lichess-org / compression Goto Github PK

Hello.
I played a game of chess against someone and gave my opponent more time. So that he has 5min time to think about it. For some reason lichess always interrupts the game. The reason for this is "time out". But both players still have time. What could be the reason for that. The game ends immediately if white wants to make the first move.
Forum: https://lichess.org/forum/general-chess-discussion/lichess-time-bug

Tactics/Puzzles on lichess are taken from it's database, so the reply/continuing move of the puzzle is the move played in the puzzle.
However, the issue with this is, even if the first move is the best(according to the computer/stockfish chess engine), the move that opponent takes sometimes isn't the best. So in a way, we are expected to expect that the opponent will move do a blunder and then we would react to it.
Example:

The green lines show the move sequence of the puzzle. We clearly can see that bishopC4(by black) is a blunder and we should not expect the player to do so, however, the puzzle is making up expect it.

There are 2 solutions in my head(relatively difficult but better, and relatively easier but not the best):

1. Computer shall analyse such games and do the tactically best move omitting the blunder made by the player.
2. There could be sub-forum, wherein, people can submit such puzzles and then, the puzzle could be removed or omitted.

Flag of pin and skewer should be not checked if no bishops , no rooks and no Queen are present

Greeting and my apologies for using the issue tracker for just a question but here what I would ask :

I just read Lichess post about this compressor
https://lichess.org/blog/Wqa7GiAAAOIpBLoY/developer-update-275-improved-game-compression

And I find it really brilliant, but then I started to remember courses I took lot of years ago and just curious about how such implementation perform against different chess game sizes.

I am using my memory about Huffman encoding, so probably I am making mistakes, but I remember that can be expensive to store the tree for short messages. And for that reason is not always used as compression algorithm.
For that reason I am curious about data on this problem (compression of chess games) to understand if there is some threshold where simpler compression algorithm are cheaper (eg an adaptation of Zigzag VLQ because we are talking about byte/int).

Another way to cut head of small games, and improving in general compression rate may be also the use of shared opening book. So you can use just a number to pinpoint a sequence in the opening.

So to summarize, my question did you already thought about those two possible improvements? I think probably yes, then if possible can you please share your consideration about splitting the compression phase, because I am really curious about it.