 |
APBRmetrics
The statistical revolution will not be televised.
|
| View previous topic :: View next topic |
| Author |
Message |
Ben F.
Joined: 07 Mar 2005
Posts: 390
Location: MD
|
 Posted: Fri Jun 06, 2008 1:18 pm Post subject: PBP Analysis: Big Men with Range |
 |
|
Midrange jump shots are some of the lowest percentage shots on the floor. Through the end of February, only 12 players with at least 20 midrange attempts (defined as shots between 10 and 21 feet of the basket) shot above the league average eFG% (49.7%) on those attempts. On the surface, then, it doesn't seem to be a good shot. Of course, if you know basketball you know the reasoning isn't that simple: being able to hit a midrange shot with even moderate accuracy changes the dynamics of the matchup. It opens up the floor (and the playbook) - and if you're a big man, it can draw the opponents' best help defender out of the paint.
And it's big men that I decided to focus on for this analysis. I wanted to know what the effect was of having a big man with range on the floor (or stated another way, what disadvantages were incurred by having a big man who couldn't shoot on your team).
My first step was to figure out how to assign ranges to players. I did this by acquiring ESPN's shot chart data, and then looking at the 85th percentile of the distance of a player's shot attempts. In other words, if a player took 85% of his shots below a given distance, I assumed that was "in his range." (I chose 85% to filter out attempts that were not common attempts, that weren't in the player's everyday arsenal. I played around with this number a lot and found 85% to match intuition the best.)
The definition of range is important, because range can be defined in two main ways: as what a player can reliably make, and as what a player will reliably shoot. My theory was that what a player shoots is more important than what he makes, since that is more likely to dictate how the defense guards him. This assumption could be wrong, but I found that using attempts also fell more in line with conventional wisdom about player ranges.
I then had to choose a cutoff for "low range" players. Below what range would a player be categorized as someone that couldn't really hit a midrange jump shot? I chose 15 feet as the cutoff - if your range was below 15 feet you were a low range player. This was done mostly by eye, but I think almost everyone here would agree with the results. Players like Yao Ming, Tim Duncan, Jason Maxiell and Brandon Bass ended up just above the cutoff, as "high range" players, while players like Hilton Armstrong, David Harrison and Melvin Ely were just below. (I think perhaps you could argue for making the "low range" a bit lower, but I'm not sure it would affect the results too much.)
With the categorization of "low range" vs. "high range" players in hand, I then went to the play by play data. I went through every possession in the NBA season, categorizing its results as "low range" if the player on the floor for the offense with the lowest range was a low range shooter.
The results were unsurprising. The low range lineups shot worse, turned the ball over more, offensive rebounded better, and shot about 3.5% worse from the FT line.
| Code: | Type ORTG eFG% TO% OR% FT/FG FT% AST/FGM
High 108.7 50.2% 14.7% 29.6% 0.31 77.4% 59.5%
Low 105.9 49.5% 15.6% 30.7% 0.30 73.8% 57.4%
FT/FG refers to free throws attempted per field goal attempted
|
Those FT% results are actually the most important part, because we know that a lineup's performance at the FT line was almost certainly not affected by a player having poor range. It's a result of the fact that low range big men aren't good foul shooters. We then have to have a suspicious eye when viewing the rest of the results, because they could all just be reflecting the qualities of low range big men, not the effects these big men have on the rest of the lineup. The category most likely affected by this is turnovers: they aren't caused by having low range players on the floor, rather low range players are simply more turnover prone.
So how do we separate out the effects of the players themselves from the effects they have on teammates?
I decided to look at a subset of the low range lineups: just lineups that had low range players who weren't turnover prone. I wanted to see if we looked at players who didn't turn the ball over, what effects would still be there.
These results were far more interesting. The turnover gap disappeared almost completely. The FT% gap narrowed from 3.5% to only 1%. Both of these results indicate that we've filtered out a lot of the low range player's own effects.
| Code: | Type ORTG eFG% TO% OR% FT/FG FT% AST/FGM
High 107.7 50.1% 15.1% 29.8% 0.31 75.8% 58.7%
Low 106.3 49.0% 15.3% 31.5% 0.29 74.8% 57.4%
FT/FG refers to free throws attempted per field goal attempted
|
So what are we left with?
Still a big gap in shooting. Low range lineups shot only 49.0% eFG%, while high range lineups shot 50.1% eFG%. And why is this so fascinating? Because low range players actually shoot a whole lot better than league average. The sample of low range players with low turnovers shot a collective 52.6% from the field. And yet despite these players shooting almost 3% above league average eFG%, when they played their lineups shot 0.7% below league average.
So here's the tricky part: deciding if these effects are actually due to players having low range, or if there's something else going on that strongly correlates with low range players. One idea is passing - the passing ability of big men (measured by assists per teammate FGM) correlates fairly strongly with range. But my results show that there wasn't a large difference in AST/FGM between the two kinds of lineups: low range lineups had 57.4% of their field goals assisted, while high range lineups had 58.7%. It's a difference, but not one likely to account for the huge difference in shooting.
That makes me think these results are fairly close to capturing the effect of low range players on a lineup's shooting.
So now is where I turn it over to everyone else: do these results seems convincing, or do you think there's something else going on that I've failed to account for that would give these results? What confounding factors could be involved here? Are there ways to make this study more robust?
Last edited by Ben F. on Fri Jun 06, 2008 4:56 pm; edited 2 times in total |
|
| Back to top |
|
 |
Mountain
Joined: 13 Mar 2007
Posts: 1527
|
| Posted: Fri Jun 06, 2008 2:24 pm Post subject: |
|
|
I agree it is an important topic and your approach is helpful. I think you could extend on it by looking at subgroups of high-high,high-low and low-low big tandems and them when matched with the full range of groups of perimeter guys (from 3 high down thru low 3).
Is high range always a plus and the more bigs with high range (frequency) the better or is it best in some combination with low range focus (and good FG% at low range too, to stretch floor both directions) and what specific combination? Does it matter if there is a low from a big? And how important is that compared to having a big with high range? What big shot location frequency distribution between inside shots and mid-range & beyond jumpers correlates best with team FG% or team +/-?
I think it would also be useful to look at % of shots taken inside and inside FG% by lineups with the different big shooting range tandems. And actually worth looking at team 3 pt frequency and accuracy too (in addition to the overall FG% effect). The right big will depend on the other pieces of the attack.
And I would think that checking the adjusted +/- of high range bigs would be worthwhile. Big SFs probably ought to be included in the study too. Looking at top 50 adjusted +/- at basketballvalue I think it is fair to say big men with high range have an edge on big men with low range. Bigmen overall are about proportionately represented.
Using the file at countthebasket you could also look by offense / defense splits. High range bigs might be having an offensive impact even more often than seen by check of rollup adjusted +/-, because it sometimes (or often) offset by lesser or negative defensive impact by this type of bigman in comparison with the D provided by "traditional", often low range bigs. Jamison is #1 on adjusted offensive +/- for PFs, Radmanovic 3, Nowitski 4 Boozer 5. At C Miller 2, Duncan 5 (is he high range or not), Amare S. 6th. Okur 13. The positive offensive impact may be more common from the PF spot and a function of usage as well as % of total shots.
Last edited by Mountain on Sun Jun 08, 2008 4:50 pm; edited 1 time in total |
|
| Back to top |
|
|
NickS
Joined: 30 Dec 2004
Posts: 384
|
| Posted: Fri Jun 06, 2008 6:25 pm Post subject: |
|
|
One of the questions I would be curious about is, what happens to the total number of dunk/close shots for the team when a "high range" big man is in.
One of the things people talk about is high-post centers opening driving lanes for guards. If you were to replace Samuel Dalembert (58% inside shots) with Okur (26% inside shots) would the team as a whole take fewer inside shots, or would it allow more dunks for Iguodala? |
|
| Back to top |
|
|
Mountain
Joined: 13 Mar 2007
Posts: 1527
|
| Posted: Sun Jun 08, 2008 5:18 pm Post subject: |
|
|
| Big man FG% from 10-15 feet and its contribution to and correlation with team FG% and team +/- is another topic that would appear worthwhile. I wonder how significant is compared to big man shot frequency from beyond 15 feet. |
|
| Back to top |
|
|
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
|
| Posted: Mon Jun 09, 2008 10:43 am Post subject: |
|
|
Ben F:
| Quote: | | Are there ways to make this study more robust? |
| Quote: | | being able to hit a midrange shot with even moderate accuracy changes the dynamics of the matchup. It opens up the floor (and the playbook) - and if you're a big man, it can draw the opponents' best help defender out of the paint. |
I know it's imposible with your data, bit if you could include some "contested%", something like this study:http://www.82games.com/sacdee.htm(but from an off. perspective), we could look how much contesting and crowding the zone, can cancel the range advantage.
P.D.
A question: what would be your range cathegorization of Horford, Ilgauskas, Wallace, and Gasol? |
|
| Back to top |
|
|
Harold Almonte
Joined: 04 Aug 2006
Posts: 616
|
| Posted: Mon Jun 09, 2008 5:12 pm Post subject: |
|
|
| Since you didn't introduce big mens's usage in the comparissons, and there's no high correlation between eFG% and FT%, it would be also interesting to see the difference in points/g between the two big man range-lineups, bigs's points/g, and their percentage from team scoring. |
|
| Back to top |
|
|
mtamada
Joined: 28 Jan 2005
Posts: 338
|
| Posted: Mon Jun 09, 2008 6:19 pm Post subject: |
|
|
Fascinating results, but with a gap that small, I'd want to see more years of data to see if that gap really holds up.
If it does hold up, yet another possible explanation is that the causality runs in reverse: maybe it's not that longer-ranged big men help their teams' offense, maybe good offensive teams feel they can (or want to) stick some softie jumpshooter out there, instead of feeling they have to get some horse down low , who'd just clog things up for their hot-scoring teammates. |
|
| Back to top |
|
|
Ben F.
Joined: 07 Mar 2005
Posts: 390
Location: MD
|
| Posted: Tue Jun 10, 2008 9:39 pm Post subject: |
|
|
| Harold Almonte wrote: | | A question: what would be your range cathegorization of Horford, Ilgauskas, Wallace, and Gasol? |
I'll have a more in-depth reply to everyone else's responses later, but for now I just wanted to answer this question because it was quick and easy to look up.
I have the ranges of those players as follows (although I don't know which Wallace you're referring to, so I'll list both Rasheed and Ben):
| Code: | Horford: 17 ft
Gasol: 17 ft
Ilgauskas: 19 ft
Rasheed: 26 ft
Ben: 10 ft |
|
|
| Back to top |
|
|
Kevin Pelton
Site Admin
Joined: 30 Dec 2004
Posts: 953
Location: Seattle
|
| Posted: Wed Jun 11, 2008 1:07 am Post subject: Re: PBP Analysis: Big Men with Range |
|
|
| Ben F. wrote: | | So how do we separate out the effects of the players themselves from the effects they have on teammates? |
If you have the possessions separated out this way and you know which players are classified as high and low range, isn't it possible to take the totals and subtract out the statistics of the big men themselves so we're looking solely at the teammate impact?
My assumption (perhaps better termed a stereotype) is that big men with range tend to have a positive effect on offense, but tend to be worse on defense. I'd be curious to see the defensive numbers. |
|
| Back to top |
|
|
Ben F.
Joined: 07 Mar 2005
Posts: 390
Location: MD
|
| Posted: Wed Jun 11, 2008 1:06 pm Post subject: Re: PBP Analysis: Big Men with Range |
|
|
| Kevin Pelton wrote: | | If you have the possessions separated out this way and you know which players are classified as high and low range, isn't it possible to take the totals and subtract out the statistics of the big men themselves so we're looking solely at the teammate impact? |
This would make sense to see the effect on shooting, but I'm not sure how it would work for things like offensive rebounding or TO%. Simply subtracting out the stats of the big men from the numerator doesn't give any meaningful number, and it's not clear to me how you could compensate for that in the denominator of either percentage.
So here is the result for shooting: subtracting out low range players' stats for eFG% shows that high range players playing with low range players shoot 48.8%, while if they aren't playing with low range players they shoot 50.2%.
| Kevin Pelton wrote: | | My assumption (perhaps better termed a stereotype) is that big men with range tend to have a positive effect on offense, but tend to be worse on defense. I'd be curious to see the defensive numbers. |
The numbers bear this out. The following table shows the offensive numbers of the opponents of a given lineup type:
| Code: | Type ORTG eFG% TO% OR% FT/FG FT% AST/FGM
High 107.8 50.2% 15.4% 30.6% 0.31 75.3% 57.6%
Low 106.9 49.5% 15.0% 29.8% 0.30 75.8% 59.2% |
So we see a decently strong defensive effect, carried out by forcing misses, and grabbing slightly more defensive rebounds.
But could these numbers just be a result of the fact that low range lineups tend to play against other low range lineups? Maybe coaches generally try to matchup with the other team, and it wasn't really that a low range lineup is a better defensive lineup, but that they play against worse offensive lineups. I decided to separate it out into matchups:
| Code: | Type ORTG eFG% TO% OR% FT/FG FT% AST/FGM
LvL 105.6 49.2% 15.6% 30.4% 0.29 74.0% 57.8%
LvH 106.9 49.9% 15.7% 31.1% 0.31 73.4% 56.9%
HvL 108.3 49.9% 14.3% 29.1% 0.30 77.6% 60.8%
HvH 108.8 50.5% 15.1% 30.1% 0.32 77.1% 58.4%
LvL = low range offense vs. low range defense
LvH = low range offense vs. high range defense
HvL = high range offense vs. low range defense
HvH = high range offense vs. high range defense |
You can see that a low range defense defends better no matter whether it faces a high or low range lineup. When matched up with a high range offense, low range defenses tend to force more misses and defensive rebound better, but force turnovers at a much lower rate. Against a low range offense, both types of lineups force turnovers at a high rate and the difference in shooting and defensive rebounding gives the low range lineup the advantage. |
|
| Back to top |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
Powered by phpBB © 2001, 2005 phpBB Group
|
FAQ
Search
Memberlist
Usergroups
Register
Profile
Log in to check your private messages
Log in
