I'll add more to this study as I get time, but looking at two players brought up in recent threads, and then observing some more (Marrese Speights, for example, because his high PER has always bothered me), there appears to be a consistent correlation between a high usage rate and a high PER. 3 pt attemps also seem to inflate PER, at least in terms of TS%. Player A stats/36 with new team; generally considered a mediocre player on this board. FG - 45% FGA - 9.4 3pt - 40% 3ptA - 3.8 Points - 11.9 Assists - 6.2 Reb - 5.0 Stl - 1.3 TO - 1.9 TS - 55% Ortg - 114 Drtg - 102 Usage - 16.3 PER - 14.6 Player B/36 - MVP candidate FG - 43% FGA - 14.9 3pt - 36% 3ptA - 7.1 Points - 21.5 Assists - 5.4 Reb - 4.5 Stl - 2.0 TO - 2.8 TS% - 60 Ortg - 118 Drtg - 102 Usage - 26.1 PER - 23.0 Player C/36 - Marreese Speights 2009-10 FG - 48% FGA - 15.7 3pt - 0% 3ptA - 0.2 Points - 18.9 Assists - 1.3 Reb - 8.9 Stl - 1.0 Blk - 1.2 TO - 1.8 TS% - 52% Ortg - 107 Drtg - 109 Usage - 25.1 PER - 17.6
There is definitely a connection between higher usage rate and higher PER. You need more possessions to generate more assists and, especially, more points. PER rewards more scoring especially, which is one criticism of the metric. Most players will see their PER rise with a higher usage rate. But most players can't put up a 20+ PER no matter what their usage rate is. Looking at usage rate in addition to PER is always valuable.
Hollinger actually considered Usage Rate to be a positive in most cases... reading a past book (can't remember which season it is... 2003-04?) he wrote about how some players' low usage indicated an inability to create his own shot, etc. Ed O.
I think that a usage/PER ratio might actually be of value when assessing a player's value. Then again, maybe not. For those who haven't figured it out: Player A is Hedo Turkoglu and his stats back in his safe environment of Orlando Player B is Manu Ginobli, who has been mentioned as a potential MVP candidate.
I'm beginning to wonder if Usage Rate is PER's inherent flaw in terms of inflating perceived value, though. I think that Hedo Turkoglu has the ability to create his own shot. As a person who hasn't thought highly of Hedo in the past (to say the least), I actually can see a lot of value in his 'hard' stats and how they positively impact his team. Perhaps he just needed a coach that would give him the opportunity to have a higher usage rate? Or, perhaps the coaches he has had have maximized his value in terms of PER? The comparison intrigues me, regardless. I don't see any league where Marreese Speights is considerably more valuable than Hedo Turkoglu. I say this as a (formerly?) rabid PER partisan.
I wonder if multiplying PER by the decimal of Usage Rate (0.17 for 17% for example) would produce anything useful as a stat... you'd have to apply it over a lot of differen players, ones you know are good (LeBron, Jordan) and one you know aren't (Oberto, Marks) to see if there's any interesting normalizing or denormalizing of the PER stat.
It might be worth a shot. As Ed noted, Hollinger himself basically believes that Usage is a subjective measure of a player's ability to create his own shot. I'm just not positive that is the case. Seems odd of Hollinger to say that considering a player like Kevin Love has a 23.5 usage rate. I don't think anybody in the NBA would say that Kevin Love is a threat to create his own offense, outside of an offensive rebound.
I think you have to be a bit careful about usage%, though I think you're doing a good exercise and am interested in the results. The def of usage rate is based heavily on shots and team pace so there's a difference b/w, say, a PG pounding the ball, and passing to someone for a hot-potato shot (the shooter gets the usage rate bump, even though he barely touched the ball); or someone like Love who gets a lot of shots from ORebs (and bumps the usage), vs. someone like Redick who gets fed on a high percentage of his shots. Purely off the top of my head, I'd like to see a correlation b/w usage rate, %assisted, and PER...and maybe even a derivative of %assisted that correlates with PG's and their assist %. That's a bit involved, though. I agree with the premise that just b/c you're getting a high usage% and therefore more chances to up your PER that you're a good player, though I have to think about Minstrel's thought that if you can achieve a high PER regardless of usage% you're special.
My opinion is that if you have a high usg% and a high PER - this means that you are a good player for this system - it does not mean that if you move to a team that has a different system you will be as good a player - and likewise, if you get a low usage% and low-PER - it could be an indication that you are not used properly. So, the places where I would find interest in the correlation are: 1. If you get a reasonably high PER with a low usage% - this means that you are a good player that could be much, much better in a system better tailored for your skills 2. If you get a low PER with high usage% - there is a good chance you are not really worth much as a player Of course, either could come with exceptions - if you use a good 3P shooter that would benefit from catch and shoot situations but you are used as a ball handler - this could lead to high usage% and low PER (because of turn-overs) - so you can not just make these rules without actually watching a player, but running an analysis to see the reasonably good PER players with low usage% is a good way to look for "deals" - on good players that you can get somewhat on the cheap. Take Wes Mathews for example - his rather low PER of 12.3 (not bad for a rookie, but not earth shattering) - combined with a low usage% of 16.5 was an indicator that with more use - he might produce more - something that we see in Portland. Add the fact that he was a rookie with well deserved reputation for being a pretty decent perimeter defender - and you can see why the Blazers might have been attracted to him. A more interesting case is our boy Anthony "Magic" Randolph - his usage% was the same in GSW as it is in NYK - but his efficiency was way better in GSW. Why? I do not know - but it seems that something that D'antoni does with him is not working too well compared to what Don Nelson did. This is a cause for concern - because I consider D'antoni to be much more of a structured coach than Nelson - this might be a real red flag about AR being able to be productive in a system that is based on something more than just running around and shooting as he sees fit.
I was wondering how accurate a Usage/PER ratio would be in assigning value to a player? Using my three examples from earlier, you get the following. Hedo 1.11 Manu 1.13 Speights (2009-10) 1.42 Obviously this has its own flaws, and there should be some sort of usage multiplier, but ideally, the closer to 1.0 the player, the more 'efficient' he is in how he is used. For a player under 1.0, and with a PER of above 15, I'll also offer that the player isn't being completely utilized by his coach. I don't think anyone would say that Hedo is 'better' than Manu, but then again, neither is Speights, and this Usg/PER ratio may more fairly acount for the PER inflation that a high usage rate seems to consistently yield. One more quick example of this is Nic Batum last season. He ended up with a PER of 'only' 17.3 in his 37 games, yet his Usg/PER ratio is 16.4/17.3, which = 0.948. Considering that Nic had a TS% of .646, and an Ortg of an incredible 127, PER doesn't seem to accurately illustrate just how efficient he was in the role that was asked of him. Perhaps Nate didn't know how to use Nic because Nic was out for the first half of the season, but Nic outperformed what you expect from a guy in that role. Anyhow, I'm a stat geek, so I'm basically just brainstorming ways to more accurately use comparative statistics.
PER isn't purely about efficiency. It's also about production. A player who is efficient about scoring 30 PPG and getting 12 RPG will end up with a far higher PER than a player who is equally efficient about collecting 12 PPG and 7 RPG. I think that makes sense, as it is harder to maintain efficiency as you take more shots and possessions. That's also the problem, IMO, with dividing Usage by PER. Jordan had PERs around 30 (!) in his prime, but usage rates between 30-35. He'd end up with a similar ratio as Turkoglu and Ginobili. My guess (though it would take a pretty enormous amount of number crunching to check this) is that the vast majority of players would end up with a U/PER of around 1.0-1.1. I think most players tend to gravitate toward "using" possessions in proportion with their ability levels, since coaches determine roles to a large extent.
Player Efficiency Rating Effeciency is included when studying PER and how it applies to a given player. It's a 'per-minute' snapshot of many things, and efficiency is included in that number. I do agree that an inefficient shooter can have an inflated PER simply due to volume of shots he attempts. Also, as I posted previously, I do agree there should be some sort of multiplier to separate player's with a high PER from those with lower PERs.
Yes? I didn't say it wasn't. I said it wasn't purely about efficiency. Actual production matters also. Both in reality and in PER.
I think that the moment you define that it's efficiency per minute, not efficiency per action - PER captures "production" in it by default. If you spend 5 minutes on the court and you are only used once - the "production" is going to be rather low - and it will be reflected in your PER - this is the real correlation between usage and PER.
I disagree. Being involved in the end of more possessions consistently yields a higher PER. Hollinger even says that one problem with PER is how an inefficient shooter can have an inflated PER, simply based on volume of attempts. It's actually very rare to find a player with a PER above 17 and a usage rate lower than that number. Nic's season last year, and Cho's comments on keeping him, tell me that advanced stats are being closely watched by the front office.
Somebody went to the trouble last year of graphing this last year through midseason and noticed what I noticed about PER and usage. He also seems to be somewhat on the Usg/PER idea, although he doesn't express it that way. http://www.hardwoodparoxysm.com/2010/01/25/graphic-offense-a-look-at-usage-and-per-mid-season/ Pretty cool graphs, especially the Usg/PER graphs in the second half of the post. Also notice how rare it is to find a player who has a PER that exceeds USG rate.
Got bored, made a spreadsheet trying out an "Impact Rating" out of Per, Usage, and Minutes Per Game. Here's the results on the entire Blazers lineup: I've sorted by Impact %, which appears to be a sort of Team MVP ranking. What do y'all think?
Repped. Looking at it, I think I've been too hard on Miller. I'll tone it down, because he is balling on offense, and my own bias against him skewed the reality for me. This graph is in line with the type of conversation I was trying to stimulate in this thread. Well done, BC!
