Yellow: More common for the piece to be captured there.
Purple: Less common.
Data comes from 15,000 games played on lichess.org.
For kings the graph shows position when game ended.
From desfirsit:
Top two rows are black pieces, bottom rows are white pieces. The orientation of the board is the same in all graphs.
Made in R, using rchess and bigchess packages, visualized with ggplot, with viridis color scale.
Data is from Lichess open data (https://database.lichess.org/). These games were played in January 2013, but reflect so basic patterns that I expect the pattern would be almost identical if newer games were used.
The analysis builds on using the rchess package to extract the sequence of moves from the games, and then I created a script to keep track of all individual pieces as the game progresses.
Edit: The color scales are relative to each piece.
Some pieces are captured more often than others (knights, bishops and the central pawns live the most dangerously), so having a common color scheme would make it hard to see where for instance the pawns on the side tend to be captured.
So one cannot use these graphs to determine which pieces are captured more often.
When asked about how players contributed to the data and their rankings this was his reply:
This is all ratings, but I hadn’t thought about the number of players. I checked now and it is 1665 unique players in these games.
The patterns would probably look a little bit different if one for instance analyzed grandmaster games, but my guess is that many of the patterns are so fundamental that the changes would be very subtle.
Especially regarding pawns that are very constrained in their moveset. But maybe the more free pieces like the bishops that you mentioned would create different patterns!
Regarding the kings, these games vary a lot in length. So if a game is surrendered early the king will probably be in one of the “normal” positions.
What do you think?
Leave a Reply