Programming for style of play

[Michael Sherwin]

I got this idea from something that Ferdinand Mosca posted. Even though Ferdinand was not suggesting this idea my mind got this from it.

This idea may sacrifice some ELO. The trade off could be a better way to control playing style in an engine. It might go something like this. For the first n ply of the search get an accurate score for each move. If there are one or more moves that are close in score then evaluate the moves according to a more humanized for style evaluation and adjust the scores. An example might be as simple as searching to a shallow depth with hash tables turned off and if one of the candidate moves is a sacrifice its score is improved some. Really the processing ideas would be endless here.