Scientific paper on chess openings using lichess data

Dear chess lovers,
Dear lichess developers,
It's my pleasure to inform you that I published in a prestigious journal a study on chess openings based on lichess data.
"Quantifying the complexity and similarity of chess openings using online chess community data."

This is the link to the full article:
www.nature.com/articles/s41598-023-31658-w
where you can download its pdf version or browse it.

This here is the announcement for the general public on our institutional page:
www.csh.ac.at/chess-openings/

While the following link points to the tweet
twitter.com/CSHVienna/status/1643153294662029313?s=20

Feedback is welcome.

Thanks a lot for the opportunity to use your data and for your amazing platform.
Best wishes and good chess to all.
Vts

how horrifying!

First diagram suggests that if you are a c4/d4 White player, you are best - in terms of efficiency of effort - to learn the Dutch Defence as Black! Not something I have seen suggested.

Or have I got this wrong?

Nice, thanks!

@AlexiHarvey said in #3:
> First diagram suggests that if you are a c4/d4 White player, you are best - in terms of efficiency of effort - to learn the Dutch Defence as Black! Not something I have seen suggested.

No, it's not right. That picture is a 2-dimensional projection of a complex network.
It does not mean the Dutch opening is "closer" to c4/d4. The visualizing software put it there. If you look carefully, It does not seem connected to the c4/d4 clusters
The relevant information in the picture is the existence of 10 clusters and the proximity *inside* the same cluster.
The interesting fact is not that we find clusters of openings -- that's obvious -- but that such classification results from players' behavior. It turns out that players over 2000 Glitko2, the ones we have studied over a year, have a systematic way of choosing their opening strategies. Otherwise, if they played randomly, we would find no clusters.