Identifying a player’s position is useful for all sorts of statistical analysis of basketball, but unfortunately position in basketball is not nearly as well-defined as position in baseball. The traditional breakdown into point guard (1), shooting guard (2), small forward (3), power forward (4), and center (5) works some of the time, but breaks down at the edges. Some teams’ offensive systems don’t differentiate between the roles for the two wing positions (SG and SF), or between the two post positions (PF and C). Some players play one positional role in their team’s offense yet typically guard an opposing player that plays a different positional role in his offense (e.g. Kirk Hinrich, who plays PG for the Bulls offensively but often defends opposing SGs). Many players play different positions at different times in the same game depending on which teammates they are on the court with. For all these reasons and more, having a list saying Player X is a PG, Player Y is a PF, Player Z is a SF, etc. is bound to be lacking.

How can positions be assigned in a more objective and informative manner?

82games has used play-by-play lineup data to address the problem of players playing different positions at different times of the game. What they do is to look at all the players on a team’s roster and rank them in order from most PG-like to most C-like. The idea is that if you picked two players from the roster and both were in the game at the same time, the one who came earlier in the ordering would be playing a smaller position (i.e. PG<SG<SF<PF<C) than the one who came later. Their ordering lists can be deciphered by looking at the “Player 2″ column from each team’s Player Pairs page . Then, for each stint that a different five-man lineup was on the court, they assign each player to having played those minutes at the position determined by their ordering relative to their teammates they were on the court with. In the end, this allows them to present for each player their total number of minutes played at each of the five positions.

For example, here was their ordering for the 06-07 Spurs, from most PG-ish to most C-ish: Tony Parker, Jacque Vaughn, Beno Udrih, Manu Ginobili, Brent Barry, James White, Michael Finley, Bruce Bowen, Eric Williams, Matt Bonner, Fabricio Oberto, Robert Horry, Tim Duncan, Melvin Ely, Jackie Butler, Francisco Elson. For each minute the lineup of Parker, Ginobili, Finley, Horry, and Duncan were on the court, Parker was assigned as having played those minutes at PG, Ginobili at SG, etc. Tallying things up for the season, you get results saying that Duncan played 18% of his minutes at PF and 72% at C, Ginobili played almost 100% of his minutes at SG, etc.

This is a great way of assigning positions, but it still has some problems. First, it’s only as good as its team-by-team player orderings, and those just aren’t very good. They often list players in a clearly incorrect order, which leads to misleading final results. Fans of a particular team can probably make a very accurate ordering of the players on that team from most PG-ish to most C-ish, and perhaps a collaborative effort in the mold of Tom Tango’s baseball Fan Scouting Reports could be organized to improve matters. Of course, no matter how accurate the orderings, there will still be some issues, like the problem of a player playing one position on offense and another defensively.

An alternative using an objective ordering method and public lineup data from BasketballValue

In the meantime, I decided to try a different method for assigning positions. It’s based on the idea that typically, the shortest player on the court is the PG, the second shortest is the SG, the third shortest is the SF, the fourth shortest is the PF, and the fifth shortest (or the tallest) is the C. Of course there are a lot of exceptions to this, but I think it holds generally.

Using play-by-play lineup data from BasketballValue and player heights from NBA.com (which may not be the most accurate), for every different five-man unit stint on the court in the 06-07 season I looked at where each player ranked height-wise relative to the teammates he was on the court with (i.e. for a given stint, each player was assigned as playing those minutes as the 1st shortest player on the court for his team, or 2nd shortest, or 3rd, etc.). When there were two players on the court of the same height, their rank was split (e.g. if there were two 6-3 players in the lineup and no shorter players, those two players were each credited as playing those minutes as the 1.5th shortest player on the court [averaging the 1st and 2nd]). These ranks were totalled for each player, giving an end result listing how many minutes a given player played during the season as the 1st shortest on the court, the 2nd shortest, etc. I also took a weighted-by-minutes average of the ranks to arrive at a one number representation of where that player ranked on average height-wise for the season. Hopefully this will all become clearer by looking at some examples.

First, let’s look at some league averages. These results are interesting even if you’re not sold on the idea of using this data to assign positions. Here are the average (minutes-weighted) heights for the shortest player on the court for a team, the second shortest, etc.

  1st      2nd      3rd      4th      5th
-------  -------  -------  -------  --------
6' 1.8"  6' 4.9"  6' 7.2"  6' 9.1"  6' 11.1"

So the average on-court lineup from 06-07 was made up of players who were 6-2, 6-5, 6-7, 6-9, and 6-11. However, there were actually more minutes played with the shortest player on the court being 6-3 as opposed to 6-2. Here’s a chart that looks at that - the distribution of minutes by height for the 1st shortest player on the court (for one team), 2nd shortest, etc. (the columns all sum to 100%).

Ht    1st  2nd  3rd  4th  5th
---   ---  ---  ---  ---  ---
5-5    2%
.
.
.
5-9    1%
5-10   3%
5-11   2%
6-0   11%
6-1   20%   3%
6-2   21%   3%
6-3   26%  22%   1%
6-4   12%   9%   1%
6-5    1%  19%   8%
6-6    1%  23%  18%   3%
6-7        13%  32%   9%
6-8         7%  25%  18%   2%
6-9             11%  35%   9%
6-10             3%  19%  20%
6-11                 13%  33%
7-0                   2%  26%
7-1                        5%
7-2                        1%
7-3                        2%
.
.
7-6                        1%

So 26% of the time the shortest player on the court was 6-3, while 21% of the time the shortest player was 6-2. Remember, this is not looking at all ten players on the court (that would require columns from 1st to 10th), just at the five-man lineups for each team separately.

As for some trivia, in the 06-07 season there was a team that on one occasion sent out a lineup of five players of the same height. In the January 4 game against the Lakers, for a little over 30 seconds early in the second quarter, the Kings played a lineup of Francisco Garcia, Kevin Martin, Ron Artest, Corliss Williamson & Kenny Thomas, all of whom are listed at 6′7″ on NBA.com. They were outscored 3-0 (all from Kobe Bryant free throws) before Mike Bibby checked in for Artest, thus ending the grand experiment.

Player-by-player data

I’ve posted a table on Swivel listing the seconds that each player in the league played as the 1st shortest on the court for his team, 2nd shortest, 3rd, etc. (actually, I split things up further to account for cases when there was another player in the game of the same height). I also calculated their mean height ordering (weighted by minutes at each rank). This is the number that I think could be used as an alternative to a static position designation (or to height) in doing any type of analysis. Obviously in some cases this is not a good representation of the position a player played, but even then it may be useful simply as a quantification of their “functional” or “playing” height.

To help understand the data, I’ll go through each column for Speedy Claxton. First the season, team, and player are listed, along with the player’s ID number in the BasketballValue database. Next is the player’s height in inches. The next column, titled “1 (sec)”, lists how many seconds that player played (here 50429) as the 1st shortest player on the court for his team. “1.5 (sec)” lists how many seconds he played (4941) as the co-shortest player on the court (i.e. time Claxton was on the court with Salim Stoudamire, also 6-1, and three taller players). And so on for the other height orderings. Finally, Claxton’s mean height ordering of 1.17 is listed. This indicates that he primarily was the shortest player on the court, though that was not always the case (if it were his mean height ordering would be exactly 1).

Going forward

In the future I will look the data for the 05-06 season, since that is also available on BasketballValue. Another possibility I may consider is integrating the heights of the players in the opposing team’s lineup, which may help with determining what kinds of players a player had to defend. Ultimately, I think this kind of analysis can only go so far, and hopefully better position orderings will become available that offer more than can be found by simply looking at height.