Revisiting UWDP College Basketball Projections

For the last 5 years I’ve been putting out my own computer projections for power conference teams in men’s college basketball. The project started when the decision was being made of whether or not to let go of Lorenzo Romar. I created a database to try to evaluate to what degree the Huskies had underachieved in the recent past with Romar at the helm and the database has only grown since.

At this point it includes information since the 2011-12 season for every power 6 team plus Gonzaga. The formula takes into account a player’s age (years in college), talent (recruiting ranking), as well as prior performance to put out a per minute estimate of that player’s production. Then I add up the totals on each roster based on playing time to come up with the raw totals for each team. I match those totals against the adjusted efficiency margins found at kenpom.com for what the output total is set to predict. If a team drastically underachieves or overachieves the base talent/experience of their roster then that’s a reflection on the coaching job.

I’ll start with some talk about how things turned out for Washington before turning to the rest of the review as well as a look ahead to 2023.

WASHINGTON PROJECTION REVIEW

Coming into this season I projected the Huskies to finish with an adjusted efficiency margin of +7.83 which ranked 9th in the Pac-12 and 61st out of the 77 power conference (+Gonzaga) schools. That would’ve been a reasonably big jump up from their +3.65 total in the 2020-21 season but still wouldn’t put them anywhere close to a postseason berth. The roster itself looked like it might be enough to challenge for an NIT berth but unfortunately Mike Hopkins is the lowest rated current coach in my system which dropped the projection substantially.

And so the Huskies failed to meet even that projection and ended up this year at +5.17. That might not mesh with the general fan’s understanding of each UW squad since the records improved from 5-22 to 17-15. However last year Washington was one of the unluckiest teams in the country as far as their record matching their efficiency-based numbers and this year ended up as one of the luckiest. That’s how UW saw a very modest increase in their efficiency margin despite the big jump in win totals.

Each coach in my model has a raw number which represents their average over/under-achievement but for the projections I tone down the actual effect it has since there is a lot of variation in college basketball. It turns out for Hopkins that if I had applied his full coaching factor I would have underestimated Washington’s total this year but by only taking a portion of it I ended up a little too optimistic.

When you look at the final conference standings the Huskies technically tied for 5th. However, when you look at their adjusted efficiency margins which take into account their dreadful non-conference they did indeed finish 9th as I projected as they get passed by Arizona State and Stanford.

In order to try to account for depth since no team gets through the season unscathed when projecting minutes I assign 70% of the total minutes to 5 starters and 30% of the total minutes to 5 bench players. I ended up correctly predicting Washington’s starting 5 of Brown, Davis, Bey, Matthews, and Roberts. And the 5 bench players I included were the 5 highest minutes totals with Fuller, Bajema, Wilson, Grant, and Sorn. However the way the season played out Fuller and Bajema both played more than 30% of the team’s minutes while the other 3 all played less.

Looking forward to next year for the Huskies is a tricky proposition. Right this second I have the Huskies as the #10 team in the Pac-12 for 2023 with a +6.62 adjusted efficiency margin (starting 5 of Fuller, Bey, Bajema, Matthews, and Wilson). That would be a slight improvement over this year’s team but worse than what I had projected for last year’s roster. Of course there’s no way that Washington’s roster is set as the Huskies have their eye on several upgrades in the transfer portal.

Let’s say that the Huskies end up with what would probably represent their best case portaling scenario based on their known targets: Washington State G Noah Williams, Utah Valley C Fardaws Aimaq, and Oregon C Franck Kepnang. That would bump up Washington’s adjusted efficiency margin to +11.42. For context, Syracuse this year finished 69th nationally with an adjusted efficiency margin of +11.45. For teams that don’t also have a losing record that’s solidly in the “make the NIT but also probably miss the bubble” range.

OVERALL PROJECTION REVIEW

I’ll start with the good news. I had Gonzaga 1st and Kansas 2nd in the preseason and that’s exactly where they finished at the end of the year. The Zags were knocked out of the tournament early but I also did a pretty good job at forecasting future success as Kansas (2nd), North Carolina (4th), and Duke (6th) all made the Final Four from my top group. I had Villanova projected as the best team in the Big East but was generally lower on them than the humans and not in Final 4 territory. Ultimately, all of my top-6 power conference teams made at least the Sweet 16 even though Michigan and North Carolina underachieved throughout the regular season before rounding into form in the NCAA tournament.

For the most part if my projected adjusted efficiency margin is within 2 in either direction I consider that to be essentially a hit. The team was very slightly better or worse than I thought but I captured right about where they ended up. This year 25 of the 77 teams ended up within that range with Notre Dame, Michigan State, and Gonzaga all finishing within 0.1 of my projections.

8 teams finished +2 to +4 over my projection and thus slightly overachieved: TCU, MIssissippi State, Florida, Duke, Penn State, Boston College, Arkansas, and Providence. On the other end of the spectrum 11 teams finished -2 to -4 under my projection and thus slightly underachieved: Missouri, Washington, Virginia, Alabama, Florida State, Georgetown, Arizona State, Ohio State, North Carolina, Georgia, and Butler.

That leaves 19 teams that definitely overachieved and 14 that definitely underachieved. The top-5 overachievers list from bottom to top were Wake Forest, Baylor, Texas Tech, Arizona, and Iowa. Both Texas Tech and Arizona were under 1st year coaches who transformed the respective cultures around each team. Wake Forest and Iowa both got out of nowhere performances. Wake’s Alondes Williams was a good but not great rotation player at Oklahoma who transferred and suddenly became ACC player of the year. Iowa had Keegan Murray go from very good role player as a freshman to superstar national player of the year candidate as a sophomore. Finally, Baylor turned the #112 recruit in the country into a projected top-10 one and done pick.

The major underachiever list from best to worst was Maryland, Georgia Tech, Oregon, Louisville, and Oregon State. Both Maryland and Louisville had their coaches quit/resign mid-season although there’s a chicken and egg situation as to whether the coach losing passion led to the team playing poorly or the coach knew the team stunk and that contributed to them wanting to leave. Georgia Tech lost its 2 best pieces from a tournament team but returned nearly everyone else and it turns out those 2 pieces were irreplaceable. Oregon brought in a bunch of highly successful transfers but the team chemistry fell apart and Altman never put the pieces together. Finally, Oregon State returned 4 of their 6 best players while adding the eventual Pac-12 assist leader to a team that made the Elite 8 and fell from 60th to 320th in defensive efficiency. Scientists will be studying that season for centuries to come.

2023 PROJECTIONS 1.0

Trying to figure out rosters and playing time this time of year is absolute madness. The transfer portal has made determining the favorites right after the national championship a borderline impossible task. LSU lost all 13 players to transfer or the draft and all 4 of their previous freshman commitments after Will Wade was fired. They now have just 5 players on their roster for next year right now. Georgia, Kansas State, Kentucky, and Pittsburgh all also have fewer than 8 scholarship players on their roster for next season as it currently stands. That’s obviously going to change.

The criteria I’m using for the moment is that any player that has declared for the draft, whether testing the waters or not, I am taking off the roster until they say they’re withdrawing. I’m also assuming any player in the current version of Jonathan Givony’s 2-round mock draft is going to remain in the draft until they say otherwise. Any player in the transfer portal is off any roster until they commit. Finally, when I do projections I include 10 players to account for depth. If a team doesn’t have 10 scholarship players right now then I’m assuming the rest of the minutes are coming from walk-ons that don’t improve the overall score. So a team with fewer than 10 scholarship players will make a big rise every time they add a new commitment until they get to that total.

With all of those criteria in place, UCLA right now is the #1 team in my system like they are on several “way too early top-25” lists across the interwebs. Every player on the team is eligible to return and none of them are projected to get drafted plus they add the #2 and #16 ranked players in the 2022 recruiting class. That leaves 12 guys who realistically think they should be playing. Something has got to give and almost certainly a few of them will leave for the draft or the portal but for now the Bruins are on top.

The rest of the top-5 are: Michigan, Duke, Texas, and Gonzaga. The Wolverines are in the top-12 at both ESPN and CBS but my system projects they may be a little underrated despite not living up to expectations this year. Duke and Gonzaga are also consensus top-12 teams from those outlets with a track record of success even though obviously Duke is handing the reins to a first time coach. Texas is the wild card although it looks like at least at ESPN they’re projecting several of the seniors on the roster to leave who haven’t announced yet. If Timmy Allen and Marcus Carr both turn pro then they would fall quickly although my system loves Chris Beard so maybe not too far.

I explained why Kentucky is so low at the moment given the number of players on their roster but Arkansas and North Carolina are the biggest discrepancies at the moment between “the humans” and my model. I had the Tar Heels 4th last preseason and they finished 16th in adjusted efficiency margin even though they made it to the title game. When you take the whole season into account they underachieved with their early struggles even though they came together by year’s end. That means I almost certainly have Davis with an artificially low coaching grade until we see what he does next year and that knocks down UNC quite a bit.

The Razorbacks may be bringing in a trio of 5-stars but my past research has shown that not all 5-stars are created equally. There’s a big gap between the top-5 and 6-10 and another gap between 11-25. That combined with losing their 3 best players (at least for now) have them as a good but not great team at least for the moment.

When it comes to the Pac-12 UCLA is obviously #1 with a substantial gap but right now I have Oregon at #2 with an even bigger gap between them and 3rd place Arizona. The Ducks badly underachieved this year but are adding a pair of 5-star freshmen and still return three 4-star senior starters plus a 5-star junior. This was the first big underachieving season since Altman got to Eugene so my system still gives him a positive grade despite failing with much of the same roster this year. If the Ducks falter again next season I may need to consider adding a weighted component for the coaching grade as opposed to an average of their entire career.

Arizona drops to 3rd because I have them losing their 3 best players for now (Mathurin, Koloko, and Terry) and right now they’re only adding one low 4-star recruit. And without any incoming transfers for now it means half their roster spots are for guys who only played garbage time last year and have no past production to inform the model. I also tamper down the weighting of the coaching grade for coaches coming off their first season so it doesn’t swing wildly in year 2 due to a fluky start. If Lloyd ends up with a top-5 team again with his current roster then he moves into Tony Bennett/Mark Few territory as guys who are trusted to consistently overachieve.

Behind the Wildcats there’s a tier with Washington State (4th), USC (5th), and Arizona State (6th). The Trojans are the most talented of that grouping but my model still doesn’t trust Andy Enfield with his years of underperformance despite turning things around a little with the Mobley brothers recently. I’m also expecting Arizona State to continue to be aggressive in the portal and potentially move up although again my model also doesn’t love Hurley.

There’s a clear drop before you get to Utah (7th), Colorado (8th), Stanford (9th), and UW (10th). If you don’t factor in my coaching adjustment then UW has the best projected roster of any of the 4 right now but Hopkins’ score drops them below the others despite Haase and Smith also earning negative grades. Although Craig Smith is in the opposite boat of Tommy Lloyd as he had a poor first season which helps him in his year 2 grade. Colorado lost 2 starters to the portal already and I have Jabari Walker set to turn pro at the moment but if he did come back it would shoot them up several spots as he’d be an instant Pac-12 player of the year contender.

Unsurprisingly given how this past year went I have Cal in 11th and Oregon State in 12th in a bottom tier of their own at the moment. Neither team has added a top-175 freshman recruit or anyone through the portal so far. Meanwhile the Beavers lost their starting backcourt and arguably 2 best players to the portal. It’s going to take one heck of a spring shopping spree for talent for either team to have any chance at seriously competing in the conference next year.

**********

Questions about any particular team? Feel free to ask in the comments and I’ll let you know why the model sees them the way it does.