Brilliant moves aren't all that brilliant

Sort:
thedelcai

The problem is that a brilliant move should not be easy to find... that's pretty much the definition, but that's not easy for a computer to determine. Initially, chess.com tried selecting moves that the engine didn't like until it reached a certain depth in its analysis, but that didn't work out very well:

from 2021:

Erik: "We are totally redoing all of this, and we have a huge group of brilliant people who are working on analysis classifications and all these other projects… The previous way of delivering on Brilliancies was a little bit about what the engine didn’t expect and found later... as it parsed through the line. Engines and humans don’t think the same. And so, that’s just not a great definition for that. It tends to pull out some really weird stuff. It works well some percentage of the time. It just really doesn’t in another percentage."

fChess.com, 2023:

"As popular as the Brillant move has been, we've heard your feedback that they can be too rare and sometimes not that brilliant. We agree, and we've updated the way that they're calculated. The new Game Review has been fine tuned to better understand what should be identified as a Brilliant move. Brilliant moves must now sacrifice material in some way and must be the best move in a position. Brilliant moves will also be a bit more generously awarded to newer players, recognizing that some strong sacrifices that may be standard for experienced players are quite an achievement for newer players. We think you'll love the update. Try it out and give us your feedback!"