Official Release of Ethereal 11.75, supporting MultiPV

[AndrewGrant]

Hello all,

Its been been about 5 months again since the last release, and I'm back with another update, which is a quite a bit stronger than the previous release. I was forced to take some time away from chess development for quite some time dealing with some medical issues, but I think I'm back now for the time being.

In addition to some speed gains, search improvements, eval improvements, and general code clarity improvements, this is the first released version of Ethereal which supports MultiPV analysis. I was able to run about a hundred thousand games using MultiPV on both single core and multi core matches without having any crashes, timelosses, or other illegal behavior. However, if you notice any issue with MultiPV or have any suggested improvements, please contacted me here on talkchess, on the github page, or using my email, andrew@grantnet.us

I've done a good bit of work on OpenBench since the release of Ethereal 11.50. A number of new authors have adopted their own instances of OpenBench for development, which pleases me greatly, and I'm thankful to all those who have suggested changes to OpenBench to make it more versatile.

Binaries for Windows {PEXT/POPCNT/NOPOPCNT} and Android {ARMV7/ARMV8} can be found here.

Elo estimates are still running, but here are the partial results, and a link to the running test

Code: Select all

ELO   | 33.87 +- 4.68 (95%)
SPRT  | 60.0+0.6s Threads=1 Hash=64MB
Games | N: 5485 W: 990 L: 457 D: 4038

A quick summary of the changes, one patch a time ...

11.51: Speedup moveIsSingular() with our knowledge about the TT move's legality
11.52: Encode Castle moves as KxR. Report UCI currmove and UCI currmovenumber
11.53: Skip some bad captures after skipping quiets during moveIsSingular()
11.54: Do not let a Queen attack through our Rooks/Bishops for Mobility scoring
11.55: Speedup and simplify the logic behind storeTTEntry()
11.56: Full engine Tuning session using an average of a static and dynamic eval
11.57: Do not consider a pawn to be Isolated if it may capture another pawn
11.58: Syzygy probing speedup and simplification
11.59: Evaluate Complexity to adjust eval based on conversion chances
11.60: Add Complexity terms to the Tuner and run a Tuning session
11.61: Allow Tuning terms to be considered Midgame or Endgame only
11.62: Evaluate Kings based on nearest file-wise Pawn
11.63: Run a Tuning session on the King Pawn File Proximity term
11.64: Remove the Passed Pawn contribution to the Complexity calculation
11.65: Evaluate Stacked/Doubled Pawns differently based on ability to unstack
11.66: Add a few options to the Tuner in relation to static vs dynamic eval
11.67: Fix a bug in the logic behind patch 11.65
11.68: Fix a newline printing issue in the PERFT suite between Python 2/3
11.69: Do not ever consider an Isolated Pawn to be Backwards
11.70: Halve the eval bonus (but not malus) for Outposts on the A/H files
11.71: Never consider a King and lone Knight/Bishop to be winning
11.72: Do not perform history updates on illegal or skipped quiet moves
11.73: Apply Futility Move Pruning to moves with good history scores in some cases
11.74: Add support for MultiPV analysis
11.75: Add MultiPV information to the README

Once again, a special thanks to Bojun Guo (noobpwnftw), for providing his machines to OpenBench, which is my Fishtest-like testing platform. His compute gives me the freedom to try ideas and concepts which would otherwise be untestable, or too time consuming to justify.

A big thanks to Alayan (Alayan-stk-2) for his patches. He is responsible for a large number of the elo gaining patches since V11.50, including the addition of complexity terms, the addition of the Pawn King file-wise distance evaluation term, and the correction for KB and KN endgames.

Another thank you to all the members of CCC, TCEC, Talkchess, and elsewhere, that have been supporting me in the last months and year. This has been a very hard year of my life, and I'm grateful for the help and advice and support I have received from so many members of this community.

And as always, thanks to the testers from CCRL, CEGT, FASTGM, Ipmanchess, CCC, TCEC, and anyone else who takes the time to run Ethereal on their machines. I'll be checking all those rating lists obsessively for the next few weeks.

Cheers from the Ethereal team

[Ovyron]

However, if you notice any issue with MultiPV or have any suggested improvements, please contacted me here on talkchess

Well, I don't expect you to implement it, but here's this idea.

Komodo MCTS claims that its MultiPV is the best thing around and would provide better results than Stockfish, but if the engine takes advantage of the knowledge that the user wants to know scores and moves for a certain depth (achieved by mixing the MultiPV and Fixed Depth Play Move features - but a MultiPV Depth option could be added so this is used for analysis), the information could be used to give as much Depth as possible to the most relevant moves and relax the Depth reached in the other ones, so Ethereal could well produce better results than Komodo or Stockfish (as the idea escalates with more MultiPV lines, it'd be easy for it to use MultiPV 8 and beat MultiPV 8 Stockfish because of how poor Stockfish performs with more PV lines added.)

[Graham Banks]

Thanks Andrew. I hope you're well now.

[Damir]

Same here Andrew. I wish you all well, and hope it was nothing serious. As always many thanks for the new Ethereal. Your hard work is always appreciated.

[chysiddh14]

alwz happy to see android version ,thanks
i wish it will be no. 1 strongest

[xr_a_y]

Just starting to use OpenBench for Minic, a must have ! Thanks

[carldaman]

Thanks for sharing your work with us, Andrew, and best of health to you going forward!

[MikeB]

Welcome back and wishing you the best of health!!

[drj4759]

There seems to be a problem with Ethereal 11.75 with cutechess running at chessowl tests. First I used the executables posted at this post but was surprised to see the partial results with Ethereal at the bottom of 23 opponents gauntlet. I stopped it and compiled the source with Ubuntu 18.04 which was used in my previous Ethereal dev tests. When I run it again with cutechess cli, I noticed that it is spewing messages like "White's connection stalls" or "Black's connection stalls" . When I looked at the partial results again, it is at the bottom of the 23 players gauntlet.

I had finished running the gauntlet of Ethereal 11.73 dev version and the result was like in the previous tests with Ethereal at the top 10 ranking. I may post the Ethereal 11.73 at the rating list publication if there is no early fix available for 11.75 release version.

[AndrewGrant]

There seems to be a problem with Ethereal 11.75 with cutechess running at chessowl tests. First I used the executables posted at this post but was surprised to see the partial results with Ethereal at the bottom of 23 opponents gauntlet. I stopped it and compiled the source with Ubuntu 18.04 which was used in my previous Ethereal dev tests. When I run it again with cutechess cli, I noticed that it is spewing messages like "White's connection stalls" or "Black's connection stalls" . When I looked at the partial results again, it is at the bottom of the 23 players gauntlet.

I had finished running the gauntlet of Ethereal 11.73 dev version and the result was like in the previous tests with Ethereal at the top 10 ranking. I may post the Ethereal 11.73 at the rating list publication if there is no early fix available for 11.75 release version.

Can you provide some more information about the setup and the configuration for cutechess-cli? I've had no crashes or losses on OpenBench, and no one else has reported any issues to me as of this moment.