Statistical Analysis. Humor. Knicks.

Friday, August 22, 2014

Four Factors and Five-Man Units

Knickerblogger.net will be in full playoff mode until the 2007 champion is crowned. But before we let go of the Knicks’ forgettable season I wanted to take a brief look back through the window of five-man units. Because Thomas? played a lot of people a fairly high number of minutes I thought it would be worthwhile to look at how they fared on Dean Oliver?s four factors with help from our good friends at 82games.com. I?ll look at offense today and defense later.

I thought looking at five-man units might be particularly interesting for NY because so many different units played significant minutes. We generally expect starters to play the lion?s share and other units to play comparatively fewer minutes. That is, minutes tend to have a skewed distribution. Of course some teams play their reserves more than others but the top minute-getting units play a lot of minutes together. For example Chicago?s top unit played 618 minutes, Detroit?607, Cleveland?467, New Jersey?460, and Toronto?321. Miami?s opening night starting five (Williams-Wade-Kapono-Haslem-Mourning) played fewer than 160 minutes because of injuries, but its top unit (Williams-Jones-Posey-Haslem-O’Neal) still played 301 minutes. By contrast, New York?s top unit played only 192 minutes together.

Unfortunately, the raw data needed to calculate all four factors directly are not available by five-man unit. Shooting is available directly. I can also use net turnover percentage and net free-throw attempts to create a ?quick and dirty? picture of turnovers and free throws. I must leave rebounding out of this analysis however, because the available data doesn?t break it out into its offensive and defensive components.

Before getting to the factors, let?s take a quick look at the best and worst five-man units based solely on plus-minus (+/-).

The Best and Worst Five-Man Units

The 3 best:

# Minutes rank Unit Minutes +/-  
1 13 Robinson Francis Collins Rose Curry 58 33
2 5 Marbury Crawford Lee Frye Curry 137 26
3 20 Robinson Crawford Q-Rich Lee Frye 34 14

The 3 worst:

# Minutes rank Unit Minutes +/-
18 12 Marbury Crawford Q-Rich Curry James 62 -30
19 4 Marbury Crawford Jeffries Frye Curry 142 -38
20 3 Marbury Crawford Q-Rich Frye Curry 162 -68

Some of the heaviest-minute units were just awful. The Marbury Crawford Q-Rich Frye Curry unit was undeniably terrible, outscored by 68 points in 162 minutes, roughly -20 points per 48. Swapping in Jeffries for Q-Rich ?improves? this unit to just plain bad. But note, swapping David Lee into this unit for Q-Rich or Jeffries makes a substantial improvement. In comparable minutes Lee’s presence is the difference between the unit ranking 19th or 20th in +/- or fifth. He appears to really click with Marbury, Crawford, Frye, and Curry. His impact is not nearly as striking with any other mix of players. (Lee also plays on NY’s top minute-getting unit?Marbury Crawford Q-Rich Lee Curry?which ended up only -2 on the season.) Another interesting feature of these two tables is the surprisingly good +/- of the unit that ended the season: Robinson Francis Collins Rose and Curry.

The Four err… Three Factors

Top 5 shooting units:

# Minutes rank (minutes played) Unit Eff FG (eFG) Close%
1 10 (67) Marbury Crawford Jeffries Lee Curry 0.614 52
2 13 (58) Robinson Francis Collins Rose Curry 0.604 59
3 9 (78) Robinson Crawford Lee Q-Rich Curry 0.59 46
4 14 (54) Marbury Crawford Q-Rich Rose Curry 0.58 48
5 15 (47) Robinson Crawford Balkman Lee Curry 0.561 55

It is probably no surprise that the five best shooting units all include Eddy Curry. However it may surprise that only two of the best shooting units include Stephon Marbury while three include Nate Robinson. The little man quietly had himself a nice sophomore season?well maybe ?quietly? isn?t the right word, but you get where I?m going. For all the people trying to run him out of town for his immaturity, consider that Robinson was fairly efficient (39% 3-pt shooter, 55.3% TS) and one of the least turnover prone players on the roster despite some of his shenanigans.

Perhaps the most unpleasant surprise here is that none of the team?s top five units in minutes was among its five best shooting. Channing Frye?s sophomore-season-to-forget is certainly one culprit, though not the only one.

Marbury-Crawford-Q-Rich-Lee-Curry (.503 eFG/192 min)
Marbury-Francis-Q-Rich-Frye-Curry (.506 eFG /162 min)
Marbury-Crawford-Q-Rich-Frye-Curry (.460 eFG /162 min)
Marbury-Crawford-Jeffries-Frye-Curry (.519 eFG /142 min)
Marbury-Crawford-Lee-Frye-Curry (.509 eFG /137 min)

Top 5 turnover units:

# Minutes rank (minutes played) Unit Net TO% +/-
1 20 Robinson Crawford Q-Rich Lee Frye 5 14
2 11 Marbury Crawford Q-Rich Lee Frye 4 -7
3 18 Marbury Crawford Q-Rich Jeffries Curry 3 2
4 15 Robinson Crawford Balkman Lee Curry 2 8
5 6 Marbury Francis Jeffries Frye Curry 1 -17

Turnovers have plagued this team like it is stuck inside some sort of biblical curse plague dome. The table shows the only five units that managed to create more turnovers than they lost. I feel reasonably confident that none of these units created many turnovers but rather were better than others at taking care of the ball. Note again that only one of these units was in the top 10 in minutes played.

I also added in the +/- numbers for those units. Two of them still managed to be badly outscored despite holding onto the ball better than opponents. The worst unit on net turnover percentage was Robinson Collins Jeffries Frye Curry unit (-13 TO%). There appears to be no truth to the rumor that these players will petition David Stern to bring back the composite ball.

Top 5 free-throw units:

# Minutes rank (minutes played) Unit Net FT Attempts
1 5 (137) Marbury Crawford Lee Frye Curry 53
2 2 (162) Marbury Francis Q-Rich Frye Curry 35
3 7 (93) Marbury Q-Rich Jeffries Frye Curry 29
4 1 (192) Marbury Crawford Q-Rich Lee Curry 28
5 8 (88) Marbury Francis Q-Rich Lee Curry 27

Overall, we know the Knicks are good at getting to the free throw line. They rank seventh in FTAs per FG. For five-man units only net free-throw attempts is available. The Marbury Crawford Lee Frye Curry unit has a whopping 53-attempt advantage. That?s more than a 2.5 more attempts per minute. New York?s top 15 units all take more free-throws than their opponents. Last season only their top 11 units took more free throws.

8 comments on “Four Factors and Five-Man Units

  1. Owen

    Quite an interesting post. Dont know quite what to make of it all, beginning with the fact that Collins featured in the top unit. His stats were pretty terrible all year. There are 20,000 minutes in a season. What do 58 and 34 minutes of unit really playing time really telling us?

    I agree Robinson was better than Marbury and Crawford this year. You are exactly right about his TS%, and despite being 5’6 he managed to get an extra rebound while comitting 1.3-1.7 turnovers less per 40 than Marbury and Crawford.

    Isn’t it strange that Crawford leads the team in average shot attempts but has the lowest ts% amongst the top seven scorers?

    Also, what happens if you do the same trick with Curry as you do with Lee, re substituting in and out of lineups. Looking at 82games, Curry has always looked tremendously awful to me over at 82games, what does this data suggest about him?

  2. Michael Zannettis

    What I found interesting was how effective the Lee-Frye-Curry frontcourt was. Many posters had linked Frye’s struggles to Lee and Curry clouding the low-post basket area.

    I wasn’t entirely convinced that was his problem, considering these were the same teammates he had in his promising rookie year. In the end, I wonder what Frye plans to work on in the off-season to resucitate his career path…

  3. Caleb

    This is fun to look at but the sample sizes are so small (as you point out) that it’s hard to think they’re significant.

  4. Gerry

    Its very apparent that what the Knicks need to be a competetive and a playoff team next year is to overhaul its frontline and. After Curry, and if he has an offnight the team has no offensive threat in the paint. A season of experimentation is more than enough to conclude that the existing frontliners in Frye, Jeffries, Rose, James and Cato cannot contribute in both ends of the court. The team should get a certified power forward who has perimeter touch and a shootblocker. Lee and Balkman may not be that offensive threats, but at least both can be expected to contribute in defense. In addition, the team may let go of Robinson and Francis for Collins hasty development.

  5. KnickFan

    Hey David,

    Can you run a regression on this data, showing each player’s contribution to the plus/minus taking the other players on the floor into account? If you want to get really fancy, you could run interaction terms between every combination of players to see if that’s significant. My guess — Lee is the only statistically significant positive, and Frye, Curry and Francis are statistically significant negatives.

  6. dave crockett

    Owen and Caleb-

    I’m not sure what to make of it all either. But, after thinking about this quite a bit over the past couple days I might have written this piece differently (and I may still re-do it, time permitting).

    I would say that this unit level data is probably more usefully thought of as census data, where each unit is independent of the others rather than a sample of the universe of “all NY five-man units.” I say this b/c I think the more useful comparisons are ultimately between, for example, NY’s 10th-ranked unit and the 10th ranked units on other teams–precisely because they tend to play similar amounts of minutes against relatively similar kinds of players. Your small sample size issues kinda go away then because you’re at least closer to comparing apples to apples–nobody’s 10th ranked five-man unit plays a ton of minutes. I didn’t do that explicit analysis here. I may try to do more of that next.

    What I think this data can give you is some insight on what things particular combinations did (or did not) do well–relative to the competition it faced. I’d be careful about comparing the units to each other directly.

    KnickFan –

    I’ll get our crack staff of interns right on that ;)

  7. KnickFan

    Dave — good point.

    In case anyone does want to do this, though, they should make sure to make the dependent variable is the plus/minus number
    ——————–
    minutes/19680

Comments are closed.