McCain X4 Release Final

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.
Post Reply
MikeB
Posts: 3390
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

McCain X4 Release Final

Post by MikeB » Fri Jun 21, 2019 3:52 pm

Bianries ( all flavors) , source and more details will be found at the link below. I will be updating the details throughout the day, but in summary :
  • McCain can be used with four polyglot opening books, with book 1 , should be a very focused book based on player or opening, book 2 is wider and for books 3 and 4 , I suggest Thomas Zipproth books - both his NN book and his Cerebellum.
  • UCI Options are now displayed in a few of the GUIs where they were not been displayed.
  • It has two options for weaker play - play by Elo, where you select the Elo (check the play By Elo box) , or a predefined leve ( leave the Play by Elo box unchecked. The weakest level (Novice) will see about 10 nodes per second if your are playing a game where you give the engine about 10 seconds per move ,say game 1 min with 9 sec increment. The highest level, (World Champion), will see about 80,000 nodes per second. This will vary by CPU.
  • The Jekyll and Hyde mode is designed to play different - it must be used in conjunction with the variety option , it can only be activated when variety is set to greater than zero. The most extreme setting is to set both to 15. With both set to 15, the World Champion level play perhaps near 1500. that setting is a little too bizarre for me and some moves will have you scratching your head, but try it with something where if you add the two settings together it totals between 15 and 20 ( and perhaps weaker than the World Champion level as well). All weaker setting were designed to play with just one thread. It is also designed to play stronger as you head into the endgame or if you get ahead.
https://github.com/MichaelB7/Stockfish/releases/tag/X4


Also many thanks for the great support I receive from Al Cooper, Dann Corbit , Lucas Monge and John Stanback in providing the binaries.

ChiefPushesWood
Posts: 58
Joined: Thu Nov 08, 2018 5:30 pm
Full name: Chief PushesWood

Re: McCain X4 Release Final

Post by ChiefPushesWood » Sat Jun 22, 2019 3:35 am

Mike,
Is there somewhere I can read about the differences in the binary compiles for windows?

Chief

Andrew
Posts: 119
Joined: Wed Mar 08, 2006 11:51 pm
Location: Australia

Re: McCain X4 Release Final

Post by Andrew » Sat Jun 22, 2019 4:29 am

Thanks for the update Mike!

User avatar
Rebel
Posts: 4663
Joined: Thu Aug 18, 2011 10:04 am

Re: McCain X4 Release Final

Post by Rebel » Sat Jun 22, 2019 7:27 am

MikeB wrote:
Fri Jun 21, 2019 3:52 pm
Bianries ( all flavors) , source and more details will be found at the link below. I will be updating the details throughout the day, but in summary :
  • McCain can be used with four polyglot opening books, with book 1 , should be a very focused book based on player or opening, book 2 is wider and for books 3 and 4 , I suggest Thomas Zipproth books - both his NN book and his Cerebellum.
Nice job.

I pitched the 2 books and the NN version is already close, 5000 games, 49.5%.

Although only 46 unique opening lines statistically is hardly relevant.

Code: Select all

Processing game 5.000
Storing game    46
Double openings 4954

Stored in OUTPUT.PGN

NN - CL [5000] 2472.5 - 2527.5 (49.5%)   all
NN - CL [46]     21.5 -   24.5 (46.7%)   without doubles
Everybody is unique, except me.

MikeB
Posts: 3390
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: McCain X4 Release Final

Post by MikeB » Sat Jun 22, 2019 9:08 pm

ChiefPushesWood wrote:
Sat Jun 22, 2019 3:35 am
Mike,
Is there somewhere I can read about the differences in the binary compiles for windows?

Chief
Functionally, they are all the same. The question for you are with binaries will work on your CPU and as well which is the fastest for your machine. They may not all work and to determine which one is the fastest , open one in a terminal window (DOS prompt) and type bench. Us the one that has the highest nodes per second output. The various labels are ID'ing CPUs , with the bmi2 the most current , then modern , then base and the x32 are designed for 32 machines. The names are the person who submitted them. Usually whichever binary is fastest for you now, will always be the fastest one for you down the road ( future versions).

MikeB
Posts: 3390
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: McCain X4 Release Final

Post by MikeB » Sat Jun 22, 2019 9:13 pm

Rebel wrote:
Sat Jun 22, 2019 7:27 am
MikeB wrote:
Fri Jun 21, 2019 3:52 pm
Bianries ( all flavors) , source and more details will be found at the link below. I will be updating the details throughout the day, but in summary :
  • McCain can be used with four polyglot opening books, with book 1 , should be a very focused book based on player or opening, book 2 is wider and for books 3 and 4 , I suggest Thomas Zipproth books - both his NN book and his Cerebellum.
Nice job.

I pitched the 2 books and the NN version is already close, 5000 games, 49.5%.

Although only 46 unique opening lines statistically is hardly relevant.

Code: Select all

Processing game 5.000
Storing game    46
Double openings 4954

Stored in OUTPUT.PGN

NN - CL [5000] 2472.5 - 2527.5 (49.5%)   all
NN - CL [46]     21.5 -   24.5 (46.7%)   without doubles
Just a word of caution using ZIpproth's books, they are designed to play the best move in every game - which is probably not what you want to do as a user . Hence, the need for additional books to provide great variety, to target a specific line or simply to your keep opponent off balance.

Paloma
Posts: 829
Joined: Thu Dec 25, 2008 8:07 pm

Re: McCain X4 Release Final

Post by Paloma » Sat Jun 22, 2019 9:59 pm

MikeB wrote:
Sat Jun 22, 2019 9:08 pm
snip...
They may not all work and to determine which one is the fastest , open one in a terminal window (DOS prompt) and type bench. Use the one that has the highest nodes per second output. ....
It's often said type bench in console windows and you'll see which binary is the fastest. But i get different times on each run.
1. total time (ms) : 4142
2. 4208
3. 4174
4. 4146
5. 4261
and so on ...
The same happens with other engines as well.
What's the reason for this?

Vinvin
Posts: 4369
Joined: Thu Mar 09, 2006 8:40 am
Full name: Vincent Lejeune

Re: McCain X4 Release Final

Post by Vinvin » Sat Jun 22, 2019 10:23 pm

Paloma wrote:
Sat Jun 22, 2019 9:59 pm
MikeB wrote:
Sat Jun 22, 2019 9:08 pm
snip...
They may not all work and to determine which one is the fastest , open one in a terminal window (DOS prompt) and type bench. Use the one that has the highest nodes per second output. ....
It's often said type bench in console windows and you'll see which binary is the fastest. But i get different times on each run.
1. total time (ms) : 4142
2. 4208
3. 4174
4. 4146
5. 4261
and so on ...
The same happens with other engines as well.
What's the reason for this?
Because the operating system does other things with your CPU in the same time.

MikeB
Posts: 3390
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: McCain X4 Release Final

Post by MikeB » Sat Jun 22, 2019 10:39 pm

Vinvin wrote:
Sat Jun 22, 2019 10:23 pm
Paloma wrote:
Sat Jun 22, 2019 9:59 pm
MikeB wrote:
Sat Jun 22, 2019 9:08 pm
snip...
They may not all work and to determine which one is the fastest , open one in a terminal window (DOS prompt) and type bench. Use the one that has the highest nodes per second output. ....
It's often said type bench in console windows and you'll see which binary is the fastest. But i get different times on each run.
1. total time (ms) : 4142
2. 4208
3. 4174
4. 4146
5. 4261
and so on ...
The same happens with other engines as well.
What's the reason for this?
Because the operating system does other things with your CPU in the same time.
Also many modern CPUs use "thermal throttling" ( the more accurate name for "turbo boost") , the hotter the CPUs get , the slower it goes. You only get the boost one it's turned on hardly being used. Some have setting where you can turn that off - then your nps output will be more stable. On the Sf sites, hey scripts so that you run it 50 times in a row to "see" which is really faster - but keep in mind , total time above is 4.2 seconds - with one that was 4.1 and another was 4.3, but on average 4.2

Post Reply