We have evaluated CHA over the entire
Lichess database of standard rated games,
which includes 4,004,175,546 games at the moment.
More concretely, we have applied CHA to the final position of all games that ended in a
timeout and that were classified as 1-0 or 0-1.
This represents a total of 1,272,466,944 games (about 32% of all games) which have been analyzed in
about 121 hours of CPU time (344 μs per position on average).
Our analysis led to identifying a total of 105,751 games that were unfairly classified.
Namely, games that were lost by the player who ran out of time, but their opponent could
not have checkmated them by any possible sequence of legal moves.
We present here all those positions, together with some descriptive measures (you can find
information about each measure by hovering the mouse over it) that can be used to
identify the most interesting ones among the vast and redundant
majority of positions.
Note that our analysis is definite, i.e., the catalog provided here includes
all the games that were unfairly classified.
In all other games, there exists a checkmate sequence for the player who did not
run out of time.
We conclude that CHA is ready for practical use. Chess servers could leverage our tool to
accurately classify games after a timeout, following
the FIDE Laws of Chess.