[artK]

Originally Posted by Jeremy Germita

Using this match schedule and teams' most recent OPR's, I've run Monte Carlo simulations for IRI 2015.

I've adjusted the randomization model from +- 10 points to +- 36 points to adjust for IRI rule changes, but other than that the assumptions made are about the same.

This is just for fun. I don't expect these to even remotely line up with actual rankings given the craziness of IRI and its rule changes and the amount of time teams have had to improve.

Would it be possible to run again with the deviation being +- 10 points? I think setting your random term to +- 36 is making your predictions too random, because that's three stacks of randomness per alliance. Setting it to +- 10 points would be better, because that is like saying one team per alliance could miss a stack from their average, which seems about right for IRI level consistency. It would also allow for some comparison to how competitive it is to get the first seed in comparison to your division numbers, which did a decent job at predicting the top seeds (except the predicted second seeded team in Newton was the last pick of the second round).