Non-Steal “Forced” Turnovers

A thread titled “Unrecorded Useful Stats” came up a few weeks back over at APBRmetrics. In it, Bob Chaikin (whom I believe works/consults for the Heat now) responded:

non-steal forced turnovers. any time a player commits a turnover that is not credited as a steal to a defender, credit a defender for good defense with a non-steal forced turnover….

surely there will be turnovers with no obvious defender deserving credit. but you count enough of these as credited to defenders and we’ll get a good idea of who is making an impact on defense, similar to how we count blocks or charges drawn…

right now milwaukee is best in the league with 9.0 non-ST forced TO/100poss, the lakers the worst with just 6.7. but i wouldn’t know which players on the bucks to give the credit to…

I replied:

My PBP code records these as “team turnovers”, which I split credit among teammates. With a little bit more effort I could keep track of counterparts for these.

Happy to say, with just a few extra lines of code, these “non-steal” turnovers were a cinch to tabulate. To answer Bob’s question, those players would be Keyon Dooling (#6), Brandon Jennings (#18), Andrew Bogut (#19), Ersan Ilyasova (#32), and Luc Mbah a Moute (#42), all of whom appear in the top 50. In the next post, I will see how this would affect ezPM stats. Here are the top 50 “pseudo-thiefs” (I was kind of shocked to see #3):

1 Jose Barea DAL 1546 37 2.39
2 JaVale McGee WAS 2099 50 2.38
3 Vladimir Radmanovic GSW 1147 27 2.35
4 Greg Monroe DET 1591 36 2.26
5 C.J. Watson CHI 1017 23 2.26
6 Keyon Dooling MIL 1703 38 2.23
7 Hedo Turkoglu PHX 1168 26 2.23
8 Zaza Pachulia ATL 1037 23 2.22
9 Rodney Stuckey DET 2005 44 2.19
10 Brandon Bass ORL 1740 38 2.18
11 Beno Udrih SAC 2670 57 2.13
12 Anderson Varejao CLE 1838 39 2.12
13 Andrea Bargnani TOR 2584 53 2.05
14 Emeka Okafor NOH 2555 51 2.00
15 Mike Conley MEM 2967 59 1.99
16 Hakim Warrick PHX 1459 29 1.99
17 Marquis Daniels BOS 1362 27 1.98
18 Brandon Jennings MIL 1670 33 1.98
19 Andrew Bogut MIL 2246 44 1.96
20 Marcus Camby POR 1997 39 1.95
21 D.J. Augustin CHA 2461 48 1.95
22 T.J. Ford IND 1344 26 1.93
23 Nate Robinson BOS 1441 27 1.87
24 Jameer Nelson ORL 1874 35 1.87
25 Steve Nash PHX 2419 45 1.86
26 Zach Randolph MEM 2767 51 1.84
27 C.J. Miles UTA 1682 31 1.84
28 Ben Wallace DET 1575 29 1.84
29 Josh Smith ATL 2622 48 1.83
30 Channing Frye PHX 2462 45 1.83
31 Brad Miller HOU 1260 23 1.83
32 Ersan Ilyasova MIL 1866 34 1.82
33 Nazr Mohammed CHA 1107 20 1.81
34 Eric Maynor OKC 1115 20 1.79
35 Shaquille O’Neal BOS 1227 22 1.79
36 Ty Lawson DEN 2008 36 1.79
37 Ronny Turiaf NYK 1339 24 1.79
38 Kevin Love MIN 3093 55 1.78
39 Corey Brewer MIN 2028 36 1.78
40 Amare Stoudemire NYK 3109 55 1.77
41 Jason Richardson PHX 1529 27 1.77
42 Luc Mbah a Moute MIL 1768 31 1.75
43 Dante Cunningham POR 1543 27 1.75
44 James Jones MIA 1661 29 1.75
45 Zydrunas Ilgauskas MIA 1381 24 1.74
46 Jared Dudley PHX 1844 32 1.74
47 Jason Thompson SAC 1506 26 1.73
48 Joel Anthony MIA 1279 22 1.72
49 Spencer Hawes PHI 1588 27 1.70
50 Serge Ibaka OKC 2136 36 1.69

16 responses to “Non-Steal “Forced” Turnovers

  1. I’m looking at this list and it’s filled with players who have a reputation of playing poor defense. What’s the correlation there?

  2. Good question. Not sure there is a “correlation”. Maybe they are underrated on defense. (Although I would argue the list doesn’t lean towards “bad” or “good” defenders by reputation.) It would be interesting to look at their defensive APM. It could also be that these players somehow benefit from their teammates’ help defense more.

  3. Good thing to check.

    Just looking at the first dozen players, 5 were better than neutral on earlier multi-season RAPM from Joe Sill. 8 are above average on back2newbelf’s current 1.5 season RAPM. About half to 2/3rds good on this element are good defenders by RAPM, though a few may just be gamblers or being helped by others.

    If non-steal turnovers are assigned out to individual defenders (probably a pretty good thing, though maybe it could be 2/3rds to them or maybe just 50% and the rest divided among the others) then it increases the desirability of at least some counterpart shot defense in the formula. I know, patience.

    • Counterpart d is coming soon. I promise. :)

    • Counterpart D is done. Here’s a taste (top 20):
      1 Dwight Howard ORL 4.51
      2 Andrew Bogut MIL 3.59
      3 LeBron James MIA 3.32
      4 Tim Duncan SAS 3.24
      5 Rudy Gay MEM 2.95
      6 Marc Gasol MEM 2.77
      7 Josh Smith ATL 2.75
      8 Andrei Kirilenko UTA 2.47
      9 Trevor Ariza NOH 2.40
      10 Pau Gasol LAL 2.40
      11 Elton Brand PHI 2.32
      12 Luol Deng CHI 2.31
      13 Lamar Odom LAL 2.25
      14 Gerald Wallace CHA 2.18
      15 Andre Iguodala PHI 2.13
      16 Corey Brewer MIN 2.10
      17 Glen Davis BOS 2.04
      18 Manu Ginobili SAS 1.99
      19 Joe Johnson ATL 1.96
      20 Mike Dunleavy IND 1.95

      This is going to be interesting…

      • The only guards are at the very tail of the top 20 and play some SF too.

        Look forward to seeing the data but also hearing about how you did it.

  4. The % good on RAPM may or may not change if you looked at all 50.

    An article recently made a big deal about DIPS for baseball. I was thinking that current metrics either without any shot defense or with lineup or team level shot defense are somewhat similar. Steals and blocks are the strikeouts. Sooting fouls are usually home runs given up, non-shooting fouls could be considered the walks. The opponents shots are ignored or assigned to team defense. The claim is made for DIPS that this defense of balls in play is independent of the pitcher. I have been pushing for the inclusion of counterpart defense but is there a good enough argument to not include it in basketball? I don’t think so. The counterpart defender is the first main line of defense not just 1 of 9 and barely that. I think opponent scoring is more influenced by the counterpart defense. There are far more events per game. I’d say it is less random. Sure there is probably some bad luck with shots falling at a greater rate on some guys in the distribution than others applying roughly equal coverage but it still happened and it still hurt.

    • I was just talking with my brother about these results, and he had some brilliant insight (one of those “yeah, I should have thought of that” moments). His idea is that some of the bad defenders (say Vlad Rad) show up on this list because opponents “go at” them more, thus resulting in more turnovers, not necessarily because of stellar d, but just because of opportunities to get “owned”. A better metric – rather than per 100 possessions – would be per FGAA (field goal attempts against). Like its counterpart “usage” for offense, “ownage” on defense is something I’ve wanted to calculate for a long time. Soon….

  5. Opponents will “go at” at some weaker defenders more but it will not be a precise relationship as their are some imprecise player judgments about shot defense abilities of players resulting in some missed opportunities and some guys challenged more even though they are good. And course there is random element.

    Radmanovic is similar to almost all his PF alternatives with the Warriors at about 17 FGA per 48 minutes according to 82 games and it has been that way for several years. It looks like by hoopdata that is a little less than big minute PFs average but a little more than all PFs. Jeff Green though is over 21 FGA per 48 at PF.

    With Rad the issue is that the opponents shots go in at a quite high rate and thus are probably remembered more and may create the impression that there are more attempts relative to players when that does not appear to be the case.

  6. Clarification: Rad faces 17 opponent FGAs per 48 minutes. Green faces 21.

  7. Of course these 82 game rates are based on assumed counterparts matched up by height and some sense of how they are used and not accurate to videotape. I think Synergy has the opponent FGAs from tape.

  8. A better metric than per 100 possessions or per FGAA (field goal attempts against) would probably be opponent scoring possessions (including the turnovers) or even the full usage (including the assists).

    • agreed

      Going back to your other point, another distinction that I think may be important is the percentage of assisted field goals against. It could be that opponents iso more when going up against Rad, so those 17 FGA, could be more prone to turnover. Anyway, that may or may not be the case, but it’s something I could look at with the data I collect.

  9. Yeah assisted field goals against would be good to check and it doesn’t appear 82games is giving that away. Synergy has possessions faced split into several groups including a count of isos faced . Not sure if they have the assisted field goals against.

    Should the counterpart of the opponent who dealt the assist get a partial charge for some of its value when the counterpart shot defense rating is implemented? I’d say yes. Maybe half the value of the assist charged to both counterpart defenders?

  10. Pingback: More on Non-Steal Counterpart Turnovers | The City

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s