Stats - +/- = devil's advocacy group

[thead]

Okay 15 games into the season I am seeing a lot of stats that range from informative to downright ridiculous

for instance the Hollinger Stat showing us as have a 74% chance to make the finals. So it got me thinking and after making a per 36 joke about Bruno's stats in another thread, I wondered... "When do the wheels fall off of certain statistics?"

I'll start with per 36. Obviously this is an attempt to scale a players statistics to reveal what would someone who plays less than that (typically a bench guy) give you as a starter. The stat however is meaningless because "the wheels fall off" when you factor in lineups, roles, the opposing defender etc... In fact unless a guy is the in 28 - 30 minute range I barely even engage that statistic anymore.

PLAYER EFFECIENCY RATING

Right now Brandan Wright is second in the league in PER and it features a number of high volume shooters. It would appear that PER is the antithesis to Per 36

PLUS/MINUS

Say James Johnson was starting, and a team went on a 16 point run with their starters. James Johnson did everything he was asked to do and did it well but the other four were playing matador defense. Someone checking the box score would assume that he played like shit

What stats do you like to use for argument sake?

What are the draw backs to other statistics used?

[NaijaBoy17]

On the topic of per 36.

The only applicable use for per 36 is to compare the raw stats of two players who play around 28+ minutes per game, but don't play the same amount of minutes. Let's say player A plays 29mpg and player B plays 35mpg for example.

Using it to extrapolate for players that play like 18mpg is dumb, otherwise Brandan Wright is basically a top 5 player statistically.

[TeamEd]

The key words in any stats question are "sample size."

+/- for a single game doesn't mean much, sure (personally I think it's an underrated stat) but over the course of several games ideally over several months it becomes meaningful. A player who's on/off numbers consistently out perform his teammates is probably playing winning basketball.

As for per 36 stats. They don't mean much if you're trying to extrapolate some bench player into a starting lineup. But, you shouldn't do that. Don't do that. That's not what the stats are for. Per 36 stats give you a good sense of how someone is playing in a particular role, but you always have to keep the situational stuff in mind.

/ Per 36 stats also let you say stupid things like 'the Cavs lost the Kevin Love trade because they gave up Anthony Bennett.'

[DanH]

Per 36 stats are just a good way to judge whether a player is doing well in his minutes. Is a guy who scores 4 points and grabs 3 rebounds in 10 minutes performing? Hard to say from the raw numbers, since we are used to judging most players on a full game scale - 30 to 36 minutes. Put that in per 36, and clearly the player is very solid.

It doesn't mean he can do that in 36 minutes. It's just a normalizer to simplify looking at stats for minute restricted players. Some people use it wrong. But the best use of it is to see if players deserve (some) more minutes - if a guy is posting a 20/10 per 36 in 18 minutes, he's probably a good candidate for increasing his minutes to 24 minutes, to see if he can extend that performance.

Plus minus is useless in single game situations. But over the course of a season, it can be meaningful (moreso as a net rating rather than a raw plus minus). Really though you need to be very mindful of lineups when discussing net ratings. That was the goal of my WOWY post last year - sorting out the individual impacts. Also the idea behind adjusted plus-minus.

PER is largely useless no matter what. But any catch all stat can have the downside of low-usage, extreme efficiency players skewing the results. That's why a lot of newer WS variants include minutes, possessions and even individual possession usage as factors in the total contribution.