Is Kentucky Underrated by the Computers?
It's not unreasonable for Kentucky fans and media to feel like their favorite team is being disrespected by the computer ratings systems. Voters in the AP pool judge Kentucky to be the 10th best team in the country, and the consensus from "bracketologists" has the Wildcats pegged around a 4 seed in the NCAA Tournament given their resume. However, the rating algorithms at KenPom.com have the Wildcats a more modest 28th, while a similar algorithm at BartTorvik.com slots the Cats in at 31st nationally. Both rating systems place teams like 13-12 Purdue and 17-9 Texas Tech ahead of Kentucky.
The disparity between the computer ratings and human evaluation is pretty large, and has puzzled Kentucky fans and media. Why is Kentucky rated so poorly by computerized systems? Is there evidence that they may be underrated? Are they an NCAA Tournament sleeper because of this? I take on these questions in this edition of Hoops Insight.
The Ratings Reflect Results
To understand why the ratings systems are relatively down on Kentucky, you have to first understand exactly how they work. Let's take KenPom's ratings for example. They rate each team using a metric called Adjusted Efficiency Margin, or AdjEM for short. This metric measures each team's offensive and defensive performance using points scored and allowed per possession. It is expressed as the difference between points scored and allowed per 100 possessions. Kentucky's rating as of this writing is +17.74, meaning they are 17.74 points per 100 possessions better than an average team. The metric incorporates 3 key factors:
Scoring margin of games, per possession
Opponent's rating
Location of games
KenPom's ratings are meant to be predictive, and on his site you'll see predictions for future games. If a team wins by more on a per-possession basis than predicted, their rating improves. If they don't, their rating declines. The opponent ratings are dynamic, and change as a team improves or declines during a season. If a team improves after UK beats them, UK's credit for winning that game will increase and their rating goes up. The location of games affects the score prediction, as home court advantage is included. For example, UK is predicted to beat Florida 71-67 in Lexington but lose 70-68 in Gainesville. That's the home court advantage effect.
Some of UK's Results are Hurting Their Rating
Given the key inputs to the rating system, it's not hard to pick out some areas where Kentucky is viewed as less impressive by the computers.
First, scoring margin. While UK had some blowout wins early in the season, they were against very bad teams. Since December 7, Kentucky doesn't have a single win by more than 14 points. By comparison, Purdue has a 29 point win over Michigan State, a 19 point win over Wisconsin, and a 36 point win over Iowa.
UK does have 3 instances where they outscored teams by at least 15 in a half: Jan 7 vs Georgia, Jan 29 vs Vanderbilt, and Feb 11 vs Vanderbilt. In each game they trailed at half however, by 6, 7, and 9 points respectively. If UK had been tied at the half each game, they would have won by 15, 16, and 23 points. Just this improvement in these 3 games would increase UK's KenPom rating by +1.1 points in AdjEM, moving them from 28th to 25th.
Opponent rating is hurting UK in one specific instance. The loss to Evansville looks especially bad since Evansville has lost 14 games in a row and fallen to 9-18 and 284th. Having the capacity to lose to such a poor opponent reflects badly on UK. Removing that loss from Kentucky's ledger would improve the Cats' rating by +1.5 pts of AdjEM by itself. That would be enough to move Kentucky from 28th to 23rd.
Making these two improvements to UK's results would barely budge their AP poll ranking, but would improve their KenPom rating from 28th to 17th. That would put their rating much more in line with the perception of AP voters and Bracketologists. But those results can't be changed, and the computer ratings will continue to reflect them.
Is Kentucky Underrated?
If you want to make a case that Kentucky is underrated, there is some evidence to support that. To examine this we'll use Bart Torvik's ratings at BartTorvik.com, which allow you to look at results using all kinds of filters. The ratings are fundamentally very similar to KenPom, with some very slight differences. If you enjoy this article you should check out BartTorvik.com and play around with the filters and see how they impact ratings.
First, we can look at how all teams rate in their Quad 1 games (as defined by the latest NET rankings). I'll only look at teams with at least 3 Quad 1 games on the season to avoid a single fluke result affecting ratings too much. Bart Torvik's website allows me to filter this way, and when I do I see something interesting. When considering just Quad 1 games Kentucky moves up to 7th in the country, and their rating improves over 6 points per 100 possessions! This improvement of 6 points in Quad 1 games is the 2nd biggest improvement among all teams in the top 50, behind only San Diego State (whose resume is bolstered by a 31 point win over Creighton on a neutral floor).
Second, we can look at how teams play in games played on the road or at neutral sites. In road/neutral games, Kentucky shows up as 16th in the ratings with an improvement of 1.4 points per 100 possessions. This improvement is the 6th biggest for any team in the top 50 when comparing road/neutral rating to overall. Only 7 other teams improve upon their overall rating this season in Quad 1 games and away from home: Kansas, Dayton, San Diego State, Penn State, Saint Mary's, LSU, and Stanford.
When considering games against stronger competition, or games away from home, Kentucky's results look much more impressive this season. That seems like a good thing, considering that the NCAA Tournament generally features games against strong teams and is always away from home. Is there a reason to believe that this could lead to better postseason performance?
Similar Teams to UK Had Mixed Results
I looked at results from the 2018-19 and 2017-18 seasons to see if teams who were stronger in Quad 1 games and games away from home tended to overperform in the postseason. Here's what I found:
In 2018-19, 14 teams qualified and 13 of them made the NCAA Tournament (sorry, NC State). Of those 13 teams:
1 was Virginia, the champs
2 (Duke, Kentucky) lost to lower seeds in the Elite Eight
2 (Florida State, LSU) lost to higher sees in the Sweet Sixteen
1 (North Carolina) lost to a lower seed in the Sweet Sixteen
2 (Wofford, Iowa) lost to higher seeds in the 2nd round
5 (Iowa St, Wisconsin, Kansas St, Louisville, VCU) lost lo lower seeds in the first round
So, 8 lost to lower seeds, 4 lost to higher seeds, and 1 won the title
In 2017-18, 7 teams qualified and 5 made the NCAA Tournament (miss ya, Penn State and Marquette)
1 (Michigan) lost to a higher seed in the title game
1 (Gonzaga) lost to a lower seed in the Sweet 16
1 (Florida) lost to a higher seed in the 2nd round
2 (Virginia, Miami FL) lost to a lower seed in the first round
So, 3 lost to lower seeds (including the first 1-16 upset) and 2 lost to higher seeds
Out of 18 teams, 11 underperformed in the NCAA Tournament, 6 more lost to higher seeds, and 1 won the national title. That's not great evidence to support the idea that performing well in Quad 1 games or in road/neutral site games leads to postseason success.
One Glimmer of Hope
The best hope for UK comes when comparing their resume to similar recent teams. Again, you can do this at BartTorvik.com, and here is the link for this year's Kentucky team. Comparing them to other high-major teams with similar resumes and projected seeds, you get a mixed bag. 2 first round upsets (Kansas State 2019, Wake Forest 2009), 3 Sweet 16 losses (Purdue 2009, UCLA 2017, Kentucky 2018), and one very interesting champion: 2014 Connecticut.
Kentucky fans are familiar with this team, as they beat the Wildcats in the title game. The 2014 Huskies had an ugly loss to a bad Houston team and did not beat a top 100 team by more than 12 points during the regular season. They nearly joined the ranks of first round upset victims, needing overtime to escape 10-seed St. Joseph's before going on a miracle run. One important factor makes them unique among all of UK's other resume comps, while also making them similar to this year's Kentucky: they were a great free throw shooting team. They placed 4th nationally, were over 90% in 4 of their 6 tourney games, and weren't below 77% in any tourney game. Most teams who suffer bad losses and don't get many blowout wins aren't very good at free throws, so they miss on easy points. 2014 UConn and 2020 Kentucky are different. They both have the capacity to lose focus against weaker teams, or to step up and win close games against better teams. If I had to predict, I'd say Kentucky will have a scare of a first round matchup in the NCAA Tournament. If they can survive that, they may be able to go on a lengthy run. Whatever happens, it will probably be dramatic. These Cats won't have it any other way.
I want to thank you for subscribing to my newsletter. At Hoops Insight, I hope to use data to find insights that reveal things you didn't realize about the Kentucky Wildcats. I'm using play-by-play data to track what's happening, who's doing it, and who is in the game, in order to show you things the box score can't. Check out my past newsletters in the Hoops Insight Archive, and read about the stats I like to use in my stats glossary.
If you have any questions about things I'm saying, the data behind it, or if you just want to debate a point, feel free to contact me on Twitter at @hoopsinsights or email at sean@hoopsinsight.com. I'd love your feedback on the newsletter and how I can improve. Thanks, and I hope you enjoy my work.