A brief observation on own vs. opponent turnovers

The Denver Nuggets rolled to a big win last night against the Los Angeles Clippers (see Charlie’s Rapid Reaction grades here). In that game, both teams did a great job of keeping control of the ball, each limiting their total turnovers to 11 apiece.

It’s well known among Nuggets fans that turning over the ball too much has been one of Denver’s primary weaknesses this season. In January, Matt provided some great in-depth analysis on the turnover problem, and if you haven’t read his breakdown, I’d highly recommend you check it out.

Regarding the Clippers game in particular, in looking over the box score afterwards I was struck by the even, equally low turnover numbers for both teams. What jumped out at first, given the pervasive and often discussed narrative that the Nuggets turn the ball over too much, was that Denver did a nice job of hanging onto the rock.

But the fact that the Nuggets were able to beat the Clippers so handily despite L.A. also having a low turnover count got me to wondering: Which is more harmful to the Nuggets, turning the ball over too often or failing to force opponent turnovers? Or conversely, which helps them more, keeping control of their own possessions or taking possessions away from the other team?

The answer turns out to be somewhat surprising.

In the 2012-13 season, the median number of both the Nuggets’ own and their oppoents’ turnovers is 15 per game. To create a clean separation between high and low turnover games, I removed the “soft in the middle” data for 15 own or oppenent TOs, and compiled the win/loss records for 16 or more, and for 14 or fewer turnovers by the Nuggets or by their opposition.

As it turns out, there is a significantly higher correlation between opponent turnovers and Denver wins than there is between the Nuggets’ own turnovers.

Here are the winning percentages for this season when the Nuggets or their opponents rack up more or fewer turnovers than the median, and the differences between those percentages and Denver’s current record:

As we should expect, when Denver has 14 or fewer, and when their opponents have 16 or more turnovers, the Nuggets tend to win more often. And vice-versa.

But the notable fact here is that the correlation is more robust when Denver’s opponents are turning over the ball more than when the Nuggets keep hold of it. And that pattern of opponent turnovers having a greater correlation with Denver’s win percentages holds on the flip side as well. The Nuggets lose more games when failing to force opponent turnovers than they do when turning the ball over too often themselves.

I’m not sure how much of a takeaway we can cull from this. The fact remains that the Nuggets hurt themselves by turning the ball over too often. But they hurt themselves worse when failing to force opponent turnovers.

Perhaps the lesson here is that there is a method to Karl’s madness – that all the incessant switching and risk taking on defense, costly as it may be at times, pays high dividends.

Or maybe, just as fans viewing the games, we should commit ourselves to facepalming eually as hard when the Nuggets miss a chance to force a turnover as we do when they cough it up.

Regardless of the conclusions we should draw, however, one thing is clear (as mundane and obvious a point as it may be): With winning percentages like .750 when the Nuggets limit their own turnovers, and .815 when they force a lot on the other end, controlling possession of the ball is critical to the success of this team.

 UPDATE:

I’d like to go into some detail in answering the question RMC reader Zach asked in the comments below:

Doesn’t this just suggest that the data to look at is turnover margin. I mean- if the nuggets average 15 turnovers per game, then, on average (not necessarily always), forcing 16 or more turnovers indicates that they won the turnover battle. The other figures do not correlate strongly in any direction- but I would guess that looking at TO +/- would probably clear up that picture.

So, in approaching this post I was starting with the question, “Which helps or harms the Nuggets the most, turning it over themselves, not turning it over, forcing opponent turnovers, or not forcing opponent turnovers?” I’m not sure how that question could be answered using turnover differential, since it crunches own and opponent turnovers into a single figure. In other words, after calculating the differential, there aren’t multiple data points to compare against each other. Unless I’m missing something, I don’t see how that could be done.

I would definitely agree, though, that in a broader sense, if you want to look at the overall turnover picture of a team, the differential probably gives a more accurate picture, much like point differential gives a more accurate representation of how effectively a team is playing than either points scored or points against alone can do.

But for what it’s worth, here is how Nuggets wins and losses have broken down along the lines of turnover differential this season:

  • The Nuggets have won the turnover differential in 26 games this season. In those games they went 21-5 (.808).
  • They’ve come out even (no differential) six times and gone 5-1 in those games (.833).
  • In the 31 games they’ve lost the TO diff, they’ve gone 15-16 (.484).
  • But interestingly, if you break down those 31 positive TO diff games further (positive meaning bad, ie. on the plus side in turnovers), they’ve played 17 games with a TO diff between 1 and 3. In those, they’ve gone 12-5 (.706).
  • Thus, it’s really only when the TO diff has been strongly positive (+5 or more — again that being bad) that the results really turn south. In those 14 games, they’ve gone just 3-11 (.214).

So it appears that the Nuggets win at a high percentage not only just when they win the turnover differential, but even when they just keep it neutral, or even lose it but keep it close. And it’s really only when the turnover differential is in the opposing team’s favor by a large margin that we see a strong correlation with losing results.

Follow me on Twitter

 

 

The following two tabs change content below.
Joel is a long time Denver Nuggets (and Broncos) fan from Colorado who's been living in Japan since the mid-90s, and blogging about the Nuggets since 2008. You can contact and follow him on Twitter: @denbutsu.
  • diehardnr1

    Bravo! Excellent breakdown

  • Legalize Denver nuggets

    Maybe do some kind of regression analysis and include pace of game? I found this piece very interesting and would love to see what else is going on when we are forcing turn overs. I mean, obviously better defense- steals, blocked shots, ect. But I’m curious if we are forcing more turn overs the faster we play, or if that causes us to turn it over more, or both. While there would still be the bird and egg question (did a fast pace create turn overs, or did turn overs create a fast pass), I still would be interested to see how pace influences this. My guess is that a higher pace would be indicative of more turn overs and forced turn overs and also winning. I’m no statistician, but seeing the interplay here might be interesting esp if my hypothesis is wrong in some way.

    • dynamo.joe

      Fish lay eggs, so there were eggs millions of years before there were birds, QED the egg came 1st.

      • Legalize Denver Nuggets

        Cite your sources!

        • dynamo.joe

          Now that’s funny!

    • https://twitter.com/denbutsu Joel

      Those are definitely questions worth exploring. Obviously, I was not attempting here to address anything about the cause of own/opponent turnovers. But your idea that Denver might come out better on the turnover front when playing at a faster pace seems sensible to me for the reasons you said, and also just because, generally speaking, when the Nuggets play faster they tend to play better overall.

  • will

    A fast team can punish opponent’s turnovers while defending against their own turnovers relatively well. I mean the Nuggets have Ty Lawson running the floor. Games with lots of turnovers are won by fast break teams. Seems intuitive to me.

  • Chris

    I’ve followed the Nuggets ever since the Nene/Skita draft and I have to say that this team is the best I’ve seen. I just love the energy this team shows as they play team b-ball. After watching the East Coast Nuggets pull defeat from the jaws of victory last night, it was so refreshing to see a Denver team (with the exception of Dre) that doesn’t play selfish ball, or whine and pout when something goes against us.

  • Chris

    Second, I felt like the Nuggets were a very good team before Chandler came back. With Chandler, I think we are an elite, title contender. I love seeing George playing different combinations of Chandler, Iggy and Brewer. Its almost guaranteed that two of those guys are playing at a really high level, usually helping the Nuggets break the game wide open.

  • Zach

    Doesn’t this just suggest that the data to look at is turnover margin. I mean- if the nuggets average 15 turnovers per game, then, on average (not necessarily always), forcing 16 or more turnovers indicates that they won the turnover battle. The other figures do not correlate strongly in any direction- but I would guess that looking at TO +/- would probably clear up that picture. A turnover is essentially equivalent to a shot for the team that gains possession, so, whoever wins the turnover battle is likely to have an advantage in shots attempted. Having more scoring chances than your opponent certainly is not a bad thing.

  • Aaron

    This means a TON. Because what is totally underreported in our current win streak is the defense we’re playing. It doesn’t show up in total points due to our fast pace. The Sacramento game comes to mind. . they had 113. . but if anyone doesn’t think we were playing amazing D in that game. . well.

    If we play D like we did last night and in the last month, we will beat Memphis, period. I”m a little worried about Gasol and Zbo getting Kofous and Faried into early foul trouble. I think we’re going to see Mozzy in that first round series regardless. . he’ll probably be our best defender against Marc Gasol. If we can rotate better to Conley, Prince, etc. . and get some steals, I actually don’t think that series will be close.

  • theo

    Nice stats, Joel. Here’s my interpretation. Definitely a method to Karl’s ‘madness’ defensively. Key facts: 1. Ig is a great fundamental defender, Chandler very good, Gallo and Kosta solid. Outside of that our roster is full of pretty fundamentally flawed defensive players ranging from mediocre to really bad. 2. Is spite of that we have a ton of very athletic players 3. We have to keep the tempo very high to succeed.

    I think Karl is running a defensive scheme that makes sense in light of those three facts. With so many relatively poor conventional defenders on the court we’re forced to switch a ton in a way teams with lots of fundamentally solid defenders don’t have to. Taking it a step further and turning a weakness into a strength, the Nugs use their superior athleticism to gamble and trap and jump into passing lanes to create havoc and turnovers. Again, a less athletic and more fundamentally sound defensive team wouldn’t trap and gamble or rely on turnovers nearly as much as the Nugs do because they wouldn’t have to and because they couldn’t athletically pull it off. It’s a double edged sword and we get exposed a lot, but as your numbers show, it actually works pretty effectively. A stat you left out that would be interesting is how efficient the Nugs are in turning those oppo turnovers into points–I’d guess probably very efficient. In light of fact 3–we’ve got to keep the pace fast–it makes sense we’d play more frenetically on D. I mean, one way to deal with a team of poor fundamental man to man defenders would be to play zone a lot, but that would slow the game in a way that would undermine the relentless attack the cup philosophy. So, at least in my mind, it makes sense. I haven’t had the strongly negative reaction to our defensive approach that a lot of fans do, though I totally understand why people would. It doesn’t look like conventional, solid, half court
    D, and it isn’t. But we don’t really have the personnel right now to play that way like, for example, the Bulls or Spurs do.

  • dynamo.joe

    Still working on 5 man lineup data. I revised my coach grading system. here are the results.

    Tot cap ___ team ___
    564.05 ___ okc ___ ave
    474.23 ___ mia ___ 55.13
    399.78 ___ lac ___
    339.43 ___ ind ___ stdev
    285.98 ___ den ___ 280.45
    273.86 ___ san ___
    254.06 ___ hou ___ A=475+
    226.42 ___ mem ___ B=195 to 475
    206.05 ___ atl ___ C=-85 to 195
    189.59 ___ Bos ___ D= -365 to -85
    187.45 ___ Brk ___ F= < -365
    168.78 ___ nyk ___
    165.26 ___ mil ___
    118.42 ___ tor ___
    89.21 ___ gsw ___
    79.70 ___ chi ___
    79.45 ___ lal ___
    -27.48 ___ noh ___
    -46.18 ___ phi ___
    -64.09 ___ det ___
    -69.10 ___ min ___
    -70.56 ___ was ___
    -73.95 ___ por ___
    -75.37 ___ uta ___
    -83.55 ___ orl ___
    -139.89 ___ sac ___
    -251.50 ___ dal ___
    -304.02 ___ cle ___
    -402.07 ___ pho ___
    -840.00 ___ cha ___

    • dynamo.joe

      So, what I did here is divide the +/- rate of each lineup by the +/- rate of the best lineup for that team. Then I multiply that by the minutes played by that lineup. Finally I sum over all 20 lineups for each team.

      So, you might not be playing your best lineup, but if the line up you are playing is 90% as effective as your best you still get 90% credit, rather than 1/2 or 1/3, etc.

      Also, if you play losing lineups, that counts against you, hence the negative scores.

      Grading on a curve because reasonable estimates for the maximum you could achieve aren’t so easy. You can’t just play your best lineup 48 minutes a game.

    • dynamo.joe

      According to this, measure (which I think is probably a pretty good one) GK is the 5th best coach in the NBA and just noses ahead of Pop.

      I can’t believe anyone is as bad as Charlotte, they must have been instructed to tank for all the balls.

      The top coached teams are basically your playoff teams, with the exception of Toronto who continually get screwed by inept front office work.

      • https://twitter.com/denbutsu Joel

        Joe, that is really interesting work. But I have to wonder: How much of that is coaching (ie. choosing the best lineups) and how much of it is roster talent (ie. there being a greater likelihood of good lineups being put on the floor simply because the team is stacked with better players). It seems like you’ve got a pretty complicated formula going on already. But do you think there might be some way to control for the quality of the roster? Maybe some kind of handicap based on how many minutes players of x PER are on the court, or something like that?

        Because when OKC is first and MIA is second, it would seem to me that the simpler explanation might be that they both have duos including one of the two best players in the league and another top 10 player. Not to take anything away from Brooks and Spo, but I think you know what I mean.

        • dynamo.joe

          I see your concern, but I think that is already included, at least to some extent. I am not comparing the lineups that Charlotte is playing to the lineups that Miami is playing. I am comparing the lineup that Charlotte plays the most to the lineup that Charlotte has played with the highest +/- per minute.

          As an example, the lineup Charlotte uses the most is Walker-Taylor-Kidd-Gilchrist-Mullens-Haywood with a +/- per 48 of -1.16. The most productive lineup that the kitties have put on the floor is Walker-Henderson-Kidd-Gilchrist-Mullens-Haywood, just a switch of Taylor for Henderson, witha +/- per 48 of 10.00. So I am not comparing it to what would happen if they traded for Lebron, just what would happen if Henderson ran with that team rather than Taylor.

          Advanced stats says Henderson sucks, but Taylor really, really sucks, so it makes sense that this would be an upgrade.

          Less sensical is that Haywood and Mullens are on good lineups. But one is a good defensive rebounder and the other is a good offensive rebounder. Neither can score. Haywood knows it and doesn’t shoot. Mullens hasn’t figured it out and shoots too much. Maybe on that lineup, they let the other 3 score and concentrate on rebounding?

          Anyway, here is the real point of the grading. It’s most productive lineup plays the the 16th most minutes. It’s second most productive lineup plays the 14th most minutes. It’s 3rd best lineup plays the 10th most minutes. As a coach, it’s your job to find the best lineups and then play them (unless MJ instructed them to tank), well they found them, but they don’t play them. That’s why they get the bad grade.

    • dynamo.joe

      Probably try to evaluate front office next.

      • theo

        Really nice stuff, joe. Been interested in the way Karl uses our various 5 man units, and from the numbers mostly looked like he was doing a very effective job with our rotations, but concerned that he was giving too many minutes to a couple of our really bad combos. Had no way to compare it to other coaches, though. Thanks.

        Interested to read a take by a Miami hoops writer the other day that evaluated the Nugs very positively. One of the advantages he said we had was ‘a top 5 in the league’ head coach–not surprising since coaches and players around the league make those kinds of comments about Karl all the time. Guess your stats tend to back that up. I have mixed feelings about Karl, but the dude gets a ton of flak among Nugs fans, no doubt. Guess you can be a prophet everywhere but in your own country….

  • Mike

    Interesting read. I hate to be a pessimist but I feel like the takeaway from this analysis is bad news for the Nuggets come playoff time. If limiting their own turnovers was more important to the team’s success, it seems like that would be an easier problem to solve. You make that a point of emphasis going forward, make sure the ball is moving through your primary ball handlers as much as possible, etc. Basically just get a little more conservative when the stakes get raised.

    But this suggests they will and should do the opposite, because they NEED the other team to commit those turnovers. The Nuggets rely on their opponents making mistakes, and the playoffs are exactly when the fewest mistakes are made. Don’t expect 16 turnovers out of the Spurs in a WCF game.

  • Ckwizard

    Good analysis! Early in the year the Nuggets turned the ball over a lot and that resulted in the other team scoring easy baskets and made Denver appear to be a worse defensive team than they were in reality… People would look at the final score and see the opponent had scored over 100 and would criticize Denver’s defense when in reality it was bad Denver offense and costly turnovers that lead the other team to score high ad sometimes win. The question you pose is which is more influential in winning well…. Defense matters and it depends on if you see the Nuggets as a good defensive team because if they are good defensivly then their own turnovers are not as costly and when the opponent turns it over it is advantage Denver. If Denver was bad defensivly then they would lose a lot more often when they turned the ball over a lot…. Data supports Denver is a Decent if not good Defensive team Because they win more frequently when forcing opponents into a lot of turnovers but can still draw even when turning the ball over a lot themselves.

  • Cider

    I created a plot of turnover differential vs. point differential to see what kind of correlation there was. Here’s the plot, where x-axis is Denver turnovers – opponent turnovers and the y-axis is Denver points – opponent points.

    http://img41.imageshack.us/img41/4943/screenshot20130309at133.png

    There certainly appears to be some correlation, but based on the R^2 coefficient there isn’t a lot. This is certainly because there are such a large number of factors that go into determining a game, which means there can’t be any incredibly strong correlation in a situation like this. I’m not sure what type of threshold you’d use for your R^2 in such a scenario, though. Maybe someone else is more versed in stats here and can interpret these results better.

    • https://twitter.com/denbutsu Joel

      First, let me just say that your math is over my head, so I won’t claim to understand it. That out of the way, it seems clear to me that going 3-11 (.214) in their 14 games when the TO diff was +5 or more, but 38-11 (.776) when it was better than that, is a strong correlation. So maybe there’s some kind of uneven distribution of point differentials or something? It does seem kind of odd that the point differentials and win/loss results wouldn’t jibe.

      • Cider

        Well, mathematically you can only determine correlations between two variables when they take on a range of values, so two sets of data points don’t really tell you correlation or how correlated two things are.

        Another option would be to see correlation of turnovers vs win percentage, but there is obviously the difficulty of having either sparse data (not many games for certain turnover differences) or too few data points (where you combine 1-5 turnovers as a single set and so on). There might be a balance between these two to look at through this type of method, but I didn’t want to devote too much time when I have other things I probably should be doing.

    • dynamo.joe

      I think you hit it on the head there are a lot of factors that lead to point differential. TO differential, rebound differential, 2FG%, 3FG%.

  • Giovanni

    Damn the grizzlies are on fire…
    Won 11 of last 12…

  • heykyleinsf

    You guys are giving things an extra look…
    amazing research. Good job!

    Just some observations from Clippers game..

    The Nuggets run a break neck pace…
    live and die by the sword.

    The turnovers I saw were more hasty
    rather than FB.. just sloppy..
    maybe in a hurry.. but not fast break
    as I saw it.
    E.g.. Gallo dribbling off his shoe
    Manimal sloppy pass to Miller