I believe Regan just measures how badly a given player's opponents tend to err (as determined by computer evaluation of moves). So it's just a post hoc statistical measure, not an actual evaluation of how tricky that player's moves are.
But funnily enough, all you'd need for erikig's program.
A program runs through your git/svn and tracks the user, date, commit, and reversion.
While analysing reversions vs commits and all the in between, I am confident it would be somewhat easy to tweak the analysis to be accurate.
Each reversion you encounter, you check the five next and five previous commits. If any users stand out as being in the previous tree for additional/reversional...
