The most misused stat of today: Per 36 numbers

[Granath]

Tim Duncan is another phenomenal example.   In Duncan's prime he was averaging 39-40 minutes per game.   Over his last 6 games, he was down to 28 minutes per game.   His stats were significantly lower.  From averaging 25/12/3 to averaging 15/9/2.    But look at the guy's Per-36 numbers it's EERIE how consistent he was regardless of whether he was getting 40 or 28 minutes per game.  Just look at it... it's all the proof you need that Per-36 has value.   Same player... just less minutes:

Swing and a miss. Strike one.

Your example is exactly how per 36 gets misused. Duncan played 39-40 minutes per game and was effective for that period of time. His last 6 years he couldn't perform at that level for that period of time. While he was on the court he was almost as effective but you can't take his last 6 years and say he was as good as he was before. He simply could not perform at that level for very long anymore. Furthermore, those reduced minutes likely means that while he was effective on the court he no longer had the ability to "turn it on" at the end of games or in critical moments. A guy who plays 25 mpg is conserving his strength on the bench to perform on the court. A guy who plays 40 minutes doesn't have that luxury. If that guy played only 25-28 mpg, his per 36 numbers would be significantly higher. Thus he's not the same player in fewer minutes and anyone who watched a lick of the Spurs the last 10 years knows that Duncan wasn't the same guy as he was before.

I'm not arguing that per 36 has no value. But it's often overused. I can use per 36 to compare a guy who plays 30 mpg with a guy who plays 38. I can't use it to compare a guy who plays half the minutes of another guy (or even the same guy).

Thus you can tout Embiid's per 36 numbers all you want but they're meaningless when the guy is playing 20 minutes per night and not allowed to play in back-to-back games. He's effective when he's in the game but he's not a 30/10 player because he can't sustain it. Thus per 36 is useless in this case.

[GC003332]

I would love to hear from the users of the per 36 stat in regards to foul trouble, tough to project out stats for many big guys,  when asked to play extended minutes they simply wouldn't be able to due to being in foul trouble and sitting on the bench.

[Ilikesports17]

Tim Duncan is another phenomenal example.   In Duncan's prime he was averaging 39-40 minutes per game.   Over his last 6 games, he was down to 28 minutes per game.   His stats were significantly lower.  From averaging 25/12/3 to averaging 15/9/2.    But look at the guy's Per-36 numbers it's EERIE how consistent he was regardless of whether he was getting 40 or 28 minutes per game.  Just look at it... it's all the proof you need that Per-36 has value.   Same player... just less minutes:

Swing and a miss. Strike one.

Your example is exactly how per 36 gets misused. Duncan played 39-40 minutes per game and was effective for that period of time. His last 6 years he couldn't perform at that level for that period of time. While he was on the court he was almost as effective but you can't take his last 6 years and say he was as good as he was before. He simply could not perform at that level for very long anymore. Furthermore, those reduced minutes likely means that while he was effective on the court he no longer had the ability to "turn it on" at the end of games or in critical moments. A guy who plays 25 mpg is conserving his strength on the bench to perform on the court. A guy who plays 40 minutes doesn't have that luxury. If that guy played only 25-28 mpg, his per 36 numbers would be significantly higher. Thus he's not the same player in fewer minutes and anyone who watched a lick of the Spurs the last 10 years knows that Duncan wasn't the same guy as he was before.

I'm not arguing that per 36 has no value. But it's often overused. I can use per 36 to compare a guy who plays 30 mpg with a guy who plays 38. I can't use it to compare a guy who plays half the minutes of another guy (or even the same guy).

Thus you can tout Embiid's per 36 numbers all you want but they're meaningless when the guy is playing 20 minutes per night and not allowed to play in back-to-back games. He's effective when he's in the game but he's not a 30/10 player because he can't sustain it. Thus per 36 is useless in this case.

TP.

[LarBrd33]

Doesn't have much value if you're looking at players who average like less than 15 minutes off the bench.  Like, we can't look at Jordan Mickey's stats in garbage time and realistically say that he'd flourish with 36 minutes in the starting lineup.

Where it's helpful is for looking at guys who put up minutes in the 25-46 minute range, imo.   People lost their mind over how great Rondo's stats were during the 2012 playoffs... 17.3 points, 11.9 assists, 6.7 rebounds and 2.4 steals.... but a key reason those stats were high is because he was playing an obscene 42.6 minutes per game.   Adjust it to the 36 minute average and his stats were pretty well in line with his other per 36 minute stats:  14.6 points, 10.1 assists, 5.7 rebounds and 2 steals.   In his final season in Boston he averaged 31.8 minutes... adjust it to 36 and he arguably had the best stats of his career:  9.4 points, 12.2 assists, 8.5 rebounds, 2 steals.   Rondo's career per-36 of 12 points, 9.5 assists, 5.3 rebounds and 2 steals are not dramatically different than the per-36 stats he put up during that playoff run everyone overreacted to.

The fact is, there are a great many players who will consistently put up per minute numbers.  You give them 25 minutes, they'll produce at a consistent pace.  You give them 40 minutes, they'll produce at a consistent pace.   Just because on one team he's getting 40 while on another team he might get 25 does not mean he was "better" one year and worse the other.  Isaiah Thomas is an excellent example of this.  He had his minutes dive to 25 on the Suns and his numbers were career low.   BUt his per-36 numbers that season were well in line with his numbers any other year:  21.4 points, 5.2 assists, 1.4 steals... his career per-36 was 21 points, 6 assists and 1.2 steals.   Same player less minutes.

Likewise, it's valuable for looking at a guy like Jimmy Butler who put up big stats while averaging 39 minutes... in comparison to a guy like Jae Crowder or Avery Bradley (averaging low 30 minutes).   I've been arguing for a while that with enough minutes and a big enough role, guys like Crowder and Bradley could probably put up 80% of what Jimmy Butler puts up.

During Butler's break-out season he avearged 18.6 points, 5.4 rebounds, 3 assists and 1.6 steals per-36.

Jae last year put up per-36 of 16.2 points, 5.8 rebounds, 2 assists and 2 steals with 44%/34%/82% shooting.

Avery last year put up per-36 of 16.4 points, 3.1 rebounds, 2.2 assists and 1.7 steals with 45%/36%/78% shooting.

I think that's valuable and important.  Clearly neither Jae or Avery has been given the same 1st option role that Butler has received.  But to me, it spoke to an idea that if you give those guys enough minutes, they could potentially put up near all-star numbers.   So far through 3 games, both of those guys are putting up all-star numbers... but Bradley in particular (who is averaging 36.7mpg) is currently Top 15 in the league.  Still early, but it doesn't surprise me.   

Role matters as well... that's why I tend to favor players with high shooting percentages.  Even when Steph Curry was a young player, he was getting plenty of minutes.  But his shooting percentages were astronomical.  I brought up multiple times that I would have loved to see what Steph could do with a more pronounced role in the offense.  Here was a kid shooting 49%/46%/90% and yet he was only getting 11-14 shots per night.   This was because Monta Ellis was the ball-dominating force on that team  You had Ellis taking 22 shots per night while putting up significantly lower shooting percentage.   I had to wonder what Curry would do with those kind of shot attempts.  Last couple years, we've seen it.  Curry had his shot attempts raise to 18-20 and his field goal percentages stayed consistent. 

These are things to consider when looking at a player like Avery Bradley.  He and Olynyk have been our most consistent shooters over the past 3 seasons.   What might happen to Avery Bradley if a team (and it might not ever happen on the Celtics) gives him consistent 36+ minutes and makes him a focal-point of the offense.   Was the last game a glimpse at that?  I think it probably was.   Though, that's not to say that he ever actually will get 36+ minutes and a consistently focalpoint role in an offense.


Per-36 is a valuable stat in context.   I get why you created this thread... it's a reaction to me tongue-in-cheek pointing out that Embiid's per-36 stats are incredible.   But for what Per-36 is typically used for, it has value for sure.

Honestly LB don't flatter yourself that much.

Tim Duncan is another phenomenal example.   In Duncan's prime he was averaging 39-40 minutes per game.   Over his last 6 years, he was down to 28 minutes per game.   His stats were significantly lower.  From averaging 25/12/3 to averaging 15/9/2.    But look at the guy's Per-36 numbers it's EERIE how consistent he was regardless of whether he was getting 40 or 28 minutes per game.  Just look at it... it's all the proof you need that Per-36 has value.   Same player... just less minutes:

Code: [Select]

                               
Season     FG%  TRB AST BLK  PTS
1997-98   .549 11.0 2.5 2.3 19.4
1998-99   .495 10.5 2.2 2.3 19.9
1999-00   .490 11.5 2.9 2.1 21.5
2000-01   .499 11.3 2.8 2.2 20.6
2001-02   .508 11.3 3.3 2.2 22.6
2002-03   .513 11.8 3.6 2.7 21.3
2003-04   .501 12.2 3.0 2.6 21.9
2004-05   .496 12.0 2.9 2.8 21.9
2005-06   .484 11.4 3.3 2.1 19.2
2006-07   .546 11.2 3.6 2.5 21.1
2007-08   .497 12.0 3.0 2.1 20.5
2008-09   .504 11.4 3.8 1.8 20.7
2009-10   .518 11.6 3.6 1.7 20.6
2010-11   .500 11.3 3.4 2.4 17.1
2011-12   .492 11.5 2.9 1.9 19.7
2012-13   .502 11.9 3.2 3.2 21.3
2013-14   .490 12.0 3.7 2.3 18.7
2014-15   .512 11.4 3.7 2.4 17.3
2015-16   .488 10.5 3.8 1.8 12.2
Career    .506 11.5 3.2 2.3 20.1

It's really not that shocking.  If you give a player 12 more minutes per night, he's going to put up more stats.  Obviously.   If you take a guy like Michael Carter Williams and cut his minutes in half, don't be shocked that he puts up half the stats, but his per-minute rebounding and assist rates stay consistent. 

Some people may misuse the stat or misinterpret it, but it absolutely has value when looking at a guy like Bradley who might average 28-31 minutes vs a guy like Jimmy Butler who might average 39 minutes.    Adjust them both to 36 to have a better look at how they compare. 

There might be some belief that these guys averaging 25-30 minutes would be incapable of putting up bigger numbers in 36-40 minutes, but that has never seemed to be the case.  Bismack Biyombo is a great example of a player who consistently put up solid stats in his limited minutes.  His per-36 numbers were pretty consistent.   When he finally got more minutes, he put up the expected stats and got himself a fat contract as a result. 

Always consider role as well, but adjusting minutes is important.  Rondo's stats were consistent even when he was receiving a ridiculous 43 per night.   Thomas's stats were consistent even when he was down to 25 minutes on the Suns.   Duncan's stats were consistent whether he was getting 28 or 40 minutes per game.   It's not a misused stat.  Some people just might not understand it.  Others might just be grumpy that Joel Embiid's per-36 stats are incredible right now.

More examples of where Per-36 was taken into account.


James Harden was considered valuable, in large part, because despite his limited minutes off OKC's bench (31.4), he was putitng up per-36 numbers of 19.3 points, 4.2 assists and 4.7 rebounds.   What might happen if you raised his minutes and gave him a more pronounced role?   We saw it the next year... his minutes were jacked up to 38, his role became more pronounced, and he put up 25.9 points, 5.8 assists, 4.9 rebounds (per-36 of 24.4/5/5/4.6)

Similar situation with Eric Bledsoe.  He was averaging just 20.4 minutes on the Clippers behind Chris Paul.  Putting up mediocre 8.5 points, 3.1 assists, 3 rebound and 1.4 steal stats.    But folks looked at his per-36 numbers and saw that he was putting up Per-36 stats of 15 points, 5.4 assists, 5.2 rebounds and 2.5 steals.  What might Bledsoe do with more minutes and a more consistent role?  We saw the next season as he ended up on the Suns averaging 17.7 points, 5.5 assists, 4.7 rebounds and 1.6 steals.

Also saw it with Anthony Davis.  He averaged just 28 minutes per game as a rookie... putting up 13.5 points, 8.2 rebounds, 1 assist, 1.2 steals and 1.8 blocks.   I wondered what he might do with more minutes and pointed out that his per-36 stats were 17 points, 10.2 rebounds, 1.2 assists, 1.5 steals and 2.2 blocks.   He's since seen his minutes rise into the 35-36 range and his rebounding, assist, steal and block have remained relatively consistent.  His per-36 scoring has increased from 17 to 24 as his offensive role has increased.  If you're curious what kind of stats Anthony Davis can put up with inflated 40+ minutes and an even more heightened role, check out his first two losses of this season when his minutes were jacked up over 40 and he had no real help.

Jokes aside, 3 games (with an obscenely high usage rate) is not enough of a sample size to judge Joel Embiid.  But as this season progresses, if he averages in the high 20s, it will be interesting to look at his stats.  Adjusting them to 36 will be helpful in guessing what kind of player he might be when he gets minutes like Anthony Davis. 

[LarBrd33]

Tim Duncan is another phenomenal example.   In Duncan's prime he was averaging 39-40 minutes per game.   Over his last 6 games, he was down to 28 minutes per game.   His stats were significantly lower.  From averaging 25/12/3 to averaging 15/9/2.    But look at the guy's Per-36 numbers it's EERIE how consistent he was regardless of whether he was getting 40 or 28 minutes per game.  Just look at it... it's all the proof you need that Per-36 has value.   Same player... just less minutes:

Your example is exactly how per 36 gets misused. Duncan played 39-40 minutes per game and was effective for that period of time. His last 6 years he couldn't perform at that level for that period of time.

He had his minutes limited to prolong his career.   The fact is, in the minutes he played, he was still the same Tim Duncan for the most part. 

Come playoff time, his minutes once again exceeded 35.   Casuals might have been flabberghasted by a resurgent Tim Duncan averaging 18.1 points, 10.2 rebounds, 1.2 assists, 1.6 blocks in 35 minutes back in the 2013 playoffs.   Casuals were likely dumfounded to see Duncan put up 18 points, 11.1 rebounds, 3.3 assists, 1.4 blocks and 1.3 steals in 35.7 minutes in the 2015 playoffs.   But folks who knew better, like myself, understood that Duncan was Duncan.  And he had been consistently putting up those numbers for 20 years.  Give him more minutes, he'd put up more stats.

Of course, rarely do fans use Per-36 to discuss players like Tim Duncan.  We use them to talk about young guys who have lesser roles and less minutes and contemplate what they might do with more minutes and a more pronounced role.  It has value for that to a point.  Obviously, a guy averaging less than 15 minutes exclusively coming off the bench... gonna be hard to figure out what they might do with more minutes.   But players averaging 25-31 (like Avery Bradley)... it's interesting to use that stat to discuss what he might do with 36+ minutes and a greater role.

I'm also still very curious to see what Kelly Olynyk might do with a greater role and more minutes.  I'm convinced there's a scenario where he ends up averaging 16-17 points for a team some day.

[celticsclay]

To be honest this was one of the more interesting threads that I was really enjoying LB with some great points by Moranis, Phosita, BDM, GCO and others about a pretty interesting statistic and you have now hijacked it by writing overly verbose posts and responding to yourself.

If you want to take the stance that per 36 are as relevant as ever despite the fact that not a single frontcourt player averaged 36 minutes last year and the amount of players playing that many minutes is less than 20% what it once was, sure go for it.

Please don't derail this thread with embiid talk or by bumping your own points. 




 

[celticsclay]

I would love to hear from the users of the per 36 stat in regards to foul trouble, tough to project out stats for many big guys,  when asked to play extended minutes they simply wouldn't be able to due to being in foul trouble and sitting on the bench.

This is another nuance and short coming of the per 36 stat and why it is used less often by people into advanced stats.

Alen Len, at one point a fairly intriguing prospect at Center for the suns has some impressive raw stats. He is averaged 9 points and 7.5 rebounds and a block in just 23 minutes a game last year. This would be fairly robust numbers of 14 points and 11.7 rebounds and over a block per 36.  However, he  can't stay on the court because he fouls 3 times in those 23 minutes and is constantly in foul trouble. To stay on the court longer would he have to be less aggressive on rebounds and blocks? This year his foul troubles have continued early on and seems like he may never play even 30 minutes per game in his career.

Edit: Bogut is an even better example of this last year 7.3 fouls per 48 minutes. He would average 5.85 fouls per 36 and would foul out just about every game...

[Moranis]

I would love to hear from the users of the per 36 stat in regards to foul trouble, tough to project out stats for many big guys,  when asked to play extended minutes they simply wouldn't be able to due to being in foul trouble and sitting on the bench.

This is another nuance and short coming of the per 36 stat and why it is used less often by people into advanced stats.

Alen Len, at one point a fairly intriguing prospect at Center for the suns has some impressive raw stats. He is averaged 9 points and 7.5 rebounds and a block in just 23 minutes a game last year. This would be fairly robust numbers of 14 points and 11.7 rebounds and over a block per 36.  However, he  can't stay on the court because he fouls 3 times in those 23 minutes and is constantly in foul trouble. To stay on the court longer would he have to be less aggressive on rebounds and blocks? This year his foul troubles have continued early on and seems like he may never play even 30 minutes per game in his career.

Edit: Bogut is an even better example of this last year 7.3 fouls per 48 minutes. He would average 5.85 fouls per 36 and would foul out just about every game...

Bogut had 5.5 fouls per 36 last year, but it was an anomaly.  He is back to his normal range around 4.4 in his three games this year.

I really don't get what you are saying in this thread, no one is saying that Bogut can play 36 minutes a night or that he should.  But normalizing Bogut's stats to per 36 allows you to better compare him to other players in similar roles i.e. a starter who plays about half the game.  You could do it with per minute stats, but it is much harder to understand what .25 points per minute actually means or how much different that is than .33 points per minute.  That is what per 36 do, they just normalize things.

[celticsclay]

I would love to hear from the users of the per 36 stat in regards to foul trouble, tough to project out stats for many big guys,  when asked to play extended minutes they simply wouldn't be able to due to being in foul trouble and sitting on the bench.

This is another nuance and short coming of the per 36 stat and why it is used less often by people into advanced stats.

Alen Len, at one point a fairly intriguing prospect at Center for the suns has some impressive raw stats. He is averaged 9 points and 7.5 rebounds and a block in just 23 minutes a game last year. This would be fairly robust numbers of 14 points and 11.7 rebounds and over a block per 36.  However, he  can't stay on the court because he fouls 3 times in those 23 minutes and is constantly in foul trouble. To stay on the court longer would he have to be less aggressive on rebounds and blocks? This year his foul troubles have continued early on and seems like he may never play even 30 minutes per game in his career.

Edit: Bogut is an even better example of this last year 7.3 fouls per 48 minutes. He would average 5.85 fouls per 36 and would foul out just about every game...

Bogut had 5.5 fouls per 36 last year, but it was an anomaly.  He is back to his normal range around 4.4 in his three games this year.

I really don't get what you are saying in this thread, no one is saying that Bogut can play 36 minutes a night or that he should.  But normalizing Bogut's stats to per 36 allows you to better compare him to other players in similar roles i.e. a starter who plays about half the game.  You could do it with per minute stats, but it is much harder to understand what .25 points per minute actually means or how much different that is than .33 points per minute.  That is what per 36 do, they just normalize things.

Well I do disagree with you that per 36 is only about normalizing stats. Like I said, when it first started being used 36 numbers was actually what good starters and stars played, or more. So people often use it as a way of thinking if this guy got normal starting minutes he might actually even average 25 points instead of 21. If memory serves correct, I had these same thoughts looking at cards made by skybox in 1990 (or some similar per minute stat). I was huge into sports cards.

Now no frontcourt players play that many minutes and only an extremely small amount of guards do (it is like what 4% of the league at 7 players) so that portion of the stat has definitely become very outdated. If you don't think per 36 numbers at least have some element of "this is what the guy would do with full-time minutes" implicitly attached to them we can just agree to disagree.  (Though that seems like a ridiculous stance to take and if I was inclined could find hundreds examples on the board of people saying "if you gave him starters minutes he would average xx per 36."

The other part is that it is misused frequently. While perhaps you and LB and other people that understand some advanced statistics, the players we see most frequently used to have their per 36 numbers is a player like Sully. It would be unfair to compare the per 36 rebounding numbers of Sully and Horford because Horford can actually play 32 minutes a game so we know what he gets in 32. I don't think anyone would argue that giving Sully 10 extra minutes he couldn't maintain his rebounding prowess cause he would get tired.

I'll also add that I think your are arguing with me just to argue, because I see the kinds of numbers you use when debating and it is a lot more advanced stuff like true shooting percentages, rebound percentage on offense vs defense etc and per 36 is one of the last things I have seen you use. So your actions kind of support my point that this is not a super useful stat. 

[celticsclay]

You make good points.

I think for today it probably makes more sense to use a lower per-minute measurement, since the average minutes played for a starter is significantly lower than it once was.

Should probably be something along the lines of per-30 or per-32.  You're right that very few players actually average 36 minutes a game, so it's kind of a weird hypothetical to use as a measuring standard ("If this guy played 36 minutes a game he'd average ___").

I think per-minute stats are useful for assessing the rate at which players affect the box score.  Of course, we could also use things like TRB%, AST%, USG, and so on.  But putting it in terms of a box score is probably a more affecting illustration for most people.

As Moranis said, as with all stats context is required in order for it to be useful at all.


The other way we could go is just normalize the use of Per-48.  It's very rare that a player actually plays 48 minutes in a game, and then only when there is at least one overtime period.  Still, that could be one way to unmoor the numbers we're talking about from the misleading perception that the per-minute average is something that a player could actually plausibly achieve.

Phosita is actually making the point I was trying to make in a well spelled out way here that I 100% agree with the bolded.

[LarBrd33]

To be honest this was one of the more interesting threads that I was really enjoying LB with some great points by Moranis, Phosita, BDM, GCO and others about a pretty interesting statistic and you have now hijacked it by writing overly verbose posts and responding to yourself.

If you want to take the stance that per 36 are as relevant as ever despite the fact that not a single frontcourt player averaged 36 minutes last year and the amount of players playing that many minutes is less than 20% what it once was, sure go for it.

Please don't derail this thread with embiid talk or by bumping your own points.

The "no frontcourt player played 36 minutes" is irrelevant.   It's a way of averaging out minutes.   Adjust down Bradley and Butler's minutes to 20 if it makes you feel better.  The point still stands that if a player gets 40 minutes while another gets 28, the guy getting 40 minutes will have more opportunities for stats.   You're misunderstanding the purpose of per-36.  We use it, because it's widely available on stat sites.  In NCAA they seem to use Per-40 instead.

It doesn't take into account role, system and pace, but it gives you a quick-glance look at what kind of stats a player might put up with less or more minutes than they are currently averaging.

[mmmmm]

I'm not sure of the point of this thread's titled assertion.  Are we supposed to accept that "per game" stats are LESS 'misused' than per-36?

Because, um... No.

As others have said, per-36 is simply a way to normalize rates.  Like with any stat, it is misleading without context.

And yes, part of the 'context' that one needs to know when using per-36 is the likelihood of a player every playing close to 36 minutes or to only use it, as someone else pointed out, to compare players in similar roles.

The same contextual limitations also apply to using almost all other stats.

I would assert that 'per game' stats are the most misused stats of today, yesterday and probably next week as well.

[mmmmm]

I would love to hear from the users of the per 36 stat in regards to foul trouble, tough to project out stats for many big guys,  when asked to play extended minutes they simply wouldn't be able to due to being in foul trouble and sitting on the bench.

I do believe that should be captured in the PF/36 rate.  If it is higher than 6 ...

[fairweatherfan]

This argument again? Do people look at ERAs for relief pitchers and say "that's bogus, that guy would give up way more runs if he pitched a full 9 innings"?

Per-36 is not a projection, and it's never been one. The only instance where you could call it a projection is if a guy had played fewer than 36 minutes that season.

I mean I can see arguing for a lower # as minutes have dropped - that's pretty reasonable - but the idea that it's projecting what a guy would do in 36 mpg is just a misreading.

[Celtics4ever]

PER36 is a projection not a stat because only real numbers that are produced are stats.