Introducing the KYUSAG Model: Preseason CAA Predictions

(Editor’s Note: You say it “kai-usage.” Almost like you’re saying the name of the band Kyuss. Anyway, take it away, John…)

With the 2020–21 NCAA men’s basketball season tipping off today, preseason predictions abound. The CAA preseason poll, released two weeks ago, placed Hofstra in pole position as they look to repeat as conference champions. Here at No Bid Nation, we have also been working very hard to develop some rankings of our own with our new basketball rating model — Kearns’s YUSAG Model, or KYUSAG for short (you can find our methodology here). This regression-based model will assign strength ratings (YUSAG coefficients, the points above an average team) to every team in the NCAA, but we hope to focus new attention to the teams comprising the Colonial Athletic Association.

Our preseason rankings are based on our prior estimates, which are designed to balance early estimates from the model that are easily skewed by outlier games. Prior estimates are constructed by predicting this year’s YUSAG ratings based on recruiting, roster turnover, and previous performance. Crucially, it is important to note that this model does not integrate information about coaching quality, injuries, or other variables that would subjectively influence the perception of each team. We hope these rankings give insight into which teams have gotten better (or worse) since March.

However, it is important to explicitly note that these rankings can be biased in certain, and perhaps unexpected, directions because of the reliance on the prior estimates. One problem for many smaller conferences is the lack of recruitment of starred players. For each point better in 247Sports recruitment composite rating, a team’s coefficient improves by .12 points. As the season progresses and I accumulate more data, the rankings and predictions will improve mightily.

For more data and statistics on college basketball this season, including visualizations, continue to follow us here at NoBidNation.com, at our data Web App and on Twitter @databyKearns.

Here’s how our model predicts the CAA season turning out, as of this morning:

10. James Madison

2019-20 Position: 10th

2020 CAA Preseason Poll Position: 9th

YUSAG Rating: -6.09

What the model says: The conference-worst Dukes rounding out our rankings may not come as a surprise. Indeed, James Madison was not just bad last year but terrible. This year might not be as bad, but it will probably not be great. Nevertheless, there are some positive developments heading into this season. James Madison is one of two teams in the CAA to recruit a 3-star player, and this year’s transfer recruiting class adds 2.1 PORPAGATU! (which stands for Points Over Replacement Per Adjusted Game At That Usage, henceforth referred to as PPG! — read more about this stat in the methodology) above replacement level, second-most in the CAA. It remains to be seen if these new additions can make an impact on the court for JMU.

9. UNC-Wilmington

2019-20 Position: 9th

2020 CAA Preseason Poll Position: 8th

YUSAG Rating: -5.76

What the model says: After suffering through a frustrating season last year, the immediate future may not look any better for the Seahawks. Last season, James Madison and UNCW duked it out to see who could be the worst team in the CAA (both teams finished more than two YUSAG points worse than the 8th-ranked team in the conference). Last year’s rating drags on their prediction heavily, leading them to place 9th on our preseason rankings. One bright spot for UNCW is that their returning players account for 87% of minutes played last season, the highest mark in the CAA. This team is another year older, another year wiser, but it remains to be seen if it will be another year better.

8. William & Mary

Luke Loewe attempts a layup against Delaware in Kaplan Arena, wearing the gold jerseys.

2019-20 Position: 2nd

2020 CAA Preseason Poll Position: 10th

YUSAG Rating: -4.73

What the model says: For those of you following Tribe basketball, it may not come as a surprise that this team is low on any ranking out there. The departure of four starters, including newly-minted Atlanta Hawk Nathan Knight, makes this team much weaker heading into Dane Fischer’s second season as head coach. For those who say that the Tribe are too high in this ranking, it is not because we are weighting their good play from last season heavily; William and Mary actually had the 5th-worst YUSAG rating in the CAA last season. Indeed, it is hard to find reprieve for the Tribe in recruiting, roster turnover, or prior performance except that they are not the worst any of the categories. This team may have to rely on more Fischer magic if they have any ambitions for this season.

7. Delaware

2019-20 Position: 5th

2020 CAA Preseason Poll Position: 2nd

YUSAG Rating: -3.22

What the model says: The model seems to question if Delaware seems poised to improve upon a solid showing last season. With the 3rd-highest PPG! of returning players, the core of the Blue Hens roster can continue the strengths of last year’s team. The departures of Nate Darling and Justyn Mutts do place some question marks on this team, as the model expects the Delaware to play markedly worse this season. Delaware may be one of those teams that is biased by the regression approach, so we will have to wait and see.

6. Towson

2019-20 Position: 3rd

2020 CAA Preseason Poll Position: 4th

YUSAG Rating: -2.79

What the model says: Following a very successful season, it may come as a surprise that Towson is ranked this low. However, like many of the teams in the CAA this year, the Tigers are undergoing a roster transformation. Returning players only account for 49% of minutes played this year. The players that have left contributed an estimated 8.7 PPG! above replacement level last season, the fourth largest loss of quality in the conference. The players that are left had only contributed 2.2 PPG! above replacement level last season, the lowest figure for any roster. The return of Zane Martin to the Tigers brings some much needed talent, but this team may face some challenges.

5. Northeastern

2019-20 Position: 6th

2020 CAA Preseason Poll Position: 7th

YUSAG Rating: -2.61

What the model says: Last year, Northeastern were one of those teams that was much better than their record suggested; the Huskies boasted the second-best YUSAG rating in the CAA last season. Their conference tournament finals appearance proved that. However, this team is drastically different. Northeastern pips Hofstra for the largest loss in minutes played by rostered players while internalizing the largest loss in quality by far, as departing players accounted for 11.4 PPG! above replacement level. Northeastern will probably always be competitive with coach Bill Coen, but this team is much different than the one that has bullied the CAA over the past few years.

4. Drexel

2019-20 Position: 8th

2020 CAA Preseason Poll Position: 3rd

YUSAG Rating: -2.49

What the model says: If any team were to be labeled as the “dark horse” in the CAA, it might have to be Drexel. After a frustrating 2019–20 season, this team has quantitatively improved. Returning players account for 76% of minutes played last season and 7.6 PPG! above replacement level, the highest mark in the CAA. The Dragons are poised to surprise as their core players continue to gel and grow.

3. Hofstra

2019-20 Position: 1st

2020 CAA Preseason Poll Position: 1st

YUSAG Rating: -2.01

What the model says: The reigning champs only third? There are lots of reason to be positive about Hofstra: their team was legit last year (the only team in the CAA with positive offensive and defensive ratings) and their coaching is fantastic. The model, however, sees the second-largest roster turnover based on minutes played and PPG! — couple this with a below average remaining roster as measured by PPG! above replacement level, and this could be a mediocre season for Hofstra. However, this team has proven its strength and its eminently possible Hofstra once again challenges for the CAA title.

2. Charleston

2019-20 Position: 4th

2020 CAA Preseason Poll Position: 6th

YUSAG Rating: -0.54

What the model says: The data suggests that there are a lot of average components to the Cougars. Their roster turnover is average, and their returning player quality is average. However, one dark horse variable for Charleston is their incoming class. Recruits are accounting for 2.9 PPG! above replacement level, the biggest additions for any team in the CAA this season. While these players have big shoes to fill following the departure of Grant Riller, this team will be intriguing to follow in the coming months.

1. Elon

2019-20 Position: 7th

2020 CAA Preseason Poll Position: 5th

YUSAG Rating: -0.22

What the model says: The YUSAG ratings from the bottom of the CAA to the top are incredibly close — this conference is a conference in flux. The difference in ratings between first and 10th is 5.9 points. The model thinks Elon is the best team in the conference, an opinion not shared by other commentators, but not by much. Nevertheless, their returning players account for 78 percent of minutes played last year and the second-most PPG! above replacement level. They have also added four recruits with a starred designation, more than the rest of the CAA combined. Elon has improved mightily and might come to surprise some people, even if they are impacted by injuries to Zac Ervin and JaDun Michael.

A box plot modeling conference finishes in the 20-21 CAA season. The line in the middle of the box signifies the average finish, while the edges of the box represent the 25th and 75th percentile. In other words, the team’s conference finish will fall within the box 50% of the time.

Thanks to John Kearns for his hard work on this model. Follow him on Twitter @DatabyKearns and keep your eyes open for more explorations into the KYUSAG model.

Leave a Reply

Your email address will not be published. Required fields are marked *