So, in a desperate attempt to help me with my handicapping page, I quickly went back to my old try to beat Vegas days, and built the Lunatic simulator.

So, it is what you might think it is.  I quickly overfit models to predict what a team’s FG percentage, TPG percentage, FT percentage, turnover rate, FT rate, TPG / FG ratio, offensive rebounding rate, and estimated number of possessions based on playing their opponent (and vice versa).

Then, I throw those predictions into a simulator that uses random numbers and the predictions to simulate 1000 games between the two teams.

Currently, I am running it through the first round of the tournament to determine if I should use it for my “sane” picks (we all know that I will pick Purdue in my insane picks).  Also, waiting to see what happens in the last play-in game.

The reason I never got comfortable using the simulator in the past is I never was able to figure out how to adjust for a team that plays against weaker teams.  The tool tends to like smaller conferences more than I would like – for example, it says that Arizona State only has a 41% chance of beating Nevada – and my experience with the Mountain West is to not trust them in the tournament.  If the Sun Devils blow out Nevada, I might have to reconsider, but so far, it is doing well.  It would have missed on Mississippi State (who was a 52% favorite, but would have been correct when accounting for the spread (since it said take Pitt and the 2.5 points).  It was also right on everything else so far.

Seems like I could do something fun like take the underdog if it has a 45% chance or better of winning.  Then again, it is also taking a while – guess it takes a while for the computer to simulate 1000 games….

If it doesn’t finish, or looks too chalky, I will stick with what I have already put in.  I figured if I was going to play with creating a model, I should at least have a backup already in place for when the model doesn’t look good at 1:30 in the morning.

Editor note: Arizona State has jumped out to an 11-3 lead.  Maybe I should stick to just using it for judging the over/unders.