 pm

Kirk wrote:
Fri Mar 29, 2019 2:52 pm
Ferdy wrote:
Thu Mar 28, 2019 3:27 pm
Kirk wrote:
Thu Mar 28, 2019 1:37 pm
Hi Ferdy!

Is this built an extension of the Nucleus analysis tool or Chess Artist ? :)
I already have sf and lco combo where sf will handle the ending. It is not in Nucleus and not in Aiquiri either, not released. For chess artist, I might include this feature.
Is there a Windows executable for Chess Artist? Thanks!
Should be here https://github.com/fsmosca/chess-artist/releases

User avatar
Kirk
Posts: 5630
Joined: Sat Mar 11, 2006 2:44 am

Re: Yet another game analysis method

Post by Kirk » Fri Mar 29, 2019 5:40 pm

Ferdy wrote:
Fri Mar 29, 2019 2:59 pm
Kirk wrote:
Fri Mar 29, 2019 2:52 pm
Ferdy wrote:
Thu Mar 28, 2019 3:27 pm
Kirk wrote:
Thu Mar 28, 2019 1:37 pm
Hi Ferdy!

Is this built an extension of the Nucleus analysis tool or Chess Artist ? :)
I already have sf and lco combo where sf will handle the ending. It is not in Nucleus and not in Aiquiri either, not released. For chess artist, I might include this feature.
Is there a Windows executable for Chess Artist? Thanks!
Should be here https://github.com/fsmosca/chess-artist/releases
Thank you VERY Much! Thank you for sharing your hobby!
“He knew all the tricks, dramatic irony, metaphor, pathos, puns, parody, litotes and... satire. He was vicious”

dkappe
Posts: 260
Joined: Tue Aug 21, 2018 5:52 pm
Full name: Dietrich Kappe

Re: Yet another game analysis method

Post by dkappe » Sat Mar 30, 2019 3:44 pm

Ferdy wrote:
Thu Mar 28, 2019 3:27 pm
Kirk wrote:
Thu Mar 28, 2019 1:37 pm
Hi Ferdy!

Is this built an extension of the Nucleus analysis tool or Chess Artist ? :)
I already have sf and lco combo where sf will handle the ending. It is not in Nucleus and not in Aiquiri either, not released. For chess artist, I might include this feature.
Might be the perfect opportunity to upgrade to python3 and the latest version of python-chess, letting it deal with the uci communication (chess-artist doesn’t know what to make of lc0 at the moment).

Ferdy
Posts: 4111
Joined: Sun Aug 10, 2008 1:15 pm
Location: Philippines

Re: Yet another game analysis method

Post by Ferdy » Sat Mar 30, 2019 5:02 pm

dkappe wrote:
Sat Mar 30, 2019 3:44 pm
Ferdy wrote:
Thu Mar 28, 2019 3:27 pm
Kirk wrote:
Thu Mar 28, 2019 1:37 pm
Hi Ferdy!

Is this built an extension of the Nucleus analysis tool or Chess Artist ? :)
I already have sf and lco combo where sf will handle the ending. It is not in Nucleus and not in Aiquiri either, not released. For chess artist, I might include this feature.
Might be the perfect opportunity to upgrade to python3 and the latest version of python-chess, letting it deal with the uci communication (chess-artist doesn’t know what to make of lc0 at the moment).
Been rewriting CA but at slower pace, using more python-chess.
The released exe file can be used to annotate using Lc0.

Command line:

Code: Select all

chess-artist -infile chusa19.pgn -outfile out_chusa19.pgn -eng C:\engines\lc0-v0.21.1-blas\lc0.exe -engoptions "Threads value 1, SmartPruningFactor value 0" -eval search -movetime 1000 -movestart 10

dkappe
Posts: 260
Joined: Tue Aug 21, 2018 5:52 pm
Full name: Dietrich Kappe

Re: Yet another game analysis method

Post by dkappe » Sat Mar 30, 2019 6:04 pm

But no checkin for Linux users?

Dann Corbit
Posts: 10112
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: Yet another game analysis method

Post by Dann Corbit » Sat Mar 30, 2019 10:56 pm

This kind of thing is really useful to people who want to learn and improve.
An engine that is 5% stronger, but nobody understands the moves? No revolution there.
Something that explains the good and bad moves?
Genius in action.

And all the things you make are useful. I praise you sincerely for your efforts.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

Post Reply