Indiana's Weakness

If you were to force me to pick a team I think is going to win it all, I would likely take Indiana. On the Sweet 16 efficiency graph, the Hoosiers have the best offense of anyone and an average defense (relative to the other 15 teams). C.J. Moore pointed out recently that the Hoosiers' weakness is playing at a slow pace. Moore noted that Indiana's record is bad in games with 64 possessions or less.

I am naturally pretty skeptical of change in pace being a big factor in a team's success. Ty Lawson's UNC team from a while back is the prime example. They played at a tremendous pace, but were equally as deadly in the half court (contrary to popular opinion). Great teams tend to be great, pretty much regardless of pace. To examine Indiana, I wanted to go beyond simply looking at their record by pace. Instead, I looked at the effect of pace on offensive efficiency and defensive efficiency. I wanted to make sure to account for strength of schedule. To do this I calculated AdjO and AdjD for each game as:

AdjO = (Indiana's Game Points Per 100 Possessions) - (Opponent's Average AdjD)

AdjD = (Opponent's Average AdjO) - (Indiana Opponent's Game Points Per 100 Possessions)
Note: Higher numbers are a good thing for Indiana for BOTH AdjO and AdjD

The highlighted portion of the table shows the culprit of Indiana's struggles. In games with under 65 possessions, the Hoosiers are fine offensively but have been much worse defensively. This is probably an unexpected finding, but I think there are some reasonable explanations. First, I didn't think for a second that Indiana would be a worse offensive team in slow paced games. The Hoosiers have the number one offense in the country. The number one offense in the country isn't likely to have any big weakness. Regardless of the pace of game, Indiana can score the basketball. Still, the question remains: why does Indiana's defense struggle in slow paced games? 

The convenient answer is to blame it on Jordan Hulls. He has been perceived as Indiana's defensive weakness for most of the season. This very well might be a part of it. In the half court, opponents may be able to expose Hulls as an on ball defender. Maybe a better reason is Indiana's defensive style. The Hoosiers  don't let their opponents get good looks and don't foul. They are 18th in the country in opponent eFG% and 14th in the country in opponent FTRate. However, they are very average when it comes to rebounding (152 in DR%) and turning teams over (120 in TO%). I suspect that these defensive shortcomings are more exposed in a grind it out, slow paced game.

The following shows the effects of pace on Indiana in graphical form: 

As you can see, Indiana's round of 32 game was an exception to their season performance. Indiana was very good defensively on Sunday, but scored fewer points per possession than their opponent's defensive average for the first time all season. The thing that might stand out most to me from these graphs is Indiana's defense against faster paced teams. In games over 70 possessions, Indiana's defense has been great. Like I said before, IU's offense is great regardless of pace. So in uptempo games, it's going to be very difficult to beat the Hoosiers.

Final Four Values

The running narrative for the NCAA this season has been all about the parity at the top. Yet here we are in mid March and despite "the field being wide open", everyone is taking Louisville. Now I'm not saying this is a bad idea. Louisville may end up cutting down the nets in Atlanta. However, by picking Louisville you are forcing yourself to have a "near" perfect bracket in order to win your pool. In a standard scoring system, you generally have to get the national champion right to win it all (with an example of an exception being if Butler had won). If you are one of the few people in your pool to take a unique team to win it all, you don't have to be all that great with your earlier picks. With Louisville, however, you're going to have to somehow pick up points on other people with Louisville.

On ESPN, you can currently look at the percentage of people in the country taking any team to reach any round. I thought it would be interesting to compare the average person's picks to Ken Pomeroy's log 5 projections. This will help determining the "value" of final four picks. If a team's value is especially high, it might be worth it to take the gamble in order to attempt to position yourself better among other Louisville brackets. Obviously the log 5 projections aren't perfect. Thus, I decided to include the Vegas odds that each team reaches the final four in the table too. The Vegas odds weren't actually used in finding value teams, but they can either reinforce (or not) the log 5 rankings.

The DIFF column on the right is simply (log 5 final four%) - (ESPN final four%). The thing that jumps out is the "Florida Effect". Not only does Florida shatter the top spot for best value, but they also turn Georgetown and Kansas into bad values. There's a whole bunch of uncertainty surrounding the Gators, but if you trust log 5 (and Vegas) they are nearly a must in your bracket. Miami and Ohio State are the other two that stand out. The conference tournaments have a ridiculous amount of pull on ESPN brackets, or so it seems. Ohio State, Miami, Kansas, and Louisville are all in the top five most picked final four teams (with Indiana) and are coming off conference tournament championships. Syracuse has struggled, but they seem like a much better alternative to Indiana in the East than Miami. Finally, if you decide to go with Pitt in your final four, log 5 says you have an 11.7% chance of taking a commanding lead in your pool.

Quantifying Carrick Felix's Defense

Arizona State, picked to finish 11th in the Pac-12 preseason poll, has been a pleasant surprise this year. Currently they are right on the outside looking in at the bubble, but just about no one thought they would be in this position entering the season. Last month, I received an email from Arizona State's assistant coach Eric Musselman. In the email, he told me to check out Carrick Felix's "amazing" defense this season for the Sun Devils. Since then, he has also been quoted on the same topic in this article. Musselman said of Felix, "He’s got to be one of the top five defenders in all of college basketball." Naturally, I decided to try and quantify Felix's defense to dig deeper.

To look at Felix's defense, I decided to just look at Pac-12 games. Generally, ASU has used Felix to guard a top scorer. Felix's versatility is a tremendous advantage here, as he has guarded point guards, shooting guards, small forwards, and power forwards. I wanted to find conference games where Felix guarded just one player for a full game. Then, I could compare the player he guarded's game stats to season averages to get the "Felix Effect". I realize that this method has some limitations. First, the goal of defense is not solely to stop your man. Instead, teams tend to have a defensive philosophy rooted in help principles. Also, there are many circumstances where players stop guarding their assignment over the course of a game. The two most common are transition plays (where the primary goal is to stop the ball) and switches due to screens. Regardless, I identified nine conference games where I determined that Felix primarily guarded one man for nearly (if not all) of the game. The following is a graph of these nine instances. The head of the player Felix guarded represents that player's season ORtg, while Felix's head represents that player's game ORtg. When Felix's head is higher, it means he was "bad" on defense. When Felix's head is lower, it means he was "good" on defense.

 

As you can see, Felix started the year out with two great games against Andre Roberson (Colorado) and E.J. Singler (Oregon). The only player to significantly get the best of him in conference play has been Mike Ladd of Washington State. In total, Felix has held six under their average ORtg and three over their average ORtg.

The above table shows Felix's defense in more detail. Negative numbers are a good thing from Felix's perspective (hence the green formatting). On average, Felix has lowered a player's offensive rating by about seven points per 100 possessions. We can also see that although Ladd was very efficient in his game against Arizona State, his usage was lower than normal. A defender's ability to affect three point percentage has been heavily scrutinized, so we should probably take those percentages with a grain of salt. However, I fully agree with the numbers concerning free throw attempts. I thought Felix did a great job of not trying to do too much on defense and avoiding fouls. The numbers do say that Felix has had a tendency of not letting players get to the line as much as they are used to.

I wanted to take a look at Victor Oladipo, who is regarded by many as the best defender in the country. Tom Crean's usage of zone defense, however, complicated matters. I was only able to identify five conference games where Oladipo spent the vast majority of his time on one offensive player. Those games are below:

When watching film, it's easy to tell the difference in defensive style between Felix and Oladipo. The latter is very agressive, ranking 20th in the country in steal percentage and averaging 3.6 fouls per 40 minutes. it's hard to really take anything from this small of a sample size, but I was surprised that offensive players have gone at Oladipo. Ryan Evans was the only player who was used on offense less than normal against Indiana. 

It's no secret that defensive statistics are far behind offensive statistics. One of the biggest talking points at the  MIT Sloan Sports Analytics Conference last weekend was the use of new technology to better measure defense. This is especially true for the college game, where we don't have much beyond blocks and steals. I will be the first to say that the best way to judge a player's defense is to watch film. Clearly, Felix's defensive prowess is easily spotted by watching Arizona State play. My goal was to back up the film with numbers. Although the method has some flaws, I think combined with the eye it can give a better total picture of a player's defensive performance.

The Effects of Offense on Defense

Last month, I looked at the effect of Louisville's offense on its defense. When a Louisville offensive possession ends, they get to set up their press. Setting up their defense makes it dramatically more difficult for opponents to score. Both field percentage and turnover percentage are influenced by the Louisville press.

The one and done nature of the NCAA Tournament does not allow for teams to have an off night. Poor shooting can knock a heavily favored team out for good. A poor shooting night is fairly difficult to control because it happens to all teams at some point. Going back to the Louisville example, a team is going to be even more vulnerable to an upset if their bad shooting also affects their defense.

I looked at the 10 teams with the best NCAA Championship odds according to Vegas Insider (note: the odds were made before the Michigan-Michigan State game). I then took these 10 teams and compared their opponent's shooting efficiency when they make a shot leading to the defensive possessions versus when they don't. This data comes from hoop-math.com:

As you can see, Indiana and Louisville's defenses are most negatively influenced by missing shots. This does not account for the extra turnovers caused by the Louisville set defense, so Louisville is probably more dependent on making shots than Indiana. On the flip side, Michigan and Arizona's defenses hardly change regardless of their previous offensive possession. Defense is not quite Michigan's specialty (41st in AdjD), but Arizona is definitely respectable (23rd in AdjD). Thus, Arizona may be less susceptible to a bad shooting night because of strong "transition" defense than other championship contenders. I'll attempt to take a closer look at this in the near future.

Butler's Winning Blueprint

Going into last Saturday's Butler-Indiana game, the question was how would Butler's shaky defense stop quite possibly the best offense in the country. Indiana scored 1.13 points per possession, a more than respectable number but still their 3rd lowest output of the season. The gameplan for Brad Stevens was pretty  obvious and discussed all over Twitter: let Freshman guard Yogi Ferrell take long jumpers. On the season, Ferrell is 35% from 2 and 25% from 3.

Stevens chose to put the stellar defender Roosevelt Jones on Yogi Ferrell. If the gameplan was to let Ferrell shoot, then Butler conceivably could have put someone like Rotnei Clarke on Farrell, saving Jones for a more efficient offensive player. However, Stevens purposefully put Jones on Ferrell  in order to more effectively stop the Ferrell-Zeller pick and roll. The video below of the opening play demonstrates this:  

Butler chose to switch screens and use Roosevelt Jones' versatility. Although only 6'4", Jones is a very good rebounder for his height and kept a body on Zeller at all times. Meanwhile, Ferrell felt compelled to go at the bigger/slower Andrew Smith. Smith did a solid job of keeping Ferrell in front of him while still baiting him to take shots. Of course, it should be noted Jones and Smith both fouling out can probably be attributed to at times playing defensively out of position. The results?

Yogi Ferrell: 3-9 from two, 2-4 from three, 6 turnovers, used on 30% of Indiana possessions (team high)

Cody Zeller: 4-9 from two, 0-0 from three, 2 offensive rebounds, used on 21% of Indiana possessions (4th highest)