very interesting project! I'm a chess aficionado and playing vs stockfish does always feel a bit clunky at ~1,700-1,900 ELO. How are the 100-point buckets calibrated? Would the ~1,800 ELO truly win ~50% of the times against humans with a similar rating?
The 100-point buckets are fine-tuned on blitz games from users at that Lichess rating. Lichess ratings tend to be a bit high compared to FIDE/USDF/chess.com ratings. There's a good post at https://chessgoals.com/rating-comparison/ comparing them
very cool idea! Like others have said, "country" doesn't feel like the right level of aggregation. Municipality, street, or even coffee shop itself would feel more natural, although there's a sure trade-off with the effort to source the data
reply