I have just released Marvin 5.1.0. My estimate is that it is about 40-50 elo stronger than 5.0.0.
From the release notes on GitHub:
* New NNUE net, net-ca6610a.nnue
* Optimized NNUE inference code
* New razoring margins
* New futility margins
* Some SMP bug fixes
* Fixed a bug in the parsing of the setoption UCI command
* Fix calculation of hash usage
The razoring and futility margins were automatically tuned using the tool introduced here, forum3/viewtopic.php?t=71650. Thanks to thomasahle for providing this great tool!
Thanks to everyone testing and showing interest in my engine!
Binaries:https://github.com/bmdanielsson/marvin- ... tag/v5.1.0
Marvin 5.1.0
Moderators: hgm, Rebel, chrisw
-
- Posts: 51
- Joined: Sun Jul 10, 2016 9:12 pm
- Location: Lund, Sweden
- Full name: Martin Danielsson
-
- Posts: 2803
- Joined: Mon Feb 11, 2008 3:53 pm
- Location: Denmark
- Full name: Damir Desevac
Re: Marvin 5.1.0
Thanks a lot Martin for all your hard work. I hope your engine keeps on improving.
-
- Posts: 1530
- Joined: Sat Feb 06, 2021 8:06 am
- Full name: Alex Morales
Re: Marvin 5.1.0
Testing...
Chess engines and dedicated chess computers fan since 1981 Mac mini M1 8GB-256GB, Windows 11 & Ubuntu ARM64.
ProteusSF Dev Forum TROLLS KINDERGARTEN
ProteusSF Dev Forum TROLLS KINDERGARTEN
-
- Posts: 7017
- Joined: Thu Aug 18, 2011 12:04 pm
- Full name: Ed Schröder
Re: Marvin 5.1.0
Code: Select all
Gambit Rating List
Running : Gauntlet Marvin 5.1.0
Time Control : Time control 40/120
Games : 1800
Results from file gauntlet-marvin.pgn:
No. Name Win Draw Loss Unf. Score Games %
------------------------------------------------------------
1 Marvin 5.1 +712 =668 -420 *0 1046.0 1800 58.1%
2 Stash 31.0 +70 =76 -54 *0 108.0 200 54.0%
3 Combusken 1.4.0 +60 =79 -61 *0 99.5 200 49.8%
4 Berserk 4.2.0 +52 =85 -63 *0 94.5 200 47.2%
5 Seer 2.0.1 +55 =77 -68 *0 93.5 200 46.8%
6 Minic 3.08 +48 =81 -71 *0 88.5 200 44.2%
7 Clover 2.3.1 +49 =74 -77 *0 86.0 200 43.0%
8 Topple 0.8.0 +36 =57 -107 *0 64.5 200 32.2%
9 Counter 3.8 +24 =78 -98 *0 63.0 200 31.5%
10 Mr Bob_1.0.0 +26 =61 -113 *0 56.5 200 28.2%
Total Games: 1800
White Wins: 584 (32.4%)
Black Wins: 548 (30.4%)
Draws: 668 (37.1%)
Unfinished: 0 (0.0%)
Estimated elo gain for Marvin_5.1
Elo pool : 3095
Marvin 5.0 : 3071.0
Marvin_5.1 : 3148.5
Difference : 77.5
90% of coding is debugging, the other 10% is writing bugs.
-
- Posts: 1142
- Joined: Thu Dec 28, 2017 4:06 pm
- Location: Argentina
Re: Marvin 5.1.0
Great! Do you have plans for FRC?Martin wrote: ↑Sat Jul 10, 2021 10:00 am I have just released Marvin 5.1.0. My estimate is that it is about 40-50 elo stronger than 5.0.0.
From the release notes on GitHub:
* New NNUE net, net-ca6610a.nnue
* Optimized NNUE inference code
* New razoring margins
* New futility margins
* Some SMP bug fixes
* Fixed a bug in the parsing of the setoption UCI command
* Fix calculation of hash usage
The razoring and futility margins were automatically tuned using the tool introduced here, forum3/viewtopic.php?t=71650. Thanks to thomasahle for providing this great tool!
Thanks to everyone testing and showing interest in my engine!
Binaries:https://github.com/bmdanielsson/marvin- ... tag/v5.1.0
Follow my tournament and some Leela gauntlets live at http://twitch.tv/ccls
-
- Posts: 51
- Joined: Sun Jul 10, 2016 9:12 pm
- Location: Lund, Sweden
- Full name: Martin Danielsson
Re: Marvin 5.1.0
FRC support is on the todo list, but I'm not sure when I will get to it.CMCanavessi wrote: ↑Sun Jul 11, 2021 1:44 amGreat! Do you have plans for FRC?Martin wrote: ↑Sat Jul 10, 2021 10:00 am I have just released Marvin 5.1.0. My estimate is that it is about 40-50 elo stronger than 5.0.0.
From the release notes on GitHub:
* New NNUE net, net-ca6610a.nnue
* Optimized NNUE inference code
* New razoring margins
* New futility margins
* Some SMP bug fixes
* Fixed a bug in the parsing of the setoption UCI command
* Fix calculation of hash usage
The razoring and futility margins were automatically tuned using the tool introduced here, http://talkchess.com/forum3/viewtopic.php?t=71650. Thanks to thomasahle for providing this great tool!
Thanks to everyone testing and showing interest in my engine!
Binaries:https://github.com/bmdanielsson/marvin- ... tag/v5.1.0
-
- Posts: 51
- Joined: Sun Jul 10, 2016 9:12 pm
- Location: Lund, Sweden
- Full name: Martin Danielsson
Re: Marvin 5.1.0
Thank you for testing!Rebel wrote: ↑Sat Jul 10, 2021 9:30 pmMarvin 5.1.0 +77 elo.Code: Select all
Gambit Rating List Running : Gauntlet Marvin 5.1.0 Time Control : Time control 40/120 Games : 1800 Results from file gauntlet-marvin.pgn: No. Name Win Draw Loss Unf. Score Games % ------------------------------------------------------------ 1 Marvin 5.1 +712 =668 -420 *0 1046.0 1800 58.1% 2 Stash 31.0 +70 =76 -54 *0 108.0 200 54.0% 3 Combusken 1.4.0 +60 =79 -61 *0 99.5 200 49.8% 4 Berserk 4.2.0 +52 =85 -63 *0 94.5 200 47.2% 5 Seer 2.0.1 +55 =77 -68 *0 93.5 200 46.8% 6 Minic 3.08 +48 =81 -71 *0 88.5 200 44.2% 7 Clover 2.3.1 +49 =74 -77 *0 86.0 200 43.0% 8 Topple 0.8.0 +36 =57 -107 *0 64.5 200 32.2% 9 Counter 3.8 +24 =78 -98 *0 63.0 200 31.5% 10 Mr Bob_1.0.0 +26 =61 -113 *0 56.5 200 28.2% Total Games: 1800 White Wins: 584 (32.4%) Black Wins: 548 (30.4%) Draws: 668 (37.1%) Unfinished: 0 (0.0%) Estimated elo gain for Marvin_5.1 Elo pool : 3095 Marvin 5.0 : 3071.0 Marvin_5.1 : 3148.5 Difference : 77.5
-
- Posts: 4471
- Joined: Fri Apr 21, 2006 4:19 pm
- Location: IASI - the historical capital of MOLDOVA
- Full name: SilvianR
-
- Posts: 4471
- Joined: Fri Apr 21, 2006 4:19 pm
- Location: IASI - the historical capital of MOLDOVA
- Full name: SilvianR