Owners of Ethereal 13.00 & 13.25 & 13.50 should have received emails by now containing a link to download the updated binaries. If you fall into that category, and have not received an email from andrewgrantethereal@gmail.com, please first check your spam folder and then reach out to me via andrewgrantethereal@gmail.com or andrew@grantnet.us. I would prefer to provide any support needed via email, and not here on talkchess.
Changes since the 13.50 release:
Code: Select all
13.75: Release Ethereal 13.75, update the README
13.74: Update the default Fischer Network using latest best practices
13.73: Update the standard Network, removing the King-Distance factorizer
13.72: Implement double extensions for singular extensions
13.71: Update the standard Network, removing the King-Distance factorizer
13.70: Rewrite the time management system to make use of Node counts
13.69: Replace the evaluation cache by storing into the TT early in QS
13.68: Include information left out in the previous patch 13.67
13.67: Merge Follow-up move pruning and Counter move pruning
13.66: Update types.h to include information from 13.65
13.65: Use threat-to and threat-from in capture history
13.64: Use threat-to and threat-from in quiet butterfly history
13.63: return value and not simply beta during NMP excluding mates
13.62: Rewrite the history function to follow an SF-like quadratic
13.61: Make use of negative extensions similar to multi-cut
13.60: Place the MovePicker objects on the Thread's NodeState stack
13.59: Simplify away the singular picker
13.58: Use the tt-move sometimes during ProbCut, and sometimes avoid via it
13.57: Update default network using a new, Stage 17 set of data
13.56: Store the static eval into the TT earlier in search
13.55: Update some of the NNUE processing tools to include ply information
13.54: Fix some typos in the Makefile for use when building for release
13.53: Update the default neural network using a smaller L1 and bigger Input x3
13.52: Update the default neural network using a smaller L1 and bigger Input x2
13.51: Update the default neural network using a smaller L1 and bigger Input x1Code: Select all
Standard Chess using a very drawish book:
ELO | 46.67 +- 1.59 (95%)
CONF | 60.0+0.60s Threads=1 Hash=64MB
GAMES | N: 40008 W: 7221 L: 1879 D: 30908
http://chess.grantnet.us/test/26533/
Fischer Random Chess using a very shallow book:
ELO | 49.49 +- 2.74 (95%)
CONF | 60.0+0.60s Threads=1 Hash=64MB
GAMES | N: 20000 W: 4731 L: 1901 D: 13368
http://chess.grantnet.us/test/26581/My email sent out to existing customers included the following blurb, reminding people of their rights under the GNU GPLv3. I think this is very important, even if it does not impact most people in a meaningful way:
Ethereal, as well as the projects that support Ethereal like OpenBench and NNTrainer are licensed under the GPLv3. The GPLv3 gives you, the user, the right to have access to the source code of the engine, the right to redistribute the GPLv3'ed portions of the project, as well as the right to reuse the Ethereal source in any capacity so long as you continue to comply with the GPLv3's license.
Open Source chess engines have accelerated the development of computer chess in immeasurable ways. If not for the early adopters of the Open Source methods, computer chess would not be what it is today. Powerful programs like Stockfish simply would not exist in their current forms. All of this is possible because the authors have empowered users by granting them rights to the code, only asking that you carry on propagating the licenses attached to their code. This is a small ask, for such a great gift, and yet we live in a time where that gift is not appreciated by some, and worse taken advantage of.
Ethereal shares in the collective knowledge generated and maintained by the Computer Chess Community. However, there are three elements of Ethereal for which explicit attribution is necessary for a good faith effort at carrying out the GPLv3. This attribution has been given when the code was committed, but is restated here for clarity: Ethereal makes use of the universally adopted Syzygy Tablebases, a project under the GPLv2 and other compatible licenses. Ethereal makes use of a forked version of Fathom, a project under the MIT license, used to implement Syzygy. Lastly, Ethereal shares a chunk of code for dealing with the Windows Operating System, which was originally written by Texel author Peter Österlund, and has since been refined and improved in various Stockfish forks, once again under GPLv3 compatible licenses.
