Minic version 3

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

Moderator: Ras

User avatar
xr_a_y
Posts: 1872
Joined: Sat Nov 25, 2017 2:28 pm
Location: France

Re: Minic version 3

Post by xr_a_y »

cutechess can do the exact same thing : send go ponder and just after ponderhit. Seems it is related to very near mate situation

Code: Select all

info string Info  2021-07-06 12:51:44-266: Received command : "go ponder wtime 596 btime 260 winc 100 binc 100"
info string Info  2021-07-06 12:51:44-266: Received command : "ponderhit"
User avatar
CMCanavessi
Posts: 1149
Joined: Thu Dec 28, 2017 4:06 pm
Location: Argentina

Re: Minic version 3

Post by CMCanavessi »

xr_a_y wrote: Tue Jul 06, 2021 1:18 pm cutechess can do the exact same thing : send go ponder and just after ponderhit. Seems it is related to very near mate situation

Code: Select all

info string Info  2021-07-06 12:51:44-266: Received command : "go ponder wtime 596 btime 260 winc 100 binc 100"
info string Info  2021-07-06 12:51:44-266: Received command : "ponderhit"
Maybe it's because when it's the opponent's turn, it's an instant move (as it was also pondering), so it's a very fast sequence. I don't know, just thinking what may be the reason.

Engine A plays a move and starts pondering
Engine B starts thinking
Engine B plays a move and starts pondering, but engine A answers that move with another move instantly, as it guessed the Engine B's move
My chess engine Facón: https://github.com/CMCanavessi/facon
User avatar
xr_a_y
Posts: 1872
Joined: Sat Nov 25, 2017 2:28 pm
Location: France

Re: Minic version 3

Post by xr_a_y »

CMCanavessi wrote: Tue Jul 06, 2021 3:21 pm
xr_a_y wrote: Tue Jul 06, 2021 1:18 pm cutechess can do the exact same thing : send go ponder and just after ponderhit. Seems it is related to very near mate situation

Code: Select all

info string Info  2021-07-06 12:51:44-266: Received command : "go ponder wtime 596 btime 260 winc 100 binc 100"
info string Info  2021-07-06 12:51:44-266: Received command : "ponderhit"
Maybe it's because when it's the opponent's turn, it's an instant move (as it was also pondering), so it's a very fast sequence. I don't know, just thinking what may be the reason.

Engine A plays a move and starts pondering
Engine B starts thinking
Engine B plays a move and starts pondering, but engine A answers that move with another move instantly, as it guessed the Engine B's move
yes indeed : forum3/viewtopic.php?f=7&t=77650
User avatar
xr_a_y
Posts: 1872
Joined: Sat Nov 25, 2017 2:28 pm
Location: France

Re: Minic version 3

Post by xr_a_y »

I'm glad to release today Minic 3.09 with some bug fixes and a new embedded net named "nidicolous nighthawk".

https://github.com/tryingsomestuff/Mini ... s/tag/3.09

This net plays quite good FRC chess and is a little stronger than negligible nystagmus at standard chess.

As always, binaries are here : https://github.com/tryingsomestuff/Mini ... ter/Minic3 (and nets there https://github.com/tryingsomestuff/NNUE-Nets if needed).

Have fun.
User avatar
CMCanavessi
Posts: 1149
Joined: Thu Dec 28, 2017 4:06 pm
Location: Argentina

Re: Minic version 3

Post by CMCanavessi »

Great! As soon as Banksia fixes the chess 960 support I will put it in my tournament
My chess engine Facón: https://github.com/CMCanavessi/facon
User avatar
Rebel
Posts: 7607
Joined: Thu Aug 18, 2011 12:04 pm
Full name: Ed Schröder

Re: Minic version 3

Post by Rebel »

xr_a_y wrote: Tue Aug 03, 2021 1:38 pm I'm glad to release today Minic 3.09 with some bug fixes and a new embedded net named "nidicolous nighthawk".

https://github.com/tryingsomestuff/Mini ... s/tag/3.09

This net plays quite good FRC chess and is a little stronger than negligible nystagmus at standard chess.

As always, binaries are here : https://github.com/tryingsomestuff/Mini ... ter/Minic3 (and nets there https://github.com/tryingsomestuff/NNUE-Nets if needed).

Have fun.
Running 3.09 for the GRL.

1800 games.

http://rebel13.nl/a/grl.htm

Minic made a good start....
90% of coding is debugging, the other 10% is writing bugs.
User avatar
CMCanavessi
Posts: 1149
Joined: Thu Dec 28, 2017 4:06 pm
Location: Argentina

Re: Minic version 3

Post by CMCanavessi »

xr_a_y wrote: Tue Aug 03, 2021 1:38 pm I'm glad to release today Minic 3.09 with some bug fixes and a new embedded net named "nidicolous nighthawk".

https://github.com/tryingsomestuff/Mini ... s/tag/3.09

This net plays quite good FRC chess and is a little stronger than negligible nystagmus at standard chess.

As always, binaries are here : https://github.com/tryingsomestuff/Mini ... ter/Minic3 (and nets there https://github.com/tryingsomestuff/NNUE-Nets if needed).

Have fun.
You might want to check, just in case: http://talkchess.com/forum3/viewtopic.p ... 22#p901322
My chess engine Facón: https://github.com/CMCanavessi/facon
User avatar
Rebel
Posts: 7607
Joined: Thu Aug 18, 2011 12:04 pm
Full name: Ed Schröder

Re: Minic version 3

Post by Rebel »

Minic 3.09 -22

Code: Select all

Gambit Rating List
Running      : Gauntlet Minic 3.09
Time Control : Time control 40/120
Games        : 1800

Results from file gauntlet-minic-309.pgn:

No. Name             Win Draw Loss Unf.  Score Games       %
------------------------------------------------------------
  1 Minic 3.09      +431 =634 -735   *0  748.0  1800   41.6%
  2 Marvin 5.1       +97  =75  -28   *0  134.5   200   67.2%
  3 Stash 31.0       +90  =68  -42   *0  124.0   200   62.0%
  4 Halogen 10       +85  =74  -41   *0  122.0   200   61.0%
  5 Seer 2.0.1       +87  =69  -44   *0  121.5   200   60.8%
  6 Wasp 4.50        +84  =75  -41   *0  121.5   200   60.8%
  7 Combusken 1.4.0  +78  =73  -49   *0  114.5   200   57.2%
  8 Weiss 1.3        +84  =59  -57   *0  113.5   200   56.8%
  9 Clover 2.3.1     +76  =65  -59   *0  108.5   200   54.2%
 10 Orion 0.8        +54  =76  -70   *0   92.0   200   46.0%

Total Games:    1800
White Wins:      580 (32.2%)
Black Wins:      586 (32.6%)
Draws:           634 (35.2%)
Unfinished:        0 (0.0%)

Estimated elo gain for Minic_3.09
Elo pool : 3125
Minic 3.08 : 3093.0
Minic_3.09 : 3070.6
Difference : -22.4
Minic 3.04 is still the best version, I will replay this 1800 game match with the standard evaluation as Minic 3.04x

http://rebel13.nl/a/grl.htm
90% of coding is debugging, the other 10% is writing bugs.
User avatar
xr_a_y
Posts: 1872
Joined: Sat Nov 25, 2017 2:28 pm
Location: France

Re: Minic version 3

Post by xr_a_y »

Rebel wrote: Wed Aug 04, 2021 8:19 am Minic 3.09 -22

Minic 3.04 is still the best version, I will replay this 1800 game match with the standard evaluation as Minic 3.04x

http://rebel13.nl/a/grl.htm
Thanks for the test. I'm not sure to understand your last sentence. Minic 3.04 was tested HCE or with a net ?
User avatar
Rebel
Posts: 7607
Joined: Thu Aug 18, 2011 12:04 pm
Full name: Ed Schröder

Re: Minic version 3

Post by Rebel »

xr_a_y wrote: Wed Aug 04, 2021 9:23 am
Rebel wrote: Wed Aug 04, 2021 8:19 am Minic 3.09 -22

Minic 3.04 is still the best version, I will replay this 1800 game match with the standard evaluation as Minic 3.04x

http://rebel13.nl/a/grl.htm
Thanks for the test. I'm not sure to understand your last sentence. Minic 3.04 was tested HCE or with a net ?
I am pretty sure I tested 3.04 with the right NN, but looking at the results (since 3.04 is still best) it makes me doubt. Hence I replay 3.04 with HCE to be sure. And so far as the current result indicates it looks like 3.04 was tested with NN. It comes nowhere near to the 3110 elo it has on the GRL.
90% of coding is debugging, the other 10% is writing bugs.