Developer update: 275% improved game compression

I love technical insight to the website, thanks for the post.

Thanks for the hard work! Lichess is getting even better!

+1 Great work!

great work, thank you for your skill and dedication.

I'm blown away by the programming skills, (while I can code in python, it always feels a bit like wading through treacle for me).

Anyway as you've been thinking about Huffman encoding (the idea of which I've always really liked) --- you might find this interesting.

arxiv.org/abs/1402.4738

(I wrote this paper after having an idea to do with alphabets and huffman encoding --- it passed peer review -- and then they asked for money to publish so I just put it on Arxiv. )

Was expecting middle-out compression but well done!

@residualinsight Having a systematic way to group the symbols sounds really awesome. I'll definitely try it.

@revoof great! .. If you find it interesting I can talk on skype maybe?

That's great devs! Thank you for keeping us updated!

As soon as I can I'll be a patron again, promise.

@revoof Sorry for the probably stupid question, but you use move order to order the moves but then you use a prebuilt Huffman tree from all rated games together. How does that fit together?

EDIT: Nvm I get it now. That does seem like it could be compressed further (at the cost of speed though). Is there an upper bound for how long this should take?

how do you make the graphs and performance tests