The Lamp Post: Put me in, Coach
“People commonly use statistics like a drunk uses a lamp post; for support rather than illumination”
– Mark Twain
Every NBA team has 15 roster spots, and they are allowed to dress 12 players for each game. What I’m hoping to shine a little light on today is how much of that 12-man depth is regularly used. How much does it vary from team to team? How does it relate to a teams’ success? Within this analysis, I will also show where Jim O’Brien’s Pacers fall. Does he reach deep into the bench? Is he parsimonious with minutes, or does he spread the wealth around?
To answer these questions, I charted all 1,230 games played during the 2009 NBA season. Rather than looking at minutes played by specific players, I sorted the players in each game in descending order of minutes played. This created a “slot” for each player, from 1 to 12 (if all 12 played). As an example, Danny Granger would be slotted differently each game. Over the course of the season, he was slotted anywhere from 1 to 8, with “Player 1″ and “Player 2″ being the most common at 22 times each. This should theoretically allow us to see how a coach “manages” the game, without regard to the individual players.
Is Eight Enough?
Playing an eight-man rotation is relatively common approach in basketball. Given the fluid nature of the game, it’s desirable to always have five players on the floor who know each other well and provide the stability of familiarity. Pacer icon Slick Leonard is a proponent of playing an eight-man rotation.
However, the eight-man rotation would be considered a tight rotation by most NBA fans today. They would tell you that eight is most certainly not enough. It’s hard to argue with those three ABA banners and 529 career wins with the Pacers,but it’s important to remember that Leonard’s hey-day was 30 years ago. I looked at two measures to get a sense of how deep today’s NBA coaches regularly reach into their bench.
The chart above indicates how many games, on average, that NBA teams used each “slot.” Teams went eight deep over 99% of the time, and nine deep over 92% of the time. Slot #10 saw the floor in just under 64% of the games. From this perspective, an eight-man rotation would certainly appear to be “tight.” Looking at the distribution of minutes by team gives you a different picture.
League wide, the top 8 slots played 93.5% of all of the minutes played. The chart above shows that all but one team (San Antonio) used their top 8 men for over 91% of their minutes. In fact, the top 6 players (on average) played about 80% of the minutes, and only one team (again, San Antonio) used their top 6 for less than 77% of the time. So, it appears that while most teams play nine or even ten players most nights, the 9th and 10th players generally see only token minutes.
In the Middle of the Road
The 2009 Pacers stand very close to the league average in terms of how Jim O’Brien used his depth. The top eight players got 93.6% of the Pacers’ minutes, which was 13th in the league and only slightly above the 93.5% league average. In fact, the chart below shows that the Pacers are almost dead on league averages for both Games Played by Slot:
As well as the average Minutes per Game by Player Slot:
While they’re dead on the averages, it is worthwhile checking to see whether that result is purely mathematical, or if the O’Brien’s use of the Pacers’ depth can actually be considered typical. No Pacer player slot was higher than 12th (out of 30) in Minutes per Game played, and only slot #10 (23rd) was lower than 19th. Even slot #10 did not have a material variance, as the Pacers were at 6.0 Minutes per Game played, and an additional 90 seconds per game would have put them in the top 5 in this category. It seems fair to me to say that O’Brien’s use of the player was slots was “typical” of the NBA.
Of Chickens and Eggs
So, one of the questions that remains is, “Does this have any effect on wins and losses?” There are some high level numbers that argue that they do. If you break the league down, the overall winning percentage for the 10 teams that use their top eight the most is .466, then it goes to .498 for the middle 10, and .537 for the ten teams with the lowest 8-man percentages. Further, five of the 14 teams with winning records were in the bottom 10 in terms of 8-man percentages. There were only three winning teams (Atlanta, Houston, and Cleveland) among the ten teams who relied most heavily on their top eight.
However, if you look more deeply at the individual teams, the discussion becomes murkier. First, teams like the Celtics and the Lakers shared the bottom 10 with two 24-win teams (Minny and Memphis) and two 34-win teams (the Nets and the Bucks). Second, it’s important to understand just how little variation there is between these teams. Twenty three (23) of the 30 teams in the NBA are less than 30 seconds per player per game above or below the league average, meaning the extremes are within one minute per player per game of each other. More to the point, is any reasonable person going to believe that the difference between the 62-win Boston Celtics and the 19-win LA Clippers was even vaguely influenced by the fact that Boston spread less than 7 minutes a game over slots 9 through 12?
The Pacers, Pace, and the Perfect 10
There has been much talk in the preseason about Jim O’Brien wanting to use a 10-man rotation this year, with the idea of maintaining pressure on the opponent both offensively and defensively. As noted above, O’Brien played basically a 9-man rotation last year. Going through his coaching history shows that 2009 was not an anomaly. He typically used nine players in about 93% of his teams’ games, and Player # 9 normally averaged 9-10 minutes per appearance. Going to a 10-man rotation will be something more of a stretch.
O’Brien has definitely used the 10th player more liberally in Indianapolis than he has in his previous stops, but the numbers still fall well short of a minimum 10-minute per game rotation as defined by Rick Barry and Jordan Cohn in their Pro Basketball Bible series of days gone by. The truth is that no one in the NBA used a 10-man rotation last season according to the Barry & Cohn rules. San Antonio came closest, playing employee #10 in 71 games for about 8.7 minutes per appearance. Looking back at some recent Pacer history (1994-present) indicates that only Bird’s 1998 and 1999 Pacer teams used a bona fide 10-man rotation. (The 2000 Finals team was a very solid 9-man rotation.) While playing a steady 10-man rotation may be a goal, history says that it’s probably unlikely to actually happen.
Which brings us to a discussion of pace vs. rotations. It seems relatively intuitive that the faster pace, the deeper you’ll need to reach into your bench. However, this hasn’t necessarily been true in practice, either here or throughout the league. The Pacers, had the third fastest pace last season at 96.5, but still ran an 8-man rotation that was the 13th most heavily used, and 16 teams gave more minutes to their 9th and 10th player slots than the Pacers did. It gets even more out of whack when you look at the two teams that played “faster” than the Pacers: Golden State and New York. These two teams used their 8-man rotations for 95.2% (5th) and 97.1% (1st) of their minutes respectively. Of course, those two teams combined to win only 61 of their 164 games, so they might not be great role models.
In fact, if you look at the top 10 teams by Pace (or possessions per game), you see a pretty clear correlation between having a lower usage of your top eight players and having more wins. Six of the teams were in the top 13 for 8man percentages, including the Pacers. All had losing records except for Chicago, who was only .500. Meanwhile, four teams were 17th or lower in the 8man, and they all won between 46 and 65 games. However, it’s important to remember that correlation is not the same as causation.
First, it’s important to note the differences in pace throughout the league are relatively immaterial. The Pacers were less than 5 possessions per game above the league average, and less than 10 possessions per game more than the slowest team, Portland. That’s about 5 and 10%, respectively. Also, it’s important to note that the 2009 Pacers, while third in the league, were not a historically fast team. Basketball-Reference.com lists the Pace Factors for NBA teams going back to 1974. In that sample, there have been 923 teams, and 415 (45%) of them were faster than Jim O’Brien’s 2009 Pacers. The problem is that only 10 of them have been since 2000. In my opinion, this creates a problem of perception, as opposed to reality. That, however, is a can o’ worms that needs to be addressed separately and in more detail. Pace is something I don’t consider to be a huge factor, but others probably will. I can’t prove anything definitively to myself or anyone else at the moment, so we’ll come back to that another time.
What can be established more clearly (and has been in Exhibit B above) is the insignificance of the difference between the 8man percentages. The Pacers, at 13th, allotted 93.6% of their total minutes to the top eight slots. The Lakers, at 27th, allotted 91.9%. The total difference due to the percentages is 345 minutes. Over 82 games, that’s only 4.2 minutes per night, or about 31 seconds per player (in the top eight) per night.
Furthermore, there is an overriding factor that drives the minute distribution, and that is the nature of the individual game. Specifically, how close is the contest?
This chart shows the minute distribution by grouping broken out into three “margin groupings.” The far right hand group shows the actual distribution of all NBA games. As you can see, the Top Eight follows the normal distribution pretty closely. However, slots 9 & 10 begin to see a disproportionate amount of their minutes in Double Digit games, and 11 and 12 appear to be relegated almost exclusively to garbage time.
Taking this into account is very important when considering the Pacers because they were 1st the NBA in “Close & OT” (4 points or less or overtime) games with 30, and they were dead last in “Double Digit” games with 24. Further, they were outliers in both. The league average for “Close & OT” games was 18.5 games, and no other team had more than 27. In fact, only three teams (Indiana, San Antonio 27, and Detroit 26) had more than 23. The league average for “Double Digit” games was 38.5 games, and only one team besides the Pacers had fewer than 30 (Detroit 28). Going back to our earlier discussion of the 10 “fastest teams” and using this information indicates that both Phoenix and Denver played tighter rotations than O’Brien’s Pacers, but ended up with lower 8man percentages by virtue of playing in fewer close games and more blowouts. The Pacers rotation was very similar to the Lakers most games, but the Lakers had a more liberal rotation in blowouts, and also were involved in 18 more of them than Indiana.
Exhibits F & G
These aren’t perfect representations of how much “garbage” time each team had, but they are strong directional indicators.
Of Eggs and Chickens and the Price of Tea in China
So, now comes the circular argument about whether the Pacers played a tighter rotation than they might have otherwise because of all the close games, or were they in a lot of close games because they played a tighter rotation. There’s probably a little of both, but in my opinion, it would lean more towards the former than the latter. Looking at the Pacers’ 11-19 record also makes you wonder if there were some slight changes that could have been made to make that better. Perhaps reaching a little deeper would have resulted in more wins in those close games. It’s a compelling argument, particularly if you believe the pace had an adverse effect on the Pacers at the end of games.
There are two pieces of data that would not support such a position. First, Indiana was 3-11 in these “Close” games prior to January 1st of this year, 8-8 after. Also, the Pacers rotations after the first of the year tended to be tighter in close wins (about 97% of the minutes for the top eight) than in close losses (about 94%). The simplest explanation for the improvement is that the Pacers just got better at finishing games. This would be encouraging news, if it weren’t for the fact that the player who made the most dramatic improvement in close games after the first of the year was Jarrett Jack.
Ultimately, the game is always going to come down to talent. All coaches will have players they trust, and players they don’t. The better teams most often have more players to trust. There’s a tendency to exaggerate the influence that a coach can have on the game, particularly when it comes to rotations. Or, perhaps more accurately, there’s a tendency to exaggerate how differently each coach will behave when it comes to rotations. With a couple of notable outliers (the Knicks and the Spurs), most NBA coaches distribute their minutes in pretty similar ways. Almost all will rely on their top eight players, using a 9th player for short rests and foul issues, and generally allotting garbage time to the rest of the rosters. Who exactly should fill those top eight roles, or the 9th, or 10th, etc. is a debate for another time.