https://live.lczero.org/

[crem]

We are planning to run a live annotation of the upcoming WCC 24 with Lc0 at https://live.lczero.org/, similar to like https://analysis.sesse.net does it with Stockfish.

Currently, as a test, it's annotating TCEC games.

Please tell if something is broken, missing, or confusing.

It's still being developed, so restarts, database wipeouts etc are expected.

[shawn]

Nice project! I wonder if there's a way to make the board bigger

[chesskobra]

Great news and thanks.

What do different types of arrows mean? Does each colour represent one move and its response (dashed)? Also, why the dashed arrows are curved? Could you make them thinner and straight? I would also like a bigger board and the game score window on the right broader, and the three dots before black's move in the pgn window could be on the same line as black's move.

It would be even nicer if sesse and lczero did a joint project and show the analysis of the two engines together.

[crem]

the three dots before black's move in the pgn window could be on the same line as black's move.

Not sure what you mean. I assume you mean "on the same vertical line as white's move" (and still not sure what you mean).

Three dots should be aligned with the move number, or the white move itself? Should the move notation (not numbers) be aligned for white and black?

[chesskobra]

If moves require more width, I see something like

Code: Select all

6. d5
  ...
Nce7

But there is no way to adjust the column widths so that it would look like

Code: Select all

6. d5
   ... Nc7

[chesskobra]

Another suggestion about the analysis window below the board: the content of the first 3 columns (move, WDL probability, Nodes/Probability of being best) are vertically centred, while the PV can be 1 to 3 rows. This makes the alignment visually uneven. Personally I would be comfortable if the contents of all columns were top, left justified and if there were horizontal light lines separating the analyses of different moves (although I am not sure if others would have a similar preference as mine).

[towforce]

Very good!

I don't know what the prevalence of OCD is in computer chess, but even though I know they're only very rough approximations, I personally would make the win/draw/lose percentages sum to 100:

1. Sum the 3 percentages (S3P)

2. Each percentage becomes: percentage * (100 / S3P)

Then the 3 percentages would always sum to 100 (or very close to it)

[Jouni]

For analysing human games Stockfish 12 and one core in MV=4 mode is more than enough .

[crem]

make the win/draw/lose percentages sum to 100:

Aren't they already add up to 100?

In the code, probabilities for white and black are rounded up to 0.1, and then probability of a draw is 100-white-black.