For a basketball event so important as the turnover, we know too little about it. Here is an attempt to change that.
In his 2004 book “Basketball on Paper”, a standard reference for basketball statheads, Dean Oliver outlined four key factors in basketball. We know them as “The Four Factors”:
- Shooting percentage (40%)
- Turnovers (20%)
- Rebounds (15%)
- Earning and making free throws (10%)
Here’s a review of some of the information we have on the four separate factors.
There is plenty of shooting data on the net. Eli Witus, now vice president of Houston Rockets basketball operations, produced the first visuals in 2008, Hoopdata followed in non-visual form. Basketball-Reference has them and so does, at last, NBA.com. Several European leagues have shot charts in their live stats applications, and when summarized, they look like this. Kirk Goldsberry, a geography scholar at Harvard, took spatial analysis to another level over the last couple of years, combining the gold mine that is optical tracking data with his geo-mapping knowledge and tools.
Wherever the data is coming from: Every player, every team is different, but on a league-wide basis, universally, there is one common culprit: the mid range shot. Take into consideration the fact that a three point shot carries more value than a two point shot, then the three outperforms every – even the upper paint – location in the halfcourt except for the shot right at the basket. The corner three is a particularly good shot – most likely because it is a little shorter than the straight-up three, and because it is the most assisted shot in basketball (except for the alley-oop).
Unfortunately the play-by-play data at our disposal is so limited that there is only little depth to our shot charts: There is no information on the proximity of the defender, nor do we know whether or not the shot had been assisted. And they carry a shooting foul bias, as Finnish coach Harri Mannonen pointed out.
82games, however, had 51 charters keep track of the “potential assist” through an unknown number of NBA games in a 2006 study, and the results were fascinating:
|Shot Type||Assisted||Unassisted||Margin||Pot Ast% of Attempt|
|3-Pointer (eFG%)||37,9% (56,9)||34,2% (51,3)||+3,7% (+5,6)||81%|
The average assisted mid range shot, it appears, is a slightly subpar but not outrageously bad shot; still it is outperformed by assisted close- and three point shots and even unassisted shots from both areas. When you isolate shots from context, that is.
What’s more? In an eye-opening Sloan Sports Conference presentation in year 2011, Sandy Weil used STATS, LLC data to explain the importance of being open. Tight defense, so the study concluded, reduces field goal percentage by 12 percent. Still, that did not help the mid range jumpshot’s case:
Tightly contested layup, moderately contested 3-pt jumper, and open 19-foot jumper have similar expected values.
In limited cases, however, the expected value of a mid range shot taken by a good jump shooter can still be beneficial to the offense when you are facing elite defense, as this ESPN Insider piece on the Portland Trail Blazers’ statistical analyst Ben Falk explains:
The Pacers are the best defensive team in the league, particularly in the metrics that matter. They rank first in effective field goal percentage, allow the second-fewest shots at the rim and give up the second-fewest corner threes. Falk knows that basketball is a game of constantly shifting probabilities, so tonight the Blazers go against the stats. “Baseline percentages are only broad summaries,” he says. “They may not always apply for a lot of reasons, including the other team’s scheme and personnel. Against a team like Indiana, getting open shots for the right shooters, even if they are in midrange, can be a better-percentage play than forcing a tough shot at the rim.
Both Witus and Goldsberry have done tremendous work here. Goldsberry’s fascinating interactive graphic reveals that close range shots, as they are defined in his graphic, are rebounded by the offense 35,4 percent of the time. The short jumpshot- and three point zones come (a distant) next, whereas the long two-point jumpshot zones, especially the frontal jumpshot zones, produce the lowest offensive rebound percentages. Missed free throws had an offensive rebound percentage of 19,2 in the 2012/13 Euroleague season.
Furthermore, quality offensive rebounders impact team offensive rebounding far more than quality defensive rebounders impact team defensive rebounding. A player with strong defensive rebounding numbers takes a good portion of rebounds away from his teammates, a strong offensive rebounder usually does not.
There is quite a bit of strategy involved, too, as crashing the offensive glass with multiple players is generally believed to weaken transition defense. There is little doubt crashing the glass with four players will do just that, but where is the break-even point? How many players do you use and what positions should they be attacking from? What type of offense are you facing? The 2012/13 “party-crashing” Indiana Pacers managed to do both: Put up strong offensive rebounding numbers AND play top-level transition defense.
Earning and Making Free Throws
We have a pretty good idea where shooting fouls are drawn: 90 percent in the paint, according to this 82games study, and 7 percent in what they call the “2-pt Jumpers” area. You need to get to the basket to earn free throws. This long-standing wisdom is not being challenged.
Then there is the team foul bonus shot: Looking at data from quarters one, two and three (to eliminate trips to the foul line coming from stop-the-clock situations in fourth quarters) from the 2012/13 Euroleague season, the average team made at least one bonus trip to the foul line due to the opponent reaching the team foul limit in 47 percent of quarters they played. Included are trips that were the result of actual shooting fouls rather than bonus trips on non-shooting fouls; the two are inseparable in play-by-play data. If the average shooting-fouls-to-total-fouls ratio from non-bonus situations (42,9 percent of all defensive fouls led to trips to the foul line through the 2012/13 Euroleague season; 82games has similar numbers) is any indication, however, teams took 0,4 bonus free throws per quarter. That may seem insignificant, but look at how marginal team season averages generally differ from one another: it is not. Non-shooting fouls can be game changing, and their impact as early as the mid-quarter is evident by the minute-by-minute averages.
According to 82games, the upper paint area is right on the level of the deep paint (at rim) area in terms of defensive fouls per possessions called. There is a larger percentage of non-shooting fouls in upper paint fouls, though, than right at the rim.
A free throw has a super high expected value, which is why — thankfully — we are seeing the hack-a-shaq so little. Even a poor, 60 percent free throw shooter has a 1,2 points per possessions expected value on a two-shot trip to the foul line. Plus: Team fouls. Plus: A one-in-five threat of an offensive rebound if the second shot misses.
Turnovers … ?
Rodhig‘s February 2013 piece on Vassilis Spanoulis’ areas of efficiency was a real eye-opener. Charting Spanoulis’ turnovers, passes and shot attempt from a small number of mid-season games, his conclusion was:
Spanoulis’ forays into crowded spaces closer to the basket often help his efficiency. Even if his percentage in mid-range shots drops, those four games have indicated that this area of the floor can serve as a platform for a bunch of low-risk assists – twelve of them, to be exact (as opposed to only three TOs), eleven of which led to quality shots at the rim and behind the arc.
Even though [advanced stats] help us ask all the right questions, sometimes the answers are more complicated – especially in the case of volume shot creators, who have to make the most of the space afforded to them by a defense designed to stop them. Players like Spanoulis can’t act as a homo economicus out there, limiting their game into high percentage areas. They need to put a system into motion. For most coaches, the endgame of this system is a quality shot. For most shot creators, the key is finding free space and making the right decision once they are there.
In rodhig‘s study, ten of Spanoulis’ 13 turnovers were committed on passes towards the high-value close range area, eight of those ten coming on passes from beyond the three point arc. Too small and too individual a sample to draw definite conclusions, but the questions he asked I found fascinating: The dunk/layup and corner three are — we understand that — the best shots in basketball, but what is the risk of creating a quality shot? Where do assist- and turnover locations come into play?
As for assists, I have nothing to offer other than to refer to Goldsberry’s “Where do corner 3s come from?”-chart published in September 2012. This chart screams “high risk” right at me:
- The highlighted area must be reached via pass or drive. And then a pass needs to happen — either to the strong side corner or even cross-court to the weak side corner.
- There are plenty of assists from under the basket. This is an area the defense does not want you to be in. And the pass right along the baseline into the corner is a difficult one.
- A number of the short wing-to-corner passes may actually result from an initial drive and kickout. This is a common sequence in basketball: Drive, kickout to the weak side wing, extra-pass against the closeout defender to the corner.
As for turnovers, I am sure my following study of 634 turnovers from games from the 2011/12 to (current) 2013/14 Euroleague seasons leaves plenty of space to fill. But it’s a start.
I charted 340 half court passing turnovers (half court defined as: all players must have passed the half-way line) for passer and intended recipient. They were fairly clear in who and where the pass was aiming for.
The following graphic shows the origin zones of passing turnovers.
More than two thirds of all passing turnovers originated from the central three point area. The left short two area (13.5 percent) is where plenty of post-ups started. The large majority (roughly three in four) of post-ups I came across were started on the left block, which is a preferred position for right-handed post players.
Here’s what areas picked-off passes from just this central three point area were aiming for:
Nearly half of all passing turnovers from the central three point area were aimed towards the deep paint area! Within those, one third of the turnovers were committed by big men who were looking to complete high-low feeds. This is a difficult pass and there appear to be plenty of players who do not master the high-low.
Plenty of turnovers happen on long passes that target a player two- or even three zones away. That is one of the reasons why defenses love to collapse on the ballhandler: The long pass is difficult to master. The top level passing point guards (Huertas, Diamantidis, Jasikevicius in particular) rarely commit turnovers on this play, but lesser quality playmakers and especially big guys often have their passes picked off or have them end up in the press row.
The large quantity of turnovers on passes to the left wing area outside the three point line (12.2 percent of all turnovers on passes from the central three point area) might surprise. These were two types of passes: First, simple passes to the wing, for whatever reason (finding a good angle to feed the post is one of them). Second, long passes towards the extended free throw line position on the wing. When teams run the pick and roll away from the single (off-ball player) side, this is the position the off-ball player drifts towards from the corner.
Below are the target areas of our 340 passing turnovers. This is the price teams pay for trying to consistently produce the best shot in basketball, the assisted close range shot: They turn the ball over, in numbers.
53,2 percent of all passing turnovers teams committed came from passes targeted towards either the deep paint (40,6 percent) or the corner three point areas (12,6 percent). Those are usually “potential assist” type of passes: Passes into the deep paint and in the corner create immediate scoring opportunities.
Non-passing on-ball turnovers
Plenty of non-passing turnovers (not included: moving screens, three seconds, five seconds and the like) are committed in the deep paint area, largely by big men who were crowded by the defense on the catch and then had the ball stripped or committed a travel.
46 percent (17,9; 19,2; 8,9) of non-passing turnovers were committed in the three short two zones (excluding the two baseline zones). Keeping in mind Goldsberry’s corner three assist chart, it appears as though these zones are absolutely crucial to get in or through to create quality shots.
From the sample of turnovers I looked at, plenty of pick and roll turnovers were committed in deeper zones, whereas isos were largely stopped in the outer zones. Perhaps the ball screen’s success, especially when the defending big man is sitting back at the free throw line, is in parts due to the fact that you gain enough separation from your defender to attack the defense and get deep into the upper paint/short two zones.
Plenty of turnovers in the three point zones were due to traveling violations and — in the two corners — out of bounds violations.
A three point shot off the dribble with a bit of separation between offensive player and defender is a low-risk play. Which may be one of the reasons why some coaches ask their big men to force the offensive player into a drive after the switch, rather than give up the jump shot — especially in low-shotclock situations.
The post-up (due to its poor average efficiency on an individual points per play basis) has not exactly gained in popularity since Synergy Sports data came into play, but the creation of a post up or a wing isolation is a low-risk process. Although turnovers are committed occasionally on simple passes to the wing (largely by low-level, poorly organized teams), the average post feed was turned over just once every four games (per team) in my sample. Fighting for post position leads to the occasional offensive foul, but this is usually limited to a good dozen of rather physical specimen. The post-up immediately gains in complexity, though, once the defense decides to double- or triple the post player.
What needs to be done
This study is limited to turnovers. We do see how many turnovers were committed on passes from Zone A to Zone B. We do not see how many passes were completed.
Ideally, one would chart every pass made or attempted and then come up with a completion rate for passes and drives, and then look at what number and what types of different drives and passes the average, say, corner three is created through. This is likely what some of the statistics folks are doing for their NBA teams, with the help of optical tracking data.