Page 1 of 2

asmFish or McBrain

Posted: Tue Jan 23, 2018 6:42 am
by MikeB
If SF-McBrain happens to win this little match, it will be posted for download within 10 days. If not, it be a while yet.

My apologies for posting here - but with asmfish getting getting nearly 1000 downloads in 48 hours, I figure there would be interest in this.

As an a side, the separate forum for matches was actually started when I was a moderator and the sole reason it was started was because most of the posts were very, very long with lots of games in pgn format. I promise no games or results will be posted here except for a few lines when the final result is in - good or bad.

Engine 1: dev-SF-McBrain-v5.0
Engine 2: latest asmFish - which runs about 20% faster than SF and SF-McB

100 games, 50 positions randomly selected by a script from recent ICCF games among highly rated players. The draw rate will likely be high.

Time Control: 15 minutes with 15 second increment
CPUS: 2
Computer: 12 Core 2010 Mac Pro 3.46 Ghz with PopCnt, 6 games will run concurrently.
Started: Match started: 01/24/18 : 00:00:00 EST
Target completion: 01/25/18 : 00:00:00

Match Software: Bob Hyatt's own self-written match program and script (modified).

Time Updates: 1st Update at 01/24/18 : 00:01:30 EST and every 45 minutes thereafter. There may be zero games completed at the first update.
https://www.dropbox.com/s/qdvoswcwdl2i3 ... 1.txt?dl=1

Match Updates: in addition to Time Updates Updates, there will also be completion updates , based on every 10 games completed. Might be a little bit over every two hours. We will have to see.
https://www.dropbox.com/s/gkwh8cybmcc6b ... a.txt?dl=1

PGN File: Updated whenever there is Time Update or a Match Update
https://www.dropbox.com/s/2bw8xgvgz8phife/all.pgn?dl=1

Why two cores, there is no question that asmFish would win most single core matches vs McBrain, but this version of McBrain has a little patch written by Günther Demetz that does amazing well in SF-McBrain. The patch occasionally sends the second core off on a special assignment under certain conditions.

How well with two cores?

Code: Select all

Mon Jan 22 15:57:30 EST 2018
time control = 300 + 5.00 seconds 
Date: 01/22/18 : 15:57:31
Rank Name                      Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR 
---------------------------------------------------------------------------------------------------------
   1 SF-McB v5.0 64 POPCNT      3103   0.0   16   16   300  152.5  50.8   29   24  247   9.7  82.3  3097 
   2 asmFishX_2018-01-22_popc   3097   5.7   16   16   300  147.5  49.2   24   29  247   8.0  82.3  3103 
---------------------------------------------------------------------------------------------------------

Re: asmFish or McBrain

Posted: Tue Jan 23, 2018 11:07 am
by Kohflote
Hi Michael,

What is the difference between SF-McBrain and BrainFish?

Thank you & regards,
Koh, Kah Huat

Re: asmFish or McBrain

Posted: Tue Jan 23, 2018 12:50 pm
by MikeB
Kohflote wrote:Hi Michael,

What is the difference between SF-McBrain and BrainFish?

Thank you & regards,
Koh, Kah Huat
https://github.com/MichaelB7/Stockfish/releases

Re: asmFish or McBrain

Posted: Tue Jan 23, 2018 2:02 pm
by zullil
MikeB wrote:
My apologies for posting here - but with asmfish getting getting nearly 1000 downloads in 48 hours, I figure there would be interest in this.
Am I correct that the asmfish binaries you're providing are generated from the current assembly code at the standard asmfish GitHub site? All you're doing is running the assembler; you haven't written any assembly code to update asmfish to match the latest development version of Stockfish? It seems that Mohammed is involved with another project, so asmfish hasn't been updated in a while.

Re: asmFish or McBrain

Posted: Tue Jan 23, 2018 8:14 pm
by tpoppins
You are correct on all counts, Louis.

Re: asmFish or McBrain

Posted: Wed Jan 24, 2018 12:09 am
by zullil
zullil wrote:
MikeB wrote:
My apologies for posting here - but with asmfish getting getting nearly 1000 downloads in 48 hours, I figure there would be interest in this.
Am I correct that the asmfish binaries you're providing are generated from the current assembly code at the standard asmfish GitHub site? All you're doing is running the assembler; you haven't written any assembly code to update asmfish to match the latest development version of Stockfish? It seems that Mohammed is involved with another project, so asmfish hasn't been updated in a while.
Was just seeking clarity. Didn't intend to offend anyone.

Re: asmFish or McBrain

Posted: Wed Jan 24, 2018 2:31 am
by MikeB
zullil wrote:
MikeB wrote:
My apologies for posting here - but with asmfish getting getting nearly 1000 downloads in 48 hours, I figure there would be interest in this.
Am I correct that the asmfish binaries you're providing are generated from the current assembly code at the standard asmfish GitHub site? All you're doing is running the assembler; you haven't written any assembly code to update asmfish to match the latest development version of Stockfish? It seems that Mohammed is involved with another project, so asmfish hasn't been updated in a while.
Correct - the last asmfish updates were from about two weeks ago. My github page shows where it was forked from. The ONLY thing I have done was turn the compiler options for UCI Limit Strength, Cerebellum opening book and variety in addition to offering the same options turned in the matefinder version, the default is that some of these options are turned off. I am not an assembly programmer and I have no intentions of becoming one.

Re: asmFish or McBrain

Posted: Wed Jan 24, 2018 2:35 am
by MikeB
amsfish is still king in my book.

Code: Select all

100 rounds and 100 games completed...
Time control: 900 seconds + 15.00 seconds
Target completion: 01/23/18 : 20:39:59
Date: 01/23/18 : 14:49:42
100 game(s) loaded
Rank Name                      Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR 
---------------------------------------------------------------------------------------------------------
   1 asmFishX_2018-01-22_popc   3103   0.0   22   22   100   51.0  51.0    7    5   88   7.0  88.0  3097 
   2 SF-McB v5.0 64 POPCNT      3097   6.7   22   22   100   49.0  49.0    5    7   88   5.0  88.0  3103 
---------------------------------------------------------------------------------------------------------

Time to complete: 14:49:44

Re: asmFish or McBrain

Posted: Wed Jan 24, 2018 5:30 am
by MikeB
MikeB wrote:amsfish is still king in my book.

Code: Select all

100 rounds and 100 games completed...
Time control: 900 seconds + 15.00 seconds
Target completion: 01/23/18 : 20:39:59
Date: 01/23/18 : 14:49:42
100 game(s) loaded
Rank Name                      Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR 
---------------------------------------------------------------------------------------------------------
   1 asmFishX_2018-01-22_popc   3103   0.0   22   22   100   51.0  51.0    7    5   88   7.0  88.0  3097 
   2 SF-McB v5.0 64 POPCNT      3097   6.7   22   22   100   49.0  49.0    5    7   88   5.0  88.0  3103 
---------------------------------------------------------------------------------------------------------

Time to complete: 14:49:44
Kicked off a second run using a normal test suite - same tc, same links , first results in about 1-1/2 hours from this post or about 1AM EST (NY) time , updates on time s/b every 48 minutes - not the 1 hour it was last night. Also after every 10 games. as before. This run will be 126 games. Links will also have last night's run in case you missed them. The pgn file will only have tonight's games. Will take close to 20 hours to complete.

Re: asmFish or McBrain

Posted: Wed Jan 24, 2018 7:57 am
by Dann Corbit
400 games is pretty much a bare minimum for a sensible Elo approximation. Of course, several thousand is a lot better.

I know that this is a big strain on resources, which is why organizations like CCRL and CEGT are such a blessing.