Project Defense: Charting The Jayhawks
(Ed: Today's guest post is from David Hess, who runs The Audacity of Hoops, a blog dedicated to statistical analysis of college basketball, and also writes for TeamRankings.com. Hess' Herculean attempt to chart Kansas players' individual defensive efficiencies through the Big 12 season -- by meticulously logging film of every Jayhawks game -- caught our attention, and we asked him to summarize his findings in the Tourney Blog. This is what he learned.)
When an offensive player takes a shot, the scorekeeper pays attention, crediting him with a field-goal attempt and, if he's lucky, a made field goal. The man defending the play, on the other hand, is usually ignored. We know 13th-seeded Oakland's Keith Benson blocks 3.6 shots per game, but we don't know how many shots he faces, and we don't know how many other misses he forces.
In Basketball On Paper, Dean Oliver, the Godfather of modern tempo-free analysis, laments this lack of useful individual defensive data, and writes about a WNBA pet project of his: Project Defensive Score Sheet. This was an effort to keep track of individual defensive credit or blame for all of the shots and turnovers a team allows. This January, when I finally read Oliver's book for the first time, I was inspired to apply his method to college basketball.
Having grown up in Kansas as a Jayhawks fan, I decided to start charting every KU game I could get my hands on. That turned out to be 19 in all -- every game from Jan. 9 through today, excluding a home game against Colorado.
For comprehensive notes on the charting process, you can read the introductory Project Defensive Score Sheet post on my blog, but the basic idea is that I kept track of every time a defensive player forced a missed shot or a turnover, allowed a field goal, grabbed a defensive rebound, or committed a foul that resulted in free throws.
I then used formulas created by Oliver to combine those stats into three summary values:
• Defensive Possession% (or DPoss%) – The portion of a team's possessions that are ended by the man a player is guarding. Twenty percent is average, and a higher number means the player is more involved with on-ball defense.
• Stop% -- The fraction of possessions a player is involved in that end without the other team scoring a point. Fifty percent is average, and higher is better.
• Defensive Rating (DRtg) – An estimate of the points per 100 possessions that the player's team allows while he's on the floor. For DRtg, 100 is about average, and lower is better.
I'll use these values to first take a look at Kansas, then spotlight a few players on NCAA tournament teams the Jayhawks faced this year.
The chart below shows the nine Kansas players who averaged 10-plus minutes per game during the Project. It's abundantly clear that Markieff Morris is the Jayhawks' most dominant defensive player. In fact, his advantage over his twin brother, Marcus, in DRtng is nearly as large as Marcus' advantage in ORtng (124.2 to 120.4).
What aspect of his play makes Markieff the better-defending Morris? I used my charting data to calculate a few more key defensive stats for the twins, as well as for Thomas Robinson, and it turns out Markieff is ahead almost across the board. Robinson is a monster rebounder, but Markieff turns opposing shooters into Alcorn State, forces more turnovers and sends people to the line less than the other Kansas bigs (despite fouling more often; apparently many of his fouls are moving screens or elbows):
|Player||DReb/40||Opp. FG%||Opp. TO%||Opp. FTA/FGA|
Now, let's look at Kansas's two point guards, who brought up the rear in terms of defensive rating. Though the end results are the same, the way they get there is very different. Tyshawn Taylor is a gambler -- the opponents he guards have a high turnover rate, but when they do get a shot off against him, they hit it at an alarmingly high percentage. Elijah Johnson eschews the turnover in favor of forcing a tough shot, but has trouble avoiding fouls:
|Player||Opp FG%||Opp TO%||Opp FTA/FGA|
Brady Morningstar has a reputation as a defensive stopper, but the numbers don't support it. Comparing him to KU's other shooting guards, both Josh Selby and Tyrel Reed have Morningstar beat in terms of forcing bad shots and turnovers. The caveats here are A) he may be matched up against better offensive players, and B) he does do a slightly better job of off-ball defense, as the man he's guarding tends to be involved in fewer plays (as shown by his low DPoss%):
|Player||Opp FG%||Opp TO%||Opp FTA/FGA||DPoss%|
Watch Morningstar's method of defending jump shooters closely next game: He prefers to put a hand right up in the shooter's face, obscuring his vision, rather than making any attempt at blocking his shot. Perhaps this technique isn't as successful as Morningstar thinks it is.
One more note: Josh Selby's numbers may be a bit misleading, overrating his current level of defensive play. Before missing a couple games due to a foot injury, his Stop% was an excellent 59.7, which would rank him second on the team. Since his return, it's been a lowly 40.7, which is worse than Taylor or Johnson. Announcers have repeatedly mentioned that Selby is pressing on offense; he seems to be lost on defense as well.
Who Defended KU Effectively?
Because I only have 1-2 games of defensive data from each of these teams, sample size is an issue, so an in-depth investigation into the numbers would be a waste of time. What I'll do instead in simply look at the best defensive performance against Kansas from the five tournament teams in my data set (Kansas State, Missouri, Texas, Texas A&M, and Michigan).
Texas A&M may have a stronger frontcourt than I -- or anyone, for that matter -- realized. Of tourney-bound forwards who faced the Jayhawks, four of the top nine defensive performers hail from the Aggies. Granted, three play relatively few minutes, but being able to rotate through them without a defensive drop could come in handy in the NCAA tournament:
|Texas A&M||Nathan Walkup||15.0||68.6%|
|Texas A&M||David Loubeau||31.0||59.8%|
|Texas A&M||Ray Turner||16.0||57.9%|
|Kansas St.||Curtis Kelly||17.5||54.5%|
|Texas A&M||Kourtney Robinson||14.0||49.9%|
Texas and Michigan, on the other hand, had great defensive guard play in their games against Kansas. Texas's performance in particular is impressive, because it comes over a two-game sample as opposed to one, and it even includes numbers from a loss, when the players likely weren't at their best.
|Kansas St||Martavious Irving||14||75.80%|
|Michigan||Tim Hardaway Jr.||38||71.80%|
As the next chart shows, three of the four worst defensive guards against Kansas were from Missouri. Once the Jayhawks were able to break the Tigers' pressure, they could do little to stop KU from scoring. Mizzou's first-round NCAA tournament opponent, Cincinnati, is only a middle-of-the-pack team in terms of holding onto the ball, though, so perhaps it won't be an issue.
|Kansas St||Jacob Pullen||35.5||44.30%|
|Texas A&M||B.J. Holmes||35||22.90%|
I'll close with one last note on charting: There's a good reason why individual defensive numbers like these aren't widely available. Compiling them is very time-intensive, at least compared to regular scorekeeping. And, perhaps more importantly, it's quite subjective. A guard gets beaten off the dribble and his man hits a floater over the outstretched arms of a helping big man. Who gets marked down for the opponent shot attempt? The guard who was at fault for the original defensive breakdown? Or the helper who switched to the newly free driver? It's a tough call, and two reasonable people might disagree. Dozens of plays like this happen every game, so having consistent numbers from play to play, night to night, and scorer to scorer would be, realistically, impossible.
That said, I don't think the subjective variation is large enough to distort Markieff Morris's Stop% advantage over Marcus. And if Kansas survives to cut down the nets in Houston, Markieff's defensive presence will surely have played a big part in that run.