A New Defensive Outlook

One subject I’ve seen addressed time and time again in the comments section is how to measure defensive ability. The topic arises shortly after I mention the three letters ‘P’, ‘E’, and ‘R’, which stand for John Hollinger’s Player Efficiency Rating. While Hollinger has done a great job in the first steps toward getting a player’s ability down into a single number, PER does a poor job in measuring a player’s total defensive contribution. Although it does account for blocks & steals, I think most of my readers are knowledgeable enough to know that this isn’t the best measurement of defensive ability. Bruce Bowen is the poster-child for PER’s failure. A uni-dimension offensive player (“go stand in the corner and only shoot if you’re wide open”) Bowen’s PER is usually among the league’s worst. However the Spurs’ guard-forward is considered among one of the best defenders at his spot, which some think overshadows his offensive liability.

This defensive conundrum is not necissarily caused by the game itself. While basketball does pose a challenge in assigning individual blame and credit, I’d argue that the primary fault is not the game, but rather how it’s being recorded on the stat sheet. In the Wizards/Bulls playoff series this last year, the NBA didn’t need to hire a host of experts to get better defensive stats. In fact they didn’t hire anyone. Kevin Broom, from the comfort of his own living room or maybe local bar, developed a simple method similar to keeping a regular box score, but on the defensive end. (BTW aspiring writers – if you want to get an article published in SportsIllustrated, that’s the way to go.). Since Roland Beech opened up 82games.com, there has been an increase in data available to the public to help come up with more information on a player’s defense.

This brings me to Dan Rosenbaum. APBRmetric members and 82games.com readers might recognize the name. Although the economics professor might be best known for his work on the monetary side of the NBA, he’s come up with something interesting: defensive ratings . Dan isn’t revealing his exact equation yet, but he’s come up with a defensive metric using 82games.com’s data. He’s only given rankings on a few players, and initial results pass the “laugh test.” The method seems to favor neither shot blockers nor man to man defenders, as both Ben Wallace and Jason Collins are in the top 5 of centers. I have to agree with Collins’ selection, because just like Bruce Bowen someone with a 9PER doesn’t get 30 minutes a game without being good on the other end of the floor. Rosenbaum’s part I is a teaser, just givng us the best & worst big men, and leaving out his exact method. Hopefully with the next installment, we’ll get all the rest of the positions, and maybe the whole list.

Oh and welcome to the blogosphere, Dan. :-)

Liked it? Take a second to support Mike Kurylo on Patreon!

Mike Kurylo

Mike Kurylo is the founder and editor of KnickerBlogger.net. His book on the 2012 Knicks, "We’ll Always Have Linsanity," is on sale now. Follow him on twitter (@KnickerBlogger).

24 thoughts to “A New Defensive Outlook”

  1. Such irresponsible journalism. It was from the comfort of my basement — not my living room or from a bar. Sheesh!

    Incidentally, one of the more interesting things (to me) I learned from that defensive charting effort is the irrelevancy of box score matchup numbers. The theory behind saying — Larry Hughes is a SG and Ray Allen is a SG, therefore, if Allen shoots 7-21, Hughes played good defense — isn’t bad, but it doesn’t hold up. There’s so much switching and help, that the standard box scores doesn’t come close to telling the real story.

  2. There’s also the problem of who they match up with. For instance, Kirk Hinrich always takes the best perimeter defender to cover for Chris Duhon, so I’m guessing Kirk’s defensive stats may suffer because of that.

    I do remember that in last year’s prospectus, Hollingers’ defensive PER always noted this when his number didn’t line up with the perception of the corresponding player’s defensive ability.

    (note: haven’t read the rosenbaum piece, cursed work firewall)

  3. Hollinger spent more time defending his defensive PER last year then he did explaining its results, so I’m anxious to see what he comes up with this year. Overall, it’s a pretty good start.

    He did confirm what I had been thinking for the bulk of 2003-04, but was too scared to admit to anyone — that, shockingly, Jamaal Tinsley and Maurice Taylor were actually damn good defenders.

    Dan’s work has been damn good so far, but I’m still not convinced that Kevin’s work and the upcoming 82games chartings aren’t the way to go — for better or worse, you have to take EVERY SINGLE POSSESSION singularly, “did the defender do an excellent, adequate, or poor job on this particular trip down court?” Even the best defenders take plays off, but lapses and aberrant play evens out after a while.

  4. I hate to take away credit from my namesake and main man, but Dean Oliver developed the method, for the record.

    What I think is great about it for the person actually doing the recording is you pay attention to a lot of things defensively you’d otherwise miss. I found it very rewarding in that sense.

  5. Caveat – I’m just a basketball fan, but I found the above postings to be really quite informative. Something I’ve always been interested in seeing in terms of stats (off. and def.) that I haven’t seen yet is a more precise version of what Kelly said above regarding taking every play on its own. I’d like to see stats broken down per 48 game minutes. What I mean to say is on the Def. end which players consistently contests defensively and which players fade and at what points in the game. Also what’s the score? Is Bruce Bowen playing the same defense 10 points down in early in 3th Q as he is 10 points up middle 4th Q? I think that a per minute stat could be useful because so much of defense is energy, even more so for the wing guys. But remember I’m just a layman stat wise, is there something out there I’m missing to find info like this?

  6. Statistics can be deceptive in basketball. Stats don’t tell the whole story on how good a player is on defense. There isn’t stats on how well a player makes other adjust their shot, stop them from penetrating to the basket, defending screens, drawing charges etc..

  7. It’s worth pointing out, though I missed it in my first reply, that the biggest problem with the Dean Oliver charting method is accounting for defensive matchup.

    A Bowen-type defensive specialist will look worse than he really is because he’s continually going up against superstars. I never really found a good way to account for this numerically.

    Dan’s analysis doesn’t suffer from this issue.

  8. Kevin Pelton makes a good point — it is actually Dean’s idea that I used last season. I made some modifications to collect info I was interested in, but it’s Dean’s basic method.

    The matchup issue is an interesting issue. Jared Jeffries may be an example of that with the Wizards. His individual defensive numbers (http://draftcity.com/dstatsper48.php) weren’t all that great, but the coaches named him as one of their better defenders, and he usually guarded the opponents best offensive forward (either SF or PF).

    Tracking the numbers, though, made me understand a lot more about the nature of defensive teamwork and help defense. The concept of “matchup” ain’t what I’d thought it was. Yeah, Jeffries might be “matched up” on Paul Pierce, but a good defensive play could be inducing Pierce to drive into help. When Jeffries allows penetration into Haywood’s help, and Pierce misses as both challenge the shot — who gets credit? Well, both do, which is exactly the point. What if Pierce gets past Jeffries, Pierce sees Haywood and pulls up for a jumper. Does Haywood get credit? Jeffries?

    The whole concept of “matchup” isn’t quite how many of us (me included) have traditionally have thought of it. Don’t believe me — try tracking a few games. As KevinP mentioned, you’ll see all kinds of defensive details that you’d never noticed before. (BTW: Tracking works best if you have TiVo, or a recording you can rewind, play in slow-mo, etc.)

  9. I also agree with Kevin Pelton’s point about good defenders always getting superstars.

    Perhaps a possible way of accounting for it is to, in some way, incorporate the opposing player’s PPG and/or FG% into the statistics, although I’m not sure how exactly this would work.

    It was funny to read someone, in reply to the SF list, explain that Trevor Ariza’s defense was overrated, using the example that Kobe Bryant was scoring on him at will. Last time I checked, Kobe Bryant pretty much scores at will against everybody, so I’m not sure that that is an indication that Trevor isn’t a good defender.

    Long story short, the best way to judge a defender is, as it always has been, is by good old common sense. Find someone who watches a lot of a players games, that has reliable opinion about basketball, and simply ask if they believe that he’s a good defender.

    It’s fairly apparent that Marbury and Crawford aren’t very good defenders, even if the numbers may not reflect that. Likewise, Trevor Ariza is not the 3rd best defensive SF in the league, but he is pretty obviously a pretty good defender.

    Your brain can take in a lot more information than you can quantify, which is why you end up knowing things that you may not be able to explain with statistics.

  10. The Ariza example shows that the brain can hold a great deal less, if it so chooses. Will that person ever be convinced his one-game opinion is wrong?

    The bigger problem is most of us don’t get to see players play, particularly in person, that often. When Antonio Daniels came to the Sonics, I thought he was a good defender because this was his reputation. It was quickly obvious this reputation was wrong.

    How many players around the league do I believe to be good or bad defenders that my opinion would change with regular observation?

  11. Just want to let everyone know that *I* was the one that wrote the Ariza line. While I find Trevor to be a good team defender, he’s generally lacking in the man to man dept. The Kobe game was one that stuck out to me, because he just got torched that day. It might be this game since it was later in the season & I remember Ariza getting yanked due to his bad D (he only played 10 mins & hit 3 of 3 – so you know it wasn’t his shooting).

  12. The brain can hold a tremendous amount of data, but if you’re not watching every game — or at least a TON of games — you may not have enough information. That’s one value of stats — the numbers are a record of every game. The interpretation of those numbers can be difficult at times, but that doesn’t diminish the value of statistical analysis.

    Also, something the brain does, is to focus attention on “standout” performances. Ariza’s bad defensive game may leap to mind, but perhaps there were several more solid (but unspectacular) performances in which his defense really helped the team.

    The brain sometimes can be guilty of violating one of my favorite John Wooden maxims — confusing activity with achievement. We may look at a guy chasing after loose balls, committing hard fouls, diving on the floor, generally putting on a great display of hustle. Typically, that kind of player is a fan favorite — whether he’s actually effective or not.

    Case in point — Etan Thomas vs. Brendan Haywood. Etan has the chiseled body, the dreadlocks, the headbands around the biceps. Etan fouls hard, glowers at opponents, shakes the rim when he gets a dunk. Haywood is kinda long and awkward looking — all elbows and stiff knees and herky-jerky movements.

    Between the two — Haywood is by FAR the more effective. Especially on defense. Tracking the games last season, I could see why. For all his show, Etan didn’t challenge as many shots as Haywood — basically, if he didn’t think he could block it, he didn’t challenge. He just did this sort of flex & glare thing and the opponent got an open shot. Whereas, Haywood would run at the shooter and get a hand up.

    The result? Opposing players had an efg% of .328 against Haywood and .494 vs. Etan. And Haywood forced about 2.8 turnovers per 48 minutes (steals + non-steal forced turnovers) to 1.4 for Etan. And Haywood fouled less — even while taking responsibility for 20.7 shots per 48 minutes to Etan’s 16.9. Etan’s display of activity didn’t come close to matching Haywood’s goofy-looking achievement.

  13. Just to see what everyones opinions are:
    Good Defenders
    1. Dwayne Wade( looks almost lackadaiscal, but is awesomely effective)
    2. Eddie Jones
    3. Kirk Hinrich
    4. Andres Nocioni(tenacious)
    5. Andre Iguodala(always a hand in the face)
    6. Luke Ridnour(quick)
    7. Chris Duhon(plays basketball the right way)
    8. Lindsey Hunter
    Bad Defenders
    1. Ben Gordon(because of size)
    2.LeBron James (unfair to put him here but I am because his defense doesnt comeclose to the other facets of his game yet)
    3. Amare Stoudemire ( looks like he doesnt care)
    4. Damon Jones (doesnt have physical tools)
    5. Darius Miles
    6. Tracy McGrady
    7. Erick Dampier (all he does is foul)

  14. Andy, I’d mainly agree with the good defenders list, although I think Hunter and Eddie Jones are getting old. I’m also not sure that Ridnour is that good but he isn’t bad either. The lesson we can learn from him is to never listen to anything Jay Bilas has to say. The bad defenders list is interesting in that Gordon and Miles both showed up in Rosenbaum’s list of the BEST defenders. I think that’s a massive fluke in Gordon’s case and a little less so in Miles’. The main problem I have with defensive stats is that perimeter defenders are very reliant on the big men behind them. Kevin Pelton mentioned Antonio Daniels as not being as good as his reputation. It’s true that according to basketball-reference his defensive ratings have been terrible in Seattle. However in San Antonio he always had above average defensive ratings. How much of that was due to Duncan and David Robinson? I guess I’m trying to say that good defensive big men are very important and that you can hide perimiter defensive liabilities with them.

  15. Eddie, I would say the reason for the change in Daniels’ Dean Oliver Defensive Ratings is largely attributable to the quality of the team defense, which is factored in. (Which is, yes, the work of Duncan and Robinson but is more a natural result of how the statistic is calculated than a statement on how Daniels’ value changed.)

    One of the interesting things to defensive reputation to me is it seems stars’ reps are always amplified. Players like Kobe, T-Mac and LeBron are occasionally considered quite good and more often (at least in the latter two cases) quite awful. I think the truth lies somewhere near the middle.

  16. I am not totally positive how those defensive ratings that Dan Rosenbaum has are calculated, but I have an idea why Ben Gordon is rated so high. Because he is percieved as a bad defender, and that is a rarity on the Chicago Bulls, I have noticed that some teams will just give the ball to the player whom Gordon is guarding and let him go to work. We all know that this is not an effective way to score points. So what I am saying is that its easier for the Bulls to keep that guy from scoring because they can give help to Gordon is his man gets around him. Basically, the team is taken out of there offense while Ben Gordon is in the game in order to take advantage of Big Ben’s perceived lack of defense.

  17. I’m of mixed opinion on Nocioni. The Bulls were much worse defensively when Nocioni was on the floor last season, despite all his tenacity, activity and effort. And, Nocioni ranks near the bottom in Dan’s defensive +/- list. In the playoffs against the Wizards (I tracked individual defensive stats for both teams during the series), I found that Nocioni performed pretty well. He’s a player I’ll be watching. If the reputation is accurate, I’d expect that the on/off numbers should start to reverse.

  18. As a Bulls fan, I’m a little nervous about Hollinger’s impending breakdown of Noc’s game. All the advanced stats for morons like me (who have been APBR freaks for a few years but still need to have their hand held during stat breakdowns) like +/- and PER suggest an nasty, nasty scouting report. And because the mainstream press gave him a great deal of pub due to his “hustling” mindset, John may negatively (for better or worse) react to that.

    For the most part, he could be downright horrible for the Bulls last year — but a lot of that can be attributed to the mindset of a 25-year old in a new country who was usually playing on about 30 minutes sleep.

  19. Kevin Broom sent me his defensive boxscore data from this year and there are some interesting patterns in it. One of my goals with it is to identify whether schemen or talent is more important in developing a defense. There are definite indications in the data that, as we’ve suspected for years, centers are the most important guys to a defense…

    Anyway, the value in that is that evaluating player defense may not be as important as how that player’s defense gets used. I’ve seen this from a bigger view with Joe Dumars, whose defense was completely ineffective after the Bad Boys left. Bowen’s defense can also be useless without the Spurs doing a good job filling in around it.

  20. Mr. Oliver, in your book, you talked about the “Project Defensive Score Sheet” you were working on in the WNBA. Have you tried bringing that project over to the NBA? Or is that basically the same thing that 82games.com is planning to work on this year?

  21. Dean’s Project Defensive Scoresheet was the basis for what I was doing on the Wizards last season. I did some things differently (collecting more detailed information, as well as some other categories of interest), but it was that same basic approach.

    The 82games project is going to be even more extreme.

  22. Project Defensive Scoresheet needs manpower, that’s for sure. The project that 82games is working on is going to cover many elements of Project DScoresheet, but also doing more. It should be a valuable data set, though understanding of it, assuming all goes well, will take years. It’s like unraveling the human genome — once you’ve got it, what does it all mean?

    I’ve worked with Kevin Broom’s data from Washington and it does provide a bunch of very useful insights about _how_ to play guys and where their strengths are. But really that was the second step of many that have to be done.

    One thing that I think we can do even with a second year of data (for the same team) is assess the variability in player ability based on how they get used. With so many changes in the Wizards’ team, we should see some changes in how the defense works. Maybe it will be Kevin’s data, but hopefully it will be 82games that provides more of that insight.

Comments are closed.