Inside Marquette’s 35 Point Game
College basketball scores in the 30s are not completely rare. But when a team rated in the preseason Top 20 scores only 35 points at home, when a team picked to win the Big East scores only 0.54 points per possession, that does raise a few eyebrows. And that is exactly what we saw as Marquette lost at home to Ohio St. on Saturday.
Marquette’s 20 turnovers were egregious, but the stat that stuck out to me was that Marquette made only 19% of its FGs (10 of 53) in the game. Since I accidentally heard that outcome before watching the game on my DVR, I decided to track exactly how Marquette missed those 43 shots. Was this simply a case of Marquette’s players missing open looks? Or was this a case of strong defense denying clean shots?
According to my informal tally, Marquette missed only 6 easy inside baskets. The most egregious came with 9:30 left in the second half when Todd Mayo stole the ball and completely whiffed on the wide-open lay-up in transition. Another egregious error came on a gorgeous give-and-go passing play which resulted in a missed lay-up with 6:45 left in the game. Marquette also missed 6 wide-open threes, and 2 wide-open jumpers. Thus I would argue that only 14 of Marquette’s 43 missed shots were good shots.
I also counted Marquette as taking 5 threes with a defender closing out late. One involved an off-balance pick, and most involved pull-up jumpers off no action, the kind of three an offensive player can get at any point in the shot-clock. There was also 1 missed half-court heave before halftime. Thus if you feel particularly generous, you might say that Marquette got 20 clean looks.
But according to my tally the other 23 shots Marquette took were extremely low percentage shots. There were clean blocks, partial blocks, underhand lay-up attempts from so far out they had no chance, contested jumpers from the free throw line, threes that players had no business taking, and a multitude of shots in the lane when surrounded by two or three defenders.
When we talk about how match-ups matter in sports, this game might just be the poster-child for that. Marquette’s offense is not based on jump-shooting. The Golden Eagles made just 30% of their threes last year, which was 323rd in the nation. Instead Marquette relies on spreading the floor and using different players to drive and attack.
But Ohio St. might just be the best team in the nation at defending penetration. Certainly Aaron Craft has a lot to do with that, but now that Ohio St. no longer has to cover up for DeShaun Thomas on defense, the Buckeyes can put a lineup of five athletic defenders on the floor at once. And that combination of speed and quickness forced Marquette to settle for some of the worst looking shots they will take all year.
I do think Marquette head coach Buzz Williams needs to do some soul searching at this point. It is inexcusable to have Marquette’s best offensive player, Davante Gardner, coming off the bench when the offense does not have many proven scorers. And it is inexcusable not to design more plays to get Gardner the ball. But this is also a case where we should give Ohio St. a lot of credit for what happened. To force an elite team into 20 turnovers and 23 terrible shots on its home floor is truly dominant.
Michigan’s Levert Falls Back to Earth in Loss to Iowa St.
Michigan’s team is extremely young and John Beilein’s teams typically get better as the year progresses. Mitch McGary was rusty on Sunday and Glenn Robinson was limping late in the game against Iowa St. There is no reason the Wolverines cannot be a Top 10 team by the end of the year.
But if you want to look for a long-term hole for the Wolverines, Michigan’s loss on Sunday still exposed one. Heading into the season I feared Michigan’s limited guard depth. And even though many people were raving about Caris Levert this summer, I remained skeptical. ESPN had Levert ranked as the 69th best SG in the country in 2012. And he had an eFG% of just 39% last year. But Levert did score 17 and 24 points to open the season. So I was fascinated to see how he would do against real competition.
Sadly, Levert did not pass the test. Despite playing 37 minutes, Levert notched just 2 assists and 5 points on 2 of 9 shooting. And if the number of wide open threes Iowa St. got are any indication, Levert’s defense was not perfect either. It was only one game, but with Levert’s good games coming against cupcakes, the jury is still out on Michigan’s guard depth.
ESPN likes to name the various weeks of basketball, “Championship Week”, “Rivalry Week”, “Feast Week”. I feel like we should just go ahead and label this week as “Cupcake Week”. With a small number of exceptions, the time between the Champions Classic and the Holiday Tournaments is some of the least watchable college basketball on the calendar.
I hate these games because I feel like you never get a real feel for any of the teams. Small conference schools play too much zone defense; coaches experiment with unrealistic lineups; players who should never take threes take them with reckless abandon; the games are just ugly.
Can you ignore these games? The short answer is yes. I don’t have an advanced model of predicting NCAA performance. But I have been playing around with a toy model. And I have found that if you just count games against quality competition, you learn almost everything you can learn. Including the margin-of-victory in games against cupcakes only improve the fit of my toy model by about 3%.
In fact, cupcake performances are sometimes completely misleading. To take one extreme example, in 2009-10 Oregon beat Winston-Salem 94-43, UC Davis 95-64, Montana St. 89-66, Mississippi Valley St. 79-51, and Arkansas Pine Bluff 73-53. Those games made it seem like Oregon had a Pythagorean winning percentage over 0.900. But it was fools’ gold. Oregon lost to all sorts of quality mid-major squads (like Portland, Montana, and St. Mary’s), finished 10-16 against quality opponents, finished 7-11 in the Pac-10, and Ernie Kent was fired.
Now this is an extreme example. Someone with a better modeling approach may be able to learn more from these cupcake games. But I don’t think saying that cupcake games have less information is that controversial. By weighting recent games more heavily and capping margin-of-victory, Ken Pomeroy essentially lowers the value of these games in his ranking system. And almost every major ranking system makes similar adjustments.
The only reason I can see to care about cupcake games is that right now we don’t know much of anything. Cupcake games might be weak signals of quality, but they are all we have right now. And there is at least some correlation between cupcake games and how teams will do against quality opponents later. The figures in the next section are designed to show that.
From Cupcakes to Quality Opponents
In the following charts, I am only looking at power conference teams. I define a cupcake game as a game against a team with a Pythagorean rating below .500. I break power conference squads into 5 buckets based on how they perform against cupcakes. Then I show how the teams in each bucket fare against quality opponents.
To take the first chart as an example, I show teams with a Pyth. winning percentage over .950 against cupcakes. I show that these teams have a 2% chance of finishing with a Pyth. rating between .700 and .800, a 17% chance of finishing with a Pyth. rating between .800 and .900, and a 81% chance of finishing with a Pyth. rating above .900.
The first chart shows teams like Wisconsin, Duke, and Indiana last year. These are teams that completely obliterated their cupcake opponents. If you do that, there is over an 80% chance your team will end up with a Pyth. rating above .900 at the end of the year, which would roughly correspond to a Top 15 squad.
Last year’s Kentucky team fell in the dreaded 2%. They crushed several cupcakes early in the year, fell apart as the season progressed, and finished with a Pyth. rating below .800.
The next figure shows where teams finish if they have a Pyth. rating between 0.900 and 0.950 against cupcakes. The Oregon 2010 squad I listed above falls in this category.
The next figure shows where teams finish if they have a Pyth. rating between .800 and .900 against cupcakes. Last year’s Marquette squad falls in this category. While they beat UMBC by 33, they also had some close games such as their 9 point win over NC Central and their 11 point win over SE Louisiana. Around 28% of these teams are going to finish below .800 and miss the tournament.
The next figure shows where teams finish if they have a Pyth. rating between .700 and .800 against cupcakes. Last year’s Providence squad fell in this category.
Finally, the last figure shows teams with Pyth. ratings below .700 against cupcakes. These are the teams that look very shaky, often with multiple losses to bad teams.
Fortunately, even when teams look shaky against cupcakes, their season doesn’t have to be over. 15% of the power conference teams that look very shaky against cupcakes will finish above .800 which is the cutoff for a typical at-large tournament team.
Last year Illinois beat Hawaii and Gardner Webb by 1 point each and yet made the round of 32 in the NCAA tournament. Illinois is an example of a team on the right hand side of this chart.
Bottom Line: Cupcake games have some predictive power, but your eyes are not deceiving you. There is only so much you can learn from these games.