19 March 2015

Rating Systems Challenge: 2015 Systems Overview


This post is part of a series tracking the successes and failures of various NCAA Men's Basketball ranking systems and bracket models throughout the 2015 NCAA Tournament. Check out the full series.

March Madness is back and that means so is the Rational Pastime Rating Systems Challenge. This year I'm pitting twenty-two separate rating systems head to head in an ESPN Tournament Challenge pool. In this, the fifth year of the Challenge, we'll see which systems picked the right upsets, and which emerge victorious from the field. Will any successfully outperform a pure chalk bracket? Over the past four years, only Nate Silver's FiveThirtyEight model has successfully done so.

There are some changes to the pool, this year. John Ezekowitz hasn't compiled a Survival Bracket for 2015. In its place, we've added three new systems. Christopher Long's system makes its debut in the pool, as well as the Massey Ratings. And, in addition to Sports Reference's simple rating system, we've added their adjusted rating system, which accounts for strength of schedule.


As expected, Kentucky (1) is dominating the rankings. Only three systems pick against Kentucky this March: Lunardi RPI picks Duke (1) and Sonny Moore and the ESPN Decision Tree both pick Arizona (2) to win it all. No system has any school save Kentucky emerging from the Midwest region.


Various systems also expect Villanova (1), Wisconsin (1) and Virginia (2) to penetrate deep into the heart of the bracket.

The table above ranks every system as to how "chalky" it is, i.e. how closely it conforms to the NCAA Selection Committee's 68-team ranking. Nolan Power Index is (by far) the least chalky of systems (as per usual), followed by the preseason polls (which do a lot better than you might expect).

For the rest of the tournament, I will post the results after each day's worth of games and highlight the systems that picked the big upsets (as well as the upsets that everyone missed). Stay tuned!

Notes on the systems:
  1. I include the preseason polls because they, in fact, tend to do a better job of picking winners in March than the postseason polls.
  2. While I aggregated the ESPN Computer picker and ESPN's Decision Tree last year, this year they picked different brackets, so I'm dis-aggregating them again.
  3. Vegas picks are based on the first round match-up, Final Four and National Championship lines as presented by Pinnacle at 9:45 AM EDT on the first day of the Tourney.
  4. Chalk is the tiebreaker in the event of any ties.

Previous Rating Systems Challenges

No comments: