The most misused stat of today: Per 36 numbers

[fairweatherfan]

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

The definition isn't accurate to begin with, but Per-36 is just real numbers adjusted to equalize for different playing time.  Do you think ERA is a projection? Is true shooting % a projection, and if so, what is it projecting?

[celticsclay]

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.

For the first point, it is a little different than that and I think you know this. If baseball regularly had a strikeout per 7.1 innings stat they flashed they screen all the time during baseball games because the average starter use to go 7.1 innings every start, and then you fast forward 15 years and the average starter was going 6.1 but they were still flashing that stat I think that would be a lot more similar (then you add in that commentators started talking about relievers and their per 7.1 numbers if they were to be turned into starters)

[LarBrd33]

The bottom line is that it's useful in a couple instances.

One is if you're looking at a guy who averages like 20-30 minutes and want to speculate what kind of stats he could put up more minutes and a more pronounced role.   Boston tends to have a pretty balanced roster.  We were overloaded with mediocre big men.   I don't think it was accurate to suggest Jared Sullinger was significantly better than Kelly Olynyk just because Sullinger averaged 13 points and 8 rebounds while Olynyk was averaging 10 points and 4 rebounds.   If Sully was averaging 27 minutes while Olynyk averaged 20 minutes, it's fair to even out both of their minutes to make a comparison.  36 minutes is the standardized number available to us.  So if you average them both out to 36, you'd see those guys were basically dead even statistically.   

The other is to temper enthusiasm about a player putting up huge stats with huge minutes.  Jimmy Butler lead the entire league in minutes during his break-out.  Some people here had the audacity to call him a Superstar and pointed to his big stats in comparison to guys like Crowder and Bradley as proof.  Consider the fact he had more minutes than any player in the league that year.  More minutes = more stats.    The same issue was present when a guy like Rondo averaged 43 minutes in the playoffs.  43 minutes is obscene.  It shouldn't be a surprise that he put up 25% less stats when getting 25% less minutes.   If you looked at Rondo's per-minute production during that playoff run, it was basically the same as any other game. 

[celticsclay]

The bottom line is that it's useful in a couple instances.

One is if you're looking at a guy who averages like 20-30 minutes and want to speculate what kind of stats he could put up more minutes and a more pronounced role.   Boston tends to have a pretty balanced roster.  We were overloaded with mediocre big men.   I don't think it was accurate to suggest Jared Sullinger was significantly better than Kelly Olynyk just because Sullinger averaged 13 points and 8 rebounds while Olynyk was averaging 10 points and 4 rebounds.   If Sully was averaging 27 minutes while Olynyk averaged 20 minutes, it's fair to even out both of their minutes to make a comparison.  36 minutes is the standardized number available to us.  So if you average them both out to 36, you'd see those guys were basically dead even statistically.   

The other is to temper enthusiasm about a player putting up huge stats with huge minutes.  Jimmy Butler lead the entire league in minutes during his break-out.  Some people here had the audacity to call him a Superstar and pointed to his big stats in comparison to guys like Crowder and Bradley as proof.  Consider the fact he had more minutes than any player in the league that year.  More minutes = more stats.    The same issue was present when a guy like Rondo averaged 43 minutes in the playoffs.  43 minutes is obscene.  It shouldn't be a surprise that he put up 25% less stats when getting 25% less minutes.   If you looked at Rondo's per-minute production during that playoff run, it was basically the same as any other game.

I think the second part is reasonable.

The first part is not a great example. Sully and KO I don't think will ever be capable of playing the 35 minutes a game you mention Butler played. KO doesn't play much more 23-25 minutes even when there are no other big men on the roster. I personally, think he gets tired cause he lacks strength to body up other big man. Maybe the staff views him as injury prone. For whatever reason, for a player like KO it is kind of a useless stat since it puts him in a context he cant ever be put into (a player playing monster minutes).

The same thing is true of Sully. He flat out sucks if you get his minutes above 27 and they actually tailed his minutes back to a lower level for him to be more effective in his last season here. You really can't scale his numbers up to compare him to anyone else.

So yes if you want to compare players that are starter quality and have shown capable of handling a lot of minutes like Bradley and Crowder versus Butler, sure there is some value there.

However, in many other situations it is misused. You add in the fact that it was developed at a time when players used to all play 36 minutes if they were a star or good starter and it makes it even more confusing (although everyone here is claiming they would NEVER do that).

[LarBrd33]

The bottom line is that it's useful in a couple instances.

One is if you're looking at a guy who averages like 20-30 minutes and want to speculate what kind of stats he could put up more minutes and a more pronounced role.   Boston tends to have a pretty balanced roster.  We were overloaded with mediocre big men.   I don't think it was accurate to suggest Jared Sullinger was significantly better than Kelly Olynyk just because Sullinger averaged 13 points and 8 rebounds while Olynyk was averaging 10 points and 4 rebounds.   If Sully was averaging 27 minutes while Olynyk averaged 20 minutes, it's fair to even out both of their minutes to make a comparison.  36 minutes is the standardized number available to us.  So if you average them both out to 36, you'd see those guys were basically dead even statistically.   

The other is to temper enthusiasm about a player putting up huge stats with huge minutes.  Jimmy Butler lead the entire league in minutes during his break-out.  Some people here had the audacity to call him a Superstar and pointed to his big stats in comparison to guys like Crowder and Bradley as proof.  Consider the fact he had more minutes than any player in the league that year.  More minutes = more stats.    The same issue was present when a guy like Rondo averaged 43 minutes in the playoffs.  43 minutes is obscene.  It shouldn't be a surprise that he put up 25% less stats when getting 25% less minutes.   If you looked at Rondo's per-minute production during that playoff run, it was basically the same as any other game.

You add in the fact that it was developed at a time when players used to all play 36 minutes if they were a star or good starter and it makes it even more confusing (although everyone here is claiming they would NEVER do that).

I don't think the number 36 matters.   I think that's a weird thing to focus on.  It's just a standardized minute number.  It would serve the exact same purpose if I said, "Rajon Rondo's stats were inflated because he played 43 minutes per game... average out those minutes to 20, and it's roughly the same stats he put up per 20 minutes any other season".    It would have the exact same purpose if I said, "Average both Sullinger and Olynyk's minutes to 20 per game, and they put up nearly identical stats".     It would serve the exact same purpose if I said, "Average Butler and Crowder's minutes down to 20, their production isn't THAT much different"...

Sites like basketball-reference just have per-36 easily available.  They also have Per 100 possessions ... so we could use that instead to compare players and it would serve roughly the same purpose.   

Start a petition to change Per-36 to Per-25 or something if it makes you feel better.  There were only about 10 guys in the league who actually averaged around 36 minutes last season:  Harden, Lowry, Butler, Caldwell-Pope, Hayward, Wall, Middleton, Derozan, Durant, Marcus Morris and Lillard. 

The idea is... if we took a guy who is averaging 25 and boosted his minutes to 36, would he put up comparable stats?

Alternatively, the idea is...

14.5 points, 3.7 rebounds, 1.8 assists and 1.4 steals

vs

14.9 points, 3.8 rebounds, 1.4 assists, 1.1 steals

The top is what Kentavious Caldwell-Pope put up last season in 36.7 minutes per game.

The bottom is what Avery Bradley put up in 2014 in 30.9 minutes per game.    Would Avery had exceeded Pope's stats with an extra 6 minutes per game?  Sure.  Would Pope have put up less stats if he hadn't averaged the 3rd highest minutes in the league?  Definitely. 

Now look at this:

15.2 points, 3.7 assists, 2.4 rebounds, 1 steal.

That's what Isaiah Thomas put up in 25.7 minutes per game on the Suns.  That's a full 11 minutes less what Caldwell-Pope got last year.   Shouldn't we consider minutes when looking at these stat lines?  Wouldn't it stand to reason that Isaiah Thomas would put up better stats than that with more than 25.7 minutes?  Yup. 

[celticsclay]

The bottom line is that it's useful in a couple instances.

One is if you're looking at a guy who averages like 20-30 minutes and want to speculate what kind of stats he could put up more minutes and a more pronounced role.   Boston tends to have a pretty balanced roster.  We were overloaded with mediocre big men.   I don't think it was accurate to suggest Jared Sullinger was significantly better than Kelly Olynyk just because Sullinger averaged 13 points and 8 rebounds while Olynyk was averaging 10 points and 4 rebounds.   If Sully was averaging 27 minutes while Olynyk averaged 20 minutes, it's fair to even out both of their minutes to make a comparison.  36 minutes is the standardized number available to us.  So if you average them both out to 36, you'd see those guys were basically dead even statistically.   

The other is to temper enthusiasm about a player putting up huge stats with huge minutes.  Jimmy Butler lead the entire league in minutes during his break-out.  Some people here had the audacity to call him a Superstar and pointed to his big stats in comparison to guys like Crowder and Bradley as proof.  Consider the fact he had more minutes than any player in the league that year.  More minutes = more stats.    The same issue was present when a guy like Rondo averaged 43 minutes in the playoffs.  43 minutes is obscene.  It shouldn't be a surprise that he put up 25% less stats when getting 25% less minutes.   If you looked at Rondo's per-minute production during that playoff run, it was basically the same as any other game.

You add in the fact that it was developed at a time when players used to all play 36 minutes if they were a star or good starter and it makes it even more confusing (although everyone here is claiming they would NEVER do that).

I don't think the number 36 matters.   I think that's a weird thing to focus on.  It's just a standardized minute number.  It would serve the exact same purpose if I said, "Rajon Rondo's stats were inflated because he played 43 minutes per game... average out those minutes to 20, and it's roughly the same stats he put up per 20 minutes any other season".    It would have the exact same purpose if I said, "Average both Sullinger and Olynyk's minutes to 20 per game, and they put up nearly identical stats".     It would serve the exact same purpose if I said, "Average Butler and Crowder's minutes down to 20, their production isn't THAT much different"...

Sites like basketball-reference just have per-36 easily available.  They also have Per 100 possessions ... so we could use that instead to compare players and it would serve roughly the same purpose.   

Start a petition to change Per-36 to Per-25 or something if it makes you feel better.  There were only about 10 guys in the league who actually averaged around 36 minutes last season:  Harden, Lowry, Butler, Caldwell-Pope, Hayward, Wall, Middleton, Derozan, Durant, Marcus Morris and Lillard. 

The idea is... if we took a guy who is averaging 25 and boosted his minutes to 36, would he put up comparable stats?

Alternatively, the idea is...

14.5 points, 3.7 rebounds, 1.8 assists and 1.4 steals

vs

14.9 points, 3.8 rebounds, 1.4 assists, 1.1 steals

The top is what Kentavious Caldwell-Pope put up last season in 36.7 minutes per game.

The bottom is what Avery Bradley put up in 2014 in 30.9 minutes per game.    Would Avery had exceeded Pope's stats with an extra 6 minutes per game?  Probably.  Would Pope had put up less stats if he hadn't averaged the 3rd highest minutes in the league?  Definitely.

bangs head. You just ignored the entire substance of my point (that sully and ko are the kind of guys that illustrate the issues with per 36 while giving you credit for a good example in crowder/bradley vs butler ) and decided to focus on a tiny point of what i said and jump off on it.

[LarBrd33]

Clay, are we no longer dating?

[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.

For the first point, it is a little different than that and I think you know this. If baseball regularly had a strikeout per 7.1 innings stat they flashed they screen all the time during baseball games because the average starter use to go 7.1 innings every start, and then you fast forward 15 years and the average starter was going 6.1 but they were still flashing that stat I think that would be a lot more similar (then you add in that commentators started talking about relievers and their per 7.1 numbers if they were to be turned into starters)

No, they're both rates adjusted over a set period, and pitchers did use to routinely pitch a full 9 innings, which has dwindled over the years - Cy Young pitched 749 complete games, Sandy Koufax 137 in just 12 years, Roger Clemens pitched 118, now the active leader has only 38. ERA has been used throughout (well, for part of Cy's career). So it's a very similar shift.

If the argument is that people misinterpret per-36 to be a projection more often than ERA, then sure, I agree, but that's not the stat's fault, it's the people misinterpreting it. But both stats are the same form of transformation operating on the same principle.

[celticsclay]

Clay, are we no longer dating?

You left the toilet seat up. 

[celticsclay]

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.

For the first point, it is a little different than that and I think you know this. If baseball regularly had a strikeout per 7.1 innings stat they flashed they screen all the time during baseball games because the average starter use to go 7.1 innings every start, and then you fast forward 15 years and the average starter was going 6.1 but they were still flashing that stat I think that would be a lot more similar (then you add in that commentators started talking about relievers and their per 7.1 numbers if they were to be turned into starters)

No, they're both rates adjusted over a set period, and pitchers did use to routinely pitch a full 9 innings, which has dwindled over the years - Cy Young pitched 749 complete games, Sandy Koufax 137 in just 12 years, Roger Clemens pitched 118, now the active leader has only 38. So it's a very similar shift.

If the argument is that people misinterpret per-36 to be a projection more often than ERA, then sure, I agree, but that's not the stat's fault, it's the people misinterpreting it. But both stats are the same form of transformation operating on the same principle.

Is this just a round about way of saying you thinking era is a bad stat because there are less complete games? That is an interesting rebuttal.

[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.

For the first point, it is a little different than that and I think you know this. If baseball regularly had a strikeout per 7.1 innings stat they flashed they screen all the time during baseball games because the average starter use to go 7.1 innings every start, and then you fast forward 15 years and the average starter was going 6.1 but they were still flashing that stat I think that would be a lot more similar (then you add in that commentators started talking about relievers and their per 7.1 numbers if they were to be turned into starters)

No, they're both rates adjusted over a set period, and pitchers did use to routinely pitch a full 9 innings, which has dwindled over the years - Cy Young pitched 749 complete games, Sandy Koufax 137 in just 12 years, Roger Clemens pitched 118, now the active leader has only 38. So it's a very similar shift.

If the argument is that people misinterpret per-36 to be a projection more often than ERA, then sure, I agree, but that's not the stat's fault, it's the people misinterpreting it. But both stats are the same form of transformation operating on the same principle.

Is this just a round about way of saying you thinking era is a bad stat because there are less complete games? That is an interesting rebuttal.

No, I think they're both fine, I'm just not sure why people criticize Per-36 on the erroneous basis that it's a projection of performance over that period, but don't apply the same standard to ERA which is a very similar stat that's nearly a century older.

On top of that, data generally shows that Per-36 stays pretty consistent even when players get more/less minutes, so it's not even likely to be that far off if it is used as a projection. It really only gives goofy results for low minute guys.

[LarBrd33]

Clay, are we no longer dating?

You left the toilet seat up. 

That's a real shame, Clay.  The two of us have been involved in a passionate entirely platonic digital bromance.  It was a very fulfilling relationship... I'd make home dingers... you'd publicly compliment me on my home dingers...  It had everything I look for in a digital bromance:  unabashed exaggerated admiration.   It only lasted a couple weeks, but boy howdy was it delightful.  Everything was going great until you took a dump on Okafor against my wishes.  Not on my watch, Clay.  I guess the two of us will always have Embiid... at least until he turns into a pumpkin and the Philly nightmare continues.    If you adjust our bromance to 36 minutes, I think it was statistically more impactful than the a lot of more celebrated bromances. 

[celticsclay]

Clay, are we no longer dating?

You left the toilet seat up. 

That's a real shame, Clay.  The two of us have been involved in a passionate entirely platonic digital bromance.  It was a very fulfilling relationship... I'd make home dingers... you'd publicly compliment me on my home dingers...  It had everything I look for in a digital bromance:  unabashed exaggerated admiration.   It only lasted a couple weeks, but boy howdy was it delightful.  Everything was going great until you took a dump on Okafor against my wishes.  Not on my watch, Clay.  I guess the two of us will always have Embiid... at least until he turns into a pumpkin and the Philly nightmare continues.    If you adjust our bromance to 36 minutes, I think it was statistically more impactful than the a lot of more celebrated bromances.

We can get back together

[LooseCannon]

The correct use of per 36 numbers is in making comparisons between players or seasons in which there is a large disparity in minutes per game.  The research tends to show that players tend to improve their production when their minutes are improved, an effect that has been referred to as the Millsap Doctrine.  The argument that reserves have inflated stats due to facing other reserves is basically false.

That being said, there are many advanced metrics which are superior to per 36 numbers, which are based on box score stats that don't capture a player's defensive contributions.  However, per 36 numbers are vastly superior to per game numbers.  Anyone using per game numbers to argue which player is better is basically an idiot using woefully outdated statistics.

[RockinRyA]

Per 36 stat's significance and accuracy goes higher as the actual minute gets higher.