Your cell phone, Alexa, Google Home and your smart TV eavesdrop on your every conversation, track your every surfing, shopping and buying experience on the Internet, know where you go when you leave your home, what music you listen to, what movies you watch, where you work, what you eat for dinner, how you spend your money, and who your friends and relatives are. You are either ok with that or at least accepting of that intrusion on your privacy, and yet you are viscerally opposed to computers ranking your college basketball teams, selecting teams for March Madness, and seeding them in NCAA tournament brackets. You are a typical American sports fan. I get it.
Sports fans are more comfortable with people they can see and hear than they are with computerized algorithms that can’t defend themselves at cocktail parties. You are happy that a panel of supposed experts—the selection committee—will invite the teams, seed the teams and position the teams in eight pods for rounds one and two of the tournament. I get it.
You focus your attention on bracketologists at ESPN (Joe Lunardi), CBS Sports (Dr. Jerry Palm) and dozens of other news outlets and faceless sports blogs in the hope of gleaning that one nugget of information that will help you fill the winning bracket in your NCAA Tournament pool. I get it.
There’s just one computer-calculated fact you should consider: Nothing is a better predictor of tournament performance than the seed a team is given, and the committee will make mistakes in seeding teams. As we discussed in NCAA Tournament Flaw #3 yesterday, high seeds are a self-fulling prophecy because the field is too large, the best teams are widely dispersed, the lower seeds are watered down by conference tournaments and yet the matchups are best against worst. Were the committee to seed teams flawlessly, you could go “chalk” and do very well in your pool. But, we know the committee will make seeding mistakes.
Last year the committee mistakenly seeded Wisconsin #8. The eyeball test was the likely culprit. Wisconsin played slowly and carefully and that never looks good to the experts. The Badgers met Villanova, mistakenly seeded #1, in the second round. In fact, the teams should both have had a #2 seed. The game was a tossup and the Badgers happened to prevail. (RPG got that one right but not many fans did.) To further compound the chaos, Florida’s last second shot eliminated Wisconsin and allowed under-seeded South Carolina (#7) to defeat its conference rival and surprisingly advance to the Final Four. All because the seedings were wrong.
Also last year: Misguided by a string of late season losses, the committee seeded Xavier at #11 when it deserved a #4 seed. This is called recency bias—what have you done lately? Commentators like to look for teams that have “momentum” going into the tournament, like Big Ten conference tournament winner Michigan (hint, hint!). We see no correlation between recent performance and tournament performance. Like professional golfers, basketball teams overachieve for a couple of weeks and then regress to their mean performance level. They may also underachieve for a couple of weeks and then progress to their mean performance level, like North Carolina may do (hint, hint!). Thus, season-long performance levels tracked by RPG are a better predictor of tournament success.
Last year Xavier defeated over-seeded Florida State (#3) to reach the Sweet Sixteen where it “upset” Arizona which should also have been a #4 seed but had mistakenly been given a #2 seed. In the Elite Eight Xavier lost, predictably, to over-seeded Gonzaga (seeded #1 but really a #3). All because the seedings were wrong.
The top three seeds in each region are the only teams with a reasonable chance of winning the national championship. However, teams that are mistakenly over-seeded in the top three are targets for “upset” by teams that are mistakenly under-seeded. Witness Wisconsin over Villanova, and Xavier over Florida State and Arizona in the example above. At the moment, the most highly regarded bracketologists predict that Xavier and Kansas will be this year’s upset targets. Bracketologists guess that these two teams will receive #1 seeds when Xavier deserves a #3 seed and Kansas deserves a #4 seed (according to our RPG rankings).
According to the bracketologists, the teams most likely to be under-seeded (compared to RPG rankings) come from Mid-Major conferences. This is called brand bias. Cincinnati, Gonzaga, and Wichita State are ranked ahead of Xavier by RPG, and, Houston and Nevada are additionally ahead of Kansas.
Almost every year, a #12 seed “upsets” a #5 seed in the opening round. This is because the committee reflexively positions Power Six conference also-rans in the #5 slot and minor conference champions in the #12 slot. The committee doesn’t have a scientific way of comparing these two categories of teams. Joe Lunardi currently projects Middle Tennessee State at #12 (seriously under-seeded) and if that holds true, an unsuspecting #5 could well become a first day casualty.
In addition to recency bias and brand bias (Power Six conference teams are given preferential treatment as compared to Mid-Major conference teams), you will hear this week a lot of talk about good wins and bad losses. A win over a ranked team or an upper division Power Six conference team is considered a good win no matter how well or how poorly the winner played. A loss to a lower division Power Six conference team or to a Mid-Major conference team is considered a bad loss no matter how well or how poorly the loser may have played. As you know, RPG grades the playing performance to differentiate wins from one another and losses from one another.
RPG also calibrates performance grades according to strength of opponent. Simplistically, a grade of B becomes a B+ or an A against a strong opponent but becomes a B- or C against a weak opponent. This year the difference between an elite team (ranked 1 – 12) and the weakest of all teams (ranked 251 – 351) is not as great as in some previous years. Our 32 ranked teams, arguably the best 32 teams in America, played precisely 13.86% better against the weakest teams than they did against the elite teams. Against teams in between our ranked teams played progressively less well (11.54% better against the weakest teams than against teams ranked 13 – 37 and so on). So, a good win this year is not all that great, and a bad loss is not all that bad. It’s a year of relative mediocrity and parity.
St. Mary’s is often criticized for losing to Washington State but their grade of 82.33 is actually better than Duke’s grade of 80.00 in its loss to NC State. Duke’s loss is considered a good, or at least ok, loss because it was to a conference rival that will be in the NCAA tournament field. However, RPG knows that St. Mary’s played better, relative to its opponent, than did Duke.
While you may hate it, the RPG computer dispassionately and consistently ranks teams according to their mean seasonal performance, calibrated for strength of opponents. There is no recency bias. There is no brand bias. Good wins and bad losses are calibrated precisely. After we hear the selections and seeds on Sunday, we’ll compare committee decisions to the cold-hearted rankings of the RPG system and make some guesses as to which teams have the most advantageous paths to the title, and which teams are the likeliest victims of “upsets” due to incorrect seeding.