I simulated Thursday’s sweet 16 games three separate times through 10,000 iterations using each rating system (avg line, kenpom, sagarin) and here are the ML results compared to Pinnacle

Do the opposite of what the ratings say, Pinnacle knows more than gambling degenerates.

The winner of these bracket contests usually, at least in retrospect, finds a solid equilibrium between maximizing variance and taking higher seeds.  Typically a stable strategy involves, to a first approximation, picking chalk all the way through, and then working backwards.  Maximizing variance should be constrained towards the occasional first round upset and regional matchups.  The teams that make the final four are good teams, and in a season where there is no clear cut favorite, any team seeded 1 through 5 (and Washington obviously), would be given odds in the range of +4 to -4 against each other.

Below is a table of a simulation using only efficiency numbers, extracting the power 16 + Utah St.

About a month ago I posted a rating system. With that as the base factor I threw the teams into a simulation. One minor adjustment however. I had attempted to present a system of analyzing teams through epistolary manifestations of brain matter. Meaning there was a thought of how teams should be appropriated using efficiency ratings and the average spread with an additional ratings adjustment that was a response to against the spread expectations (my brain matter). Efficiency ratings and average spread having more weight, I envisioned any point on a Cartesian plain of the two variables as representing an actual circle, having a circumference, and ultimately diameter, contingent on against the spread expectations.  An asymmetry in units measured forced me to seek arbitrary methods.  The logic was, a team has a measure of performance as a function of their given metrics, and then has a potential projected performance using woulda-coulda-shoulda philosophy. Essentially I was just having fun with math and geometry.

However, I scratched most of that, and added instead as a third dimension, sagarin ratings. I weighted each variable equally (spread, efficiency, sagarin), and ran a simulation threw 10,000 iterations. The top 16 are posted below, sorted by simulated odds to win championship in descending order. Pinnacle odds are for comparison.

(I removed around 5% on average from each of BYU’s ratings to pacify the loss of Brandon Davies and to anticipate future performance.)

Teams highlighted have an expected value greater than zero.  Glaring omissions are UNC and UCONN.  For UNC its perfectly understandable, since the insertion of Marshall at PG they have morphed into a legit contender.  Having seen this team perform at a subpar level on many occasions post-Lawson and pre-Marshall, the change in performance is not merely historical contingency.  Larry Drew was a liability.  I can acutely remember the ACC quarter final in 2008 ini the Georgia Dome against VaTech, when Lawson sat out a game.  The same team that absolutely destroyed teams in the tournament with Lawson running the court, needed a bad call and a last second shot by Hansbrough to defeat the bubble-trodden Hokies.  Much of the variance in UNC’s enigmatic season has been resolved by the absence of Drew and the presence of Marshall.  Pinnacle’s opinion is substantiated though unfortunately for the TarHeels, they are in the same region as Washington, who may have the best value (despite he ttimezone inconvenience), along with Utah State, of all the teams in the tournament.

UCONN was serviceable to mediocre for the better part of the year, and had a great run in the Big East Tournament.  Good for them.  I see absolutely zero qualitative measure to assign them any amount of value.  The SIM has them at 0.46% future probability compared to Pinnacle’s 4.15%, an uncharacteristic overreaction.

I’ll post simulation results of a pure efficiency simulation here shortly.