Revisiting My 2019-20 MBB Computer Projections

The college basketball season ended sooner than expected so I didn’t get the luxury of gradually entering in the information for each team as they got eliminated. But we’re all caught up now and ready to see how well my computer projection system did at predicting the 2019-20 season’s results. (Note: Back in October I did not project the NCAA tournament would be cancelled due to pandemic. Sorry for missing that one.)

I’ll try to keep the methodology brief here. I started this work following Lorenzo Romar’s dismissal and the arguments regarding the balance between star power and experience (hint: you want both). That led me to the current model which tries to balance each player’s recruiting ranking, their on-court performance, and their grade level against KenPom’s adjusted efficiency margins. If a team consistently underperforms or overperforms compared to the projections then it is attributed to the nebulous bucket of “coaching” and each new team is adjusted based on their specific coach’s rating.

For those unfamiliar with an adjusted efficiency margin, it represents how many points better than average a team is over the course of 100 possessions. A margin in the realm of 25+ is a national title contender. 20-25 puts you in contention for a top-4 NCAA tournament seed. 15-20 usually gets you into the big dance. 10-15 is bubble or NIT. And below 10 usually means a forgettable season.

If you want to start by reviewing my write-up at the beginning of the season check it out here.

Washington and the Pac-12

About 6 months ago Washington finished 3rd in the preseason media poll coming off a regular season conference title. My projections were slightly less optimistic. They saw the Huskies finishing with the 6th best adjusted efficiency margin in the Pac-12 albeit by a close margin. Colorado and Oregon were my runaway favorites but there was an extremely tight grouping between spots 3-7 with Arizona, Arizona State, Oregon State, Washington, and UCLA. All things considered I’d say I did pretty well with that projection.

You may say to yourself “Max you idiot. Washington finished last in the conference, how is thinking they’d be 6th doing well?” I can’t dispute the first part but remember that I attempted to project on the basis of adjusted efficiency margin rather than conference standings. The Huskies actually ended up with the 5th highest efficiency mark in the Pac-12 one spot ahead of USC.

I understand that it sure didn’t feel like Washington had the 5th best team in the conference as the losses mounted but the third of the season with Quade Green eligible when the team was in fact quite good did happen. And while the repeated close losses may have been the result of bad luck or may have been a lack of mental fortitude, there’s no question that the Huskies were infinitely more competitive than your average last place team in the conference standings. The Dawgs were 42nd nationally in efficiency when Quade Green went out. Yet they only fell 12 spots despite going 4-13 in their next 17 games because of the combination of blowout wins and close losses.

In the preseason I noted that this is where my projections rated Washington’s starters compared to the rest of the conference: Isaiah Stewart (2), Jaden McDaniels (15), Quade Green (17), Naz Carter (18), Hameir Wright (38). Stewart may have been a few spots worse than that but he wasn’t the problem and mostly played up to expectations. Green was playing like a top-10 guy in the conference at the time he went out but only played one-third of the season and the replacement combination of Hardy/Tsohonis came nowhere close to that level.

McDaniels and Carter had times where they played like all-conference contenders but lacked consistency and also failed to show up in several games. Wright’s placement was probably about right if you properly rate his defensive contributions.

The above chart shows the actual KenPom adjusted efficiency margin on the y-axis and my projections on the x-axis. Teams that finished above the line there finished better than I expected while teams below it finished worse than expected. The cluster of teams I predicted for spots 4-7 all underperformed to various degrees. Colorado and Cal fell quite short of my system’s expectations despite being projected on opposite sides of the conference standings.

Meanwhile Arizona and Stanford were both substantially better than I expected from an efficiency standpoint while improving from 3rd and 11th in my preseason to 2nd and 4th in reality. Finally, Oregon, USC, Utah, and Washington State all finished right about exactly where I expected them to wind up.

You’ll notice in that graph in the top right it says r2=0.453. For those unfamiliar with that term it means that my projections explained 45.3% of the variation in the actual results. That’s within a few percent of how the system has performed in past seasons and of course we’ll never know if getting to finish the conference tournaments as well as the NCAA tournament would have helped or hurt how well the numbers matched up.

I consider that total to be pretty solid considering how many completely unpredictable things happen over the course of every college basketball season. I put in standardized minute totals for every team (5 starters playing 70% of minutes and 5 backups playing 30% of minutes) with the knowledge that things will change vastly due to injury or other factors. Yet going back in and putting in the correct minutes played only raised the r2 to 0.458 which showed that for the most part things tend to balance out over the course of the year.

Best Calls and Worst Misses

ACC

North Carolina- Predicted: 12.6 and 9th in conference. Actual: 9.4 and 10th in conference.

I would’ve bet a decent chunk of change that the Tar Heels would outperform my projections considering they had only failed to be a top-8 seed in the NCAA tournament once in the previous 16 seasons under Roy Williams. Instead they completely fell apart as injuries and poor chemistry resulted in a disappointing 14-19 campaign. Factoring in actual missed time would have lowered my projection to 11.5 but still above where they ended up.

Florida State- Predicted: 13.6 and 7th in conference. Actual: 20.2 and 3rd in conference.

The Seminoles were coming off of 3 straight top-30 seasons but had lost a lot of talent from the year before. Instead of a middle of the pack season they ended up winning the ACC regular season title outright and would’ve been in line for a top-3 seed in the NCAA tournament.

Big 10

Michigan State- Predicted: 24.0 and 1st in conference. Actual: 24.1 and 1st in conference.

The Spartans weren’t the best team in the country as just about everyone thought in the preseason but they still were a top-10 team and the best team in the best conference all season long. Even if they lost the 3-team tiebreaker in the conference standings this team was just about what we thought they would be.

Rutgers- Predicted: 7.9 and 12th in conference. Actual: 17.7 and 9th in conference.

This is actually a joint bad call between Rutgers and Minnesota who I had rated as almost identical teams in the preseason and wound up finishing with almost identical adjusted efficiency margins. The problem? I had them pegged as close to top-100 teams and they instead finished 27th and 28th in the country.

Big 12

TCU- Predicted: 9.2 and 10th in conference. Actual: 9.1 and 8th in conference.

Iowa State and Kansas State ended up being worse than I expected to keep the Horned Frogs from actually finishing in the cellar in the Big-12 standings. But TCU finished well outside of NCAA tournament consideration exactly where I expected them to be.

West Virginia- Predicted: 14.4 and 5th in conference. Actual: 21.3 and 3rd in conference.

I had Kansas and Baylor as 1 and 2 by a decent margin coming into the season and they finished there although both were quite a bit better than I thought. West Virginia was an NCAA tournament contender based on my projections but instead would’ve been pretty comfortably a top-5 seed in the tourney with their actual results.

Big East

Creighton- Predicted: 20.1 and 1st in conference. Actual: 20.9 and 1st in conference.

This may be my best call of the entire season. I had the Blue Jays pegged to win the Big East and that’s exactly what they did despite being projected #7 in the conference’s media poll.

Butler- Predicted: 12.7 and 7th in conference. Actual: 18.4 and 4th in conference.

Overall there weren’t many misses in the conference as everyone else finished within an efficiency margin of 3.5 between my projections and actual. The Bulldogs had Sean McDermott emerge as one of the most efficient players in the country and outside of a pair of 3-game losing streaks were a dominant team for most of the season.

SEC

Arkansas- Predicted: 14.3 and 5th in conference. Actual: 13.6 and 5th in conference.

With Eric Musselman and a fleet of transfers coming into the program there might’ve been some extreme variability in Arkansas’ results. Instead they finished right where I thought they would as a team likely on the right side of the bubble.

Vanderbilt- Predicted: 10.3 and 13th in conference. Actual: 0.1 and 15th in conference.

I thought the Commodores were going to rebound and be respectable after going winless in the SEC last year. Instead they went 3-15 in a bad SEC and may have won the title of worst power conference team in the country. This would’ve looked a little better if I had known that Aaron Nesmith would be lost for the season partway through the year as he’s a likely 1st round draft pick. Overall though 10 of the 15 teams in the SEC did worse than my projections as a conference that was viewed as possibly the best in the country wound up finishing 6th and below the Pac-12.