v2.5.3

• Fix a rare edge case where triple repetition might not be detected if the first position in the sequence had a fake en-passant square (where the en-passant square is set but the capture is not actually legal).

v2.5.2

• The space bar will now toggle analysis on and off, except when in Play mode or Analysis Locked mode.

v2.5.0

• Translations for various languages. Mostly machine translations.
• Suggestions and fixes welcome... as a pull request is strongly preferred.
• Also ask for other languages if you like.
• See the file src/modules/translations.js

Currently only menu items are translated. I might expand this in the future to translate other in-app text, if I'm in a good mood, if I have time, if you're lucky, if the stars align, if my cat approves, if the groundhog sees his shadow, barring any unforeseen circumstances, after due consideration, God willing.

v2.4.8

• Added some backends to be up to date with the various Lc0 builds.

v2.4.6

• Add an option for Leela's WDLEvalObjectivity setting.
• Add a visual indicator if the engine is running with contempt - in case you forget.

v2.4.5

• Adds some basic settings in the Engines menu to access Leela's contempt feature.

v2.4.4

• In the event that one of Nibbler's config files contains invalid JSON (due to manual edits from the user) we attempt to fix a couple of common mistakes, then re-parse.
• Missing commas are not fixed, that's too hard.

v2.4.3

• Prevent an issue where Nibbler might temporarily display a very stale PV.
• Thanks to the anonymous reporter of this issue. I'm surprised nobody noticed this before...

v2.4.2

• Nibbler can now accept UCI format moves while parsing (whether from file or clipboard). While these would not be present in normal PGN, this is potentially useful to developers who deal with such UCI format moves.
• The save file dialog has been altered to encourage files to be saved as .pgn files.

v2.4.1

• Improvements to the graph thanks to @Naphthalin:
• The graph now draws two horizontal lines. When using centipawns, these lines correspond to pawn values of -1 and +1, while for winrates they correspond approximately (not exactly) to 25% and 75%. Thus, when the game is close the graph will fall in between those lines.
• The graph is now based on centipawns - apparently this is better, even for Leela.
• Extreme centipawn values (e.g. +20 or +40) will be drawn as basically the same, since they are. This allows more resolution for the critical values where the game is in doubt.