Saturday, 16 May 2020

Do not do this at home

One of the features offered by some of the online chess sites is an analysis of games that you have recently played. For a long time I didn't pay much attention to this feature, but I do now use it to look for interesting games (both for this blog and for my chess streams) . It normally compares the move chosen with what the best move chosen by an engine, and then rates the move based on difference in evaluation. (As an aside, I at least learned the source of annoying questions I used to get asked on Quora about "average centi-pawn loss").
Unfortunately, based on my recent observations, is that it often does not correlate with playing strength. In some games one big mistake can wreck your score (as shown below), while in other cases, it has less of an effect. This is often the case when the game has a forcing nature, where the winning plan is quite obvious, and so both players walk a single path.
In the example game White was given a accuracy score of 10.8%, which I think is a little unfair. Sure he blundered into a mate in 2, but the moves before that weren't that terrible.

mattrad (1604) - beare840 (1517) [D00]
Live Chess, 16.05.2020

No comments: