Let’s assume you’re the GM of an NBA franchise. It’s the offseason and you need to sign a free agent center. You have three options.
If these are the only stats we have, then it’s clear. Player A is a superior scorer and rebounder, and his defensive stats are about as good as the other two. On the other hand, Player C is a feeble scorer and the worst rebounder of the three. Player B is twice as good a scorer than C, and slightly better in rebounding and shot blocking. I would rank them Player A, Player B, then Player C.
Now what if your options were these three players:
In this example Player C is the best choice. He is superior to the others in 4 categories: rebounds, blocks, steals, and turnovers. Meanwhile Player B sinks to the bottom of the list, being too turnover prone and the worst rebounder of the three. This time I would rank them Player C, Player A, and Player B.
Now what if I revealed that the players in the first table are the same as the ones in the second table? How is it that a player can have two entirely different set of stats? Simple, the first table is the players’ per game stats, while the second is their per 40 minute stats. The stats changed so much from the first table to the next because per game stats are proportional to the number of minutes a player receives. And in this example, the players played varying amounts of minutes. Player A averaged 34.7 min/g, Player B 31.2 min/g, and Player C 24.1 min/g.
Playing time in the NBA is dependent on a few different factors. For younger players their talent, draft position, contract size, pre-draft hype, team depth, coach’s tendency, team record, and sneaker deal may alter their court time. A #2 overall pick playing for a rebuilding team with little depth (LaMarcus Aldridge) will see more playing time than a #4 pick playing for a playoff team with ample forwards/centers (Tyrus Thomas). Since per-game stats are proportional to playing time, and playing time is based on many different factors, then it makes sense that per-game stats are capturing some factors other than a player’s ability. In other words when Aldridge had more rebounds per game (5.0 rpg) last year than Thomas (3.7 rpg) it’s not due to Aldridge’s skill on the glass, but rather his higher draft position, team depth, team record, et al.
If we want to judge a player based on their talent alone, then we need to isolate a player’s talent from the rest of those variables. And that’s what per minute stats do. Once we remove a player’s minutes from his stats, then all the factors that go into playing time are removed as well. Per minute stats aren’t dependent on draft position, contract size, pre-draft hype, team depth, coach’s tendency, team record, or sneaker deal. When compared to per game stats, per minute stats come much closer to capturing a player’s ability.
Getting back at our tables above, it’s clear that Player C was hindered by his lack of playing time. Hence why he appeared to be inferior when using per game stats. But when we accounted for this lack of playing time with per minute stats, Player C was clearly superior to the other two. Player C, also known as Ben Wallace, would receive major minutes the next year in Detroit and win the first of his 4 Defensive Player of the Year award the season after. Player A, Dale Davis, was a fine player for 14 seasons, but was never considered great. Michael Olowokandi, Player B, never amounted to the hype of being the #1 overall pick, and despite being a year younger than Wallace, is barely holding on to his NBA career.
In the end, per game stats was unable to distinguish the perennial All Star (Ben Wallace) from the solid pro or the first round bust (Michael Olowokandi). However per minute stats identified the players correctly from most to least talented. Using per minute stats to compare players eliminates many of the unwanted factors that go into per game stats. Like sneaker contracts.
- Why do we use per 40 minutes stats? Ben Wallace averaged 2.3blk/40 and he also averaged 0.072blk/min. It’s easier to visualize 2.3 blocks instead of 0.072 blocks. You could use any number instead of 40, but we use 40 since it has become the most commonplace.
- Using per 40 minute stats isn’t an endorsement for the player to receive that many minutes. Nor should you use per 40 minute stats for one player, and per game stats for another. (And yes I mention this because I’ve seen people do it).
- You could use more advanced metrics than per-minute stats, but per minute stats are easily calculated and widely circulated.
- Don’t think that per minute stats hold up over increased minutes? Check this out:
I did a small study using player-seasons from 1978-2004. To be included in the study, a player had to (a) see an increase of at least 50% in minutes per game from one season to the next and (b) play at least 41 games in each season. These criteria gave me 465 player-seasons. In 346 of these seasons (74.41%), the player’s PER increased with an increase in playing time.
No, increased minutes do not seem to lead to decreased efficiency. In fact, the data indicates increased minutes lead to? increased efficiency. More than 70% of the players in the study (there were 251 in total) saw their PER (which is, by definition, a per-minute summary statistic) increase with the increase in minutes. Players whose minutes per game increased by five saw an average change of +1.38 in their PER.
Players who receive 10 or more minutes per game are likely to keep the same per minute stats no matter what the increase in playing time is. Therefore per minute stats remains far superior to per game stats in terms of comparing and evaluating players.