Is it possible to deduce brilliant sacrifices by analyzing a match's PGN along with the corresponding evaluation score after each move? If not, what other non-visual information would be needed to be able to deduce when a player made sacrifices to help his cause?
It would be easy to detect sacrifices- if the material count in the position suddenly drops, but the computer evaluation position doesn't drop a lot, and the material isn't restored within a short number of moves (to avoid simple exchanges being considered sacrifices), then it would be a sound sacrifice. This would miss out on a lot of unsound sacrifices or sacrifices that computers don't understand, however- there are many cases of this with fortresses etc. Detecting brilliant sacrifices would probably be even harder.