I like accurate's measurement, although I might push out the 2-score parameter a little further, maybe within 14 with up to 10 minutes left. The thing is it makes a huge difference whether in that scenario the trailing team has the ball or not, but you'd have a hell of a time collecting that data without going into play-by-play info (which you may be doing anyway)...

As far as variables to measure in relation to fatalities, one that I would look at is the intersection of home vs. away and favorite/underdog (by the point spread coming in). My wild-ass guess is that fatalities are more related to celebration than drowning sorrow, and as such I would expect a close victory by a home underdog to show more instances than any close victory by a road team, or by a favored home team.

I would also consider how "big" the game is; maybe you could add the rankings of the two teams coming in, obviously anything which is single digits combined is a highly-anticipated game....

Whether or not alcohol is served at the venue.

Whether or not the game is at a neutral site.

Historical (maybe last 10 year) winning percentage; I'll bet that a hell of lot more people would have had drunken wrecks in Orlando if Central Florida had won that game than (say) after we lost to K-State up there...

Very cool study, keep us posted...