Announcement

Collapse
No announcement yet.

Our Clutch Pace

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Our Clutch Pace

    First of all, big warning here: this analysis uses some pretty small sample sizes. Extrapolate at your own risk. I'm putting in sample-sizes throughout so you can get an idea of just how reliable or unreliable the sizes are.

    Second caveat: I'm pulling stats from stats.nba.com, where you can choose from 'ahead and tied' or 'trailing and tied'. This means that possessions in which we're tied will count in both categories. This is why we've played 42 total clutch minutes, but if you add up the leading and trailing categories, you get 50 minutes


    So, looking over some stats, I noticed something interesting in the pace we play at in the clutch: in games we lost that had clutch minutes (total sample size 11 minutes), we played at a pace of 112.98 (second fastest in the league). In games that we won that had clutch minutes (total sample size 30 minutes), we played at a pace of 89.78 (fifth slowest in the league). Now, there's obviously a reason why that is: in clutch minutes, the trailing team is going to want to speed up the pace, while the leading team is going to want to slow things down (with a few exceptions like Golden State or San Antonio). But the only other team with as stark a difference in their win/loss pace was Memphis. So I wanted to break it down and see if there was more to be learned from clutch pace.

    Status Minutes Pace Offrtg Defrtg
    Trailing, and won 16 88.12 136.8 81.9
    Leading, and won 20 90.61 142.2 81.9
    Trailing, and lost 9 118.22 90.6 142.0
    Leading, and lost 4 98.0 43.6 112.8
    In all wins 30 89.78 143.9 84
    In all losses 11 112.98 91.5 136.9
    All leading 25 91.90 127.3 88.7
    All trailing 25 99.35 117.8 110.4
    So, that's the raw data. What's impressive is how good we are at slowing the game down to a crawl when we're leading, and then having a great offensive rating and defensive rating. The only team with a higher offensive rating in the clutch when leading is GS, who play at an insane pace of 121.91 when leading in clutch minutes. Our defensive rating when leading is also top 10 in the league.

    What's unimpressive is how poorly we play when we play at a faster pace in the clutch. Even when trailing, if we keep the pace reasonable we can convert at a high rate, and turn potential losses into wins. When we allow the game to unfold at a faster rate, we've turned wins into losses.

    Our ideal clutch speed is around 90. That means 90 possessions per 48 minutes, so in the last 5 minutes of a game, our ideal speed is going to give us (and our opposition) 9.375 possessions... let's say 9. In 9 possessions, at such a pace and with an offensive rating of 130 (which seems to be around a typical scenario when we run that sort of pace in clutch), we should score about 12 points. Similarly, if we assume a clutch defensive rating of 85, (again, seems to be a typical scenario), that would give the other team 8. Now, a differential of 4 over 5 minutes might not sound like a lot, but when you consider that this is a) this is a consistent sample over 25 minutes of playing time, and b) if you're trailing, you need to at the very least get the margin down to one possession. So in other words, a 7 point margin with 5 minutes left should be no reason to panic or change their pace.

    But if you take our typical trailing pace of about 100 and the corresponding rate 117, that only buys us at most 2 extra possessions, and with the lower conversion rate we'll produce on average only 1 more point. Meanwhile, we're giving up an additional 4 points over those 5 minutes by allowing the other team more possessions and a higher conversion rate.
    To me it's a little reminiscent of a football team that has a fantastic running game, but when they fall behind in the fourth quarter, they sometimes go away from the running game because they feel a sense of urgency that they don't necessarily need to feel. We're grinders.

    I don't think there's anything I've said here that really departs from the eye-test that much. It just reinforces that we are a good clutch team because we slow things down and grind them out. Last night was a microcosm of that. In the last 5 minutes, we had a lot of 10-second offensive possessions, and overall, we did not produce many points on them (we were +1 total). In overtime, the first 3 baskets that we scored averaged 18 seconds into the shot clock, gave us a 4 point lead, and from then on we were in our comfort zone. We were +5 over those 5 minutes.

    Edit just one other thing to add on this: I was curious if we use more isolation plays in the clutch when we slow it down. But apparently not: in the 'loss' sample where we ran at a pace of 113, we had an AST Ratio of 6.7. In the 'win' sample where we ran at a pace of 90, we had an AST Ratio of 19.3. Despite our perception of our own tendencies for Iso ball late in games, we have the 8th overall clutch assist ratio, and the 4th overall clutch assist/turnover ratio.
    Last edited by octothorp; Tue Dec 9, 2014, 06:49 PM.

  • #2
    Interesting.

    Comment


    • #3
      It fits the eye-test, that the Raptors' roster is built-for a half-court team-based approach, working the ball around to get a good/open look, or a chance to attack the basket off the dribble. When shots are rushed and/or forced, the Raptors don't score nearly as efficiently and their defense doesn't have adequate time to get setup. This team should only look for opportunistic fast-break attacks, rather than forcing a more up-tempo game in general.

      Comment


      • #4
        Good work. One point: I think you'll find there's some serious sample size issues going on in this when you break the minutes down into so many categories.
        @EdTubb - edwardtubb at gmail

        Comment


        • #5
          TeamEd wrote: View Post
          Good work. One point: I think you'll find there's some serious sample size issues going on in this when you break the minutes down into so many categories.
          Clutch stats in general are almost useless because of the tiny sample. Still an interesting study, just hard to draw any conclusions considering the data is basically just noise.
          twitter.com/dhackett1565

          Comment


          • #6
            DanH wrote: View Post
            Clutch stats in general are almost useless because of the tiny sample. Still an interesting study, just hard to draw any conclusions considering the data is basically just noise.
            Obviously, I should've read the very first line of the OP.
            @EdTubb - edwardtubb at gmail

            Comment

            Working...
            X