clock menu more-arrow no yes mobile

Filed under:

Reviewing the UWDP Computer Projections

How well was I able to predict Washington, the rest of the Pac-12, and everyone else?

Oregon State v Washington Photo by Ethan Miller/Getty Images

The college basketball season is officially over. Before the season started I decided to come up with my own set of computer projections stemming from some previous research I had done. I laid out my picks in this article here from November 6th. Note that I only looked at the football Power-5 conferences plus the Big East.

I attempted to project the adjusted efficiency margins (EM) as used at for each team rather than wins or losses since I wasn’t able to factor in each team’s schedule. My rankings are based on: recruiting rankings, synergy’s points per possession player stats, experience, and positional adjustments. Each player gets a per minute score based on those factors and after everything is tallied up a small coaching adjustment is applied if a coach has demonstrated a tendency to under or over perform the projections at their current school.

To help give a baseline for what a given EM means, generally 25+ means a national title contender, 20-25 means a top-4 or 5 seed in the tournament, 15-20 is an at large NCAA qualifier, 10-15 is a NIT/bubble team, and below 10 usually means barely NIT or no post-season berth. Villanova just won the national title with an EM of 33.76 while no other team finished above 30.

I’ll start by looking at the Pac-12 and then talk about my best and worst calls from other conferences and how that may change my formula for next season.

Washington and the Pac-12

Coming into the year I was much more optimistic about the Huskies than most people. Washington had a trio of 4-star juniors in Matisse Thybulle, Noah Dickerson, and David Crisp who all performed well the previous year. They ranked 14th, 18th, and 21st in the conference in my system for overall impact. That gave the Huskies a strong base score. Ultimately, I had Washington jumping from 0.34 and 11th in the conference to 13.24 and 8th in the Pac-12. KenPom saw them staying in 11th place but with an EM of 2.28.

The Huskies ended up splitting the difference for their score with a final margin of 7.07 but made me correct in terms of ranking as they were in fact 8th in the Pac-12. I mentioned before the season that since UW had a coaching change, my system didn’t have a way to penalize them for how terrible they were last year. It turns out I overshot Washington’s overall efficiency margin but I also overshot the entire Pac-12.

2018 Pac-12 Projections vs. Actual

Team My Proj Efficiency Margin Actual Efficiency Margin My Proj Conf Rk Actual Rk
Team My Proj Efficiency Margin Actual Efficiency Margin My Proj Conf Rk Actual Rk
Arizona 22.54 17.7 2 1
Arizona State 13.41 14.13 7 2
California 8.59 -7.27 11 12
Colorado 9.69 5.27 10 10
Oregon 17.92 10.24 5 6
Oregon State 15.92 6.99 6 9
Stanford 19.32 8.37 3 7
UCLA 17.85 13.52 4 5
USC 24.49 14.06 1 3
Utah 9.99 12.99 9 4
Washington 13.24 7.09 8 8
Washington State -3.18 -1.66 12 11

The only teams in the conference who finished with a better EM than I thought were Utah (+3), Washington State (+1.52), and Arizona State (+0.72). Everyone else under-performed including me boosting up USC (-10.43), Stanford (-10.95), and Cal (-15.86!) by double digits more than actual. I wasn’t totally alone in over-estimating the conference. KenPom inflated the average Pac-12 team by 2.3 points while I did so by 5.7.

My numbers look a little better in comparison to KenPom’s projections if you look at the absolute value of how far we were off. My projections were off an average of 6.57 towards their EM while Pomeroy was off by 5.38. It just happened that mine were basically all in the same direction while he under and over estimated a few each.

Both Stanford and USC were hit by injuries this season but correcting for actual minutes played still only brings down my projections for each by 1 to 1.5 in EM. Not nearly enough to account for how wrong I was on those.

Top-10 Difference in My Projections vs. KenPom

School My Projection KenPom Projection Actual EM
School My Projection KenPom Projection Actual EM
Ohio State 22.55 7.85 21.16
Washington 13.24 2.28 7.09
NC State 14.67 4.43 14.52
Georgetown 17.55 7.54 7.44
Iowa State 6.81 16.69 6.58
Nebraska 15.5 5.64 13.54
LSU 14.77 6.14 11.62
Arizona State 13.41 5.17 14.13
Indiana 19.61 11.42 10.44
Villanova 19.47 27.48 33.76

When you just look at the biggest disagreements it actually makes me look pretty good. On 6 of the top 8 in this list I was closer with the two exceptions being Washington and Georgetown. The top-4 disagreements all have a common thread as they had a coaching change stemming from a poor record in the off-season. As mentioned above, my system didn’t penalize teams specifically for having a bad year if there was a change in head coach so my projections were significantly higher. That resulted in me going 2-2 as I almost completely nailed Ohio State and NC State while Washington was in the middle and Georgetown was right where he thought they’d be.

I was also closer on each of the next four in Iowa State, Nebraska, LSU, and Arizona State. But then things take a turn for the worse as my projections were outperformed by KenPom on 10 of the next 11 in terms of biggest disagreement and 18 of the next 22.

5 Most Accurate Projections

School My Projection My Conf Rk Actual EM Actual Conf Rk
School My Projection My Conf Rk Actual EM Actual Conf Rk
NC State 14.67 7 14.52 11
Iowa State 6.81 10 6.58 10
Maryland 15.58 9 15.03 6
TCU 17.91 3 18.47 4
Michigan State 24.76 1 25.34 2

There’s not an obvious reason as to why I was most accurate for these particular teams. Maryland and TCU both suffered injuries to some of their best players or else they would have over-performed my projections by a sizable amount. On paper, Iowa State was by far the least talented team in the Big 12 and that stayed true although that was no guarantee that their EM would end up close to where I put it.

5 Least Accurate Projections

School My Projection My Conf Rk Actual EM Actual Conf Rk
School My Projection My Conf Rk Actual EM Actual Conf Rk
Virginia 16.61 6 29.53 1
Vanderbilt 21.23 3 7.93 13
Tennessee 8.01 13 22.27 1
Villanova 19.47 2 33.76 1
California 8.59 11 -7.27 12

The trends are a little clearer on this one. Virginia and Villanova ended up being the best 2 teams in the country for most of the year and I had them as merely pretty good to really good coming into the season instead of great. It’s hard to predict outlier teams like Villanova’s EM score this year and not mess up your formula so I’m not too worried about that. But the coaching factor for both Tony Bennett and Jay Wright who are continually excellent probably needs to be examined.

I had Cal and Tennessee as almost even in the projections but then Tennessee over-achieved by a ton and Cal under-achieved by a ton. I’m not super bothered by this because no one saw Tennessee coming and for Cal, like Villanova, it’s hard to project outlier teams in either direction. The Vanderbilt one bugs me a little because I was higher on them than either KenPom or the media and they ended up being much worse than any of us thought. I’m slightly comforted in that if you substitute actual minutes played for my projected minutes their EM projection drops almost 5 points. I also had Bryce Drew with a positive coaching score which will change going into next season.

In the end my projection came closer for 35 teams and KenPom was closer for 40. I’ll take it for my first attempt at this.

You can follow me @UWDP_maxvroom for all your UW Men’s Basketball news and notes