[You must be registered and logged in to see this link.]Location: Pen Argyl, Pennsylvania
[You must be registered and logged in to see this link.] Posts: [You must be registered and logged in to see this link.]Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania
[You must be registered and logged in to see this link.] by
[You must be registered and logged in to see this link.] » Wed Jun 12, 2019 3:39 am
Only for those that can compile their own. Looking for feedback on usability with Chessbase, Areana, XBoard and others Chess GUIs.
source:
[You must be registered and logged in to see this link.]Changes since X3:
[You must be registered and logged in to see this link.] ... lB7:McCain
Specific item to test:
Now accepts 4 polyglot books
Book1 is primary and should very specific to an opening ( or player)
Book2 is wider than book 1, with the secondary lines of book 2 or expanded to more players
Book3 is wider than book 2 , I recommend Thomas Zipproth NN book
Book4 is the big book, Cerebellum or something else that covers a wide variety of lines.
Play By Elo (Engine_ELo) - check the box and pick your ELo. Rating of 1200 should be about 7 nodes per second if you a TC that provides for about 10 seconds per move ( 1 min with 9 second increment) , Goes up to about 2850 which is the vicinity of 150K nps for McCain
Engine Level - select a predefined level and make sure Play by Elo is not checked, no other box to check.
Also, the Lc0 scoring percentage is now the default, but can be picked switched back to centipawn by the user. 0 means loss, 100 means win and 50 is drawish
From Xboard:
Attachments
[You must be registered and logged in to see this link.] by
[You must be registered and logged in to see this link.] » Thu Jun 13, 2019 12:05 am
I was able to get some Windows and macOS binaries for those I who like to kick the ties and provide input.
[You must be registered and logged in to see this link.] - Quote :
- [You must be registered and logged in to see this link.]MikeB wrote: [You must be registered and logged in to see this link.]↑
Thu Jun 13, 2019 12:05 am
I was able to get some Windows and macOS binaries for those I who like to kick the ties and provide input.
[You must be registered and logged in to see this link.]
Mike
I tried the Lucas compile. The options look ok in CB15 but the hash shows 2764 even though I set it to 512 outside CB.
This is the uci file I use.
Code:
[You must be registered and logged in to see this link.] - Code:
-
[ENGINE]
Name=McCain X4
Author=M. Byrne, Stockfish authors and others
Filename=c:\mccain\McCain-X4_x64_bmi2.exe
[OPTIONS]
UCI_Chess960=true
W_Contempt=0
B_Contempt=0
Best Move 3=false
Dynamic_Contempt=false
Threads=1
Hash=512
SyzygyPath=SyzygyPath=c:\tablebases\syzygy
Setting the hash to 512 in engine properties while running McCain in CB seems to work. BTW the hash is not even shown in the screenshot from CPW.
Is this a bug?
Thanks for all your great work.
[You must be registered and logged in to see this link.] Posts: [You must be registered and logged in to see this link.]Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania
[You must be registered and logged in to see this link.] by
[You must be registered and logged in to see this link.] » Thu Jun 13, 2019 1:53 am
- Quote :
- [You must be registered and logged in to see this link.]MikeB wrote: [You must be registered and logged in to see this link.]↑
Wed Jun 12, 2019 4:58 am
I know a lot of folks appreciate the added features, but some folks want to know how does it do against SF. Most of the time, I don't know, since changes are almost always tested again the previous version of McCain.
So fo those that want to see a 500 game match with a time of 5 min with 3 second increment , enjoy!
Results:
[You must be registered and logged in to see this link.] ... a.txt?dl=1
Games:
[You must be registered and logged in to see this link.] ... s.pgn?dl=1
Early results:
[You must be registered and logged in to see this link.] ... s.pgn?dl=1
Updates will be at the completion of 20 games or every 789 seconds ( calculated by the script based on time control how often the planned showings are , i.e, every xx games ) I still use Bob Hyatt's testing code he used for Crafty , of course I have modified it over the years and it's fully integrated with his "watcher" script, using env variables - so that the watcher script know the time control and the planned output. The watcher script then calculates a time to display intermediate results that makes sense for the given time control and planned output.
[big snip]
final result and this is not a fluke:
Code:
[You must be registered and logged in to see this link.] - Code:
-
500 rounds and 500 games completed...
Time control: 300 seconds + 3.0000 seconds
Target completion: 06/12/19 : 13:37:04
Date: 06/12/19 : 13:08:54
500 game(s) loaded
Rank Name Rating Δ + - # Σ Σ% W L D W% =% OppR
---------------------------------------------------------------------------------------------------------
1 Stockfish 090619 64 POPC 3102 0.0 15 15 500 253.5 50.7 68 61 371 13.6 74.2 3098
2 McCain X4 3098 4.4 15 15 500 246.5 49.3 61 68 371 12.2 74.2 3102
---------------------------------------------------------------------------------------------------------
RR-061219-1308.pgn
It's a fluke when McCain or Honey finishes higher that current-dev-SF...
this probably close to real as well, maybe a 10 Elo improvement over X3 - thanks to the SF team:
Code:
[You must be registered and logged in to see this link.] - Code:
-
500 rounds and 500 games completed...
Time control: 60 seconds + .6000 seconds
Target completion: 06/12/19 : 15:54:18
Date: 06/12/19 : 15:22:31
500 game(s) loaded
Rank Name Rating Δ + - # Σ Σ% W L D W% =% OppR
---------------------------------------------------------------------------------------------------------
1 McCain X4 3105 0.0 17 17 500 257.5 51.5 86 71 343 17.2 68.6 3095
2 McCain X3 3095 10.9 17 17 500 242.5 48.5 71 86 343 14.2 68.6 3105
---------------------------------------------------------------------------------------------------------
RR-061219-1522.pgn
I figured out how to compile the engine... Here is a screenshot in Chessbase... Everything is working just fine including the drop downs, etc. Obviously you can see one of the words is truncated (Analysis Contempt) but I don't really care about that.
Windows and macOS binaries for those I who like to kick the ties and provide input.
[You must be registered and logged in to see this link.]
[You must be registered and logged in to see this link.]
