Please Explain VORP

Moderator: Doctor MJ

Curmudgeon
RealGM
Posts: 23,937
And1: 4,776
Joined: Jan 20, 2004
Location: Boston, MA

Please Explain VORP 

Post#1 » by Curmudgeon » Sun Jun 25, 2017 1:56 am

Can someone explain to me in layman's terms how VORP works in basketball?
I understand that it takes box score plus minus (BPM), which is how the player did per 100 possessions. If he is plus points per 100 possessions, that's good, zero is average, and -2 is replacement level. Then you multiply that number by the percentage of total minutes played and then by number of team games played divided by 82-- Right?

Why is replacement level -2 instead of -3 or -1?
Why don't playoff games count?

Can someone give a concrete example using an actual NBA player's numbers?

It strikes me that a player on a bad team can never have a good VORP, even though the player is good. For example, suppose I play all 48 minutes and all 82 games on a team that loses every 100 possessions by 5 points. What's my VORP? Does it matter that if I weren't such a good player the team would lose every possession by 10 points? Is there any adjustment for that?
"Numbers lie alot. Wins and losses don't lie." - Jerry West
"You are what your record says you are."- Bill Parcells
"Offense sells tickets. Defense wins games. Rebounding wins championships." Pat Summit
HartfordWhalers
Forum Mod - 76ers
Forum Mod - 76ers
Posts: 34,605
And1: 11,181
Joined: Apr 07, 2010
   

Re: Please Explain VORP 

Post#2 » by HartfordWhalers » Sun Jun 25, 2017 1:27 pm

Curmudgeon wrote:Can someone explain to me in layman's terms how VORP works in basketball?
I understand that it takes box score plus minus (BPM), which is how the player did per 100 possessions. If he is plus points per 100 possessions, that's good, zero is average, and -2 is replacement level. Then you multiply that number by the percentage of total minutes played and then by number of team games played divided by 82-- Right?

Why is replacement level -2 instead of -3 or -1?


http://www.basketball-reference.com/about/bpm.html#vorp is a pretty good description, and includes links to get at what you are asking in terms of replacement level being calibrated to -2.

Curmudgeon wrote:Why don't playoff games count?

Can someone give a concrete example using an actual NBA player's numbers?

It strikes me that a player on a bad team can never have a good VORP, even though the player is good. For example, suppose I play all 48 minutes and all 82 games on a team that loses every 100 possessions by 5 points. What's my VORP? Does it matter that if I weren't such a good player the team would lose every possession by 10 points? Is there any adjustment for that?


Most stats are done just for the regular season (and playoffs separately). It creates a common basis to compare players -- you aren't comparing how one player does in 82 games and another in 82 games + being swept by GS in round 1.

As for whether someone can have a positive VORP on a bad team, they definitely can. Worth remembering is BPM is not how the team does when the player is on the court. BPM is how a generic team should do on the court on average given the players box score statistics -- Rebound rates, steal rates, block rates, assist percentage, usage, turnover percentage, true shooting etc.

So, if a player on the Celtics has the same raw stats while his team goes up 20 as a player on the Nets whose team goes down 20, they would have the same BPM (there is a small team adjustment on one term but the general point stands). If the fundamental box score statistics are important, than they should capture whether the player makes the difference between losing every possession by more points than can be scored in a possession.

If the box score statistics are missing something important, than a stat based off on-offs is a good compliment to BPM(and VORP). But there is not any direct adjustment in BPM using actual game point differentials.

In terms of specific players to demonstrate how it works:
Joel Embiid 786 minutes 3.2 BPM 1.0 VORP
Richaun Holmes 1193 minutes 1.7 BPM 1.1 VORP

So, Holmes moves slightly ahead of Embiid in VORP based off playing more minutes, while Embiid had a much more positive BPM. Both players had very good BPM despite being on a less than stellar team.

And:
Marcus Smart -0.3 BPM (-1.1 OBPM 0.9 DPM) 1.0 VORP
Avery Bradley -0.5 BPM (-0.2 OBPM -0.4 DPM) 0.7 VORP

Here BPM looks at the players box score statistics and has Smart a solid defender, while Bradley is rated above replacement player but below average defensively.

If you think for instance that box score statistics are in general poor at defense, you might want to look at something like RPM and see how the actual on offs (buttressed by box score stats and other variables) rate the two players defenses:
13 Marcus Smart 0.26 ORPM 0.49 DRPM 0.75 RPM 5.98 Wins
71 Avery Bradley -0.44 ORPM -1.72 DRPM -2.16 RPM 1.13 Wins

So RPM when using on off data similarly finds Bradley poor defensively and Smart strong defensively.
Curmudgeon
RealGM
Posts: 23,937
And1: 4,776
Joined: Jan 20, 2004
Location: Boston, MA

Re: Please Explain VORP 

Post#3 » by Curmudgeon » Sun Jun 25, 2017 3:14 pm

How can you determine rebound rates and steal rates from a box score? I suppose you could take number of offensive rebounds divided by your own team's missed shots, and number of defensive rebounds over the number of shots the other team missed, and come up with a weighted average of some sort. But that ignores positional issues: some players will get fewer rebounds depending on where the are positioned in the offense his team is playing, and the defense his team is playing. The same for steals. A player's steal rate will be lower if the man he is guarding doesn't have the ball very often.

Also, as the article concedes, defense is simply not measured adequately by a box score. It isn't in baseball either.
"Numbers lie alot. Wins and losses don't lie." - Jerry West

"You are what your record says you are."- Bill Parcells

"Offense sells tickets. Defense wins games. Rebounding wins championships." Pat Summit
HartfordWhalers
Forum Mod - 76ers
Forum Mod - 76ers
Posts: 34,605
And1: 11,181
Joined: Apr 07, 2010
   

Re: Please Explain VORP 

Post#4 » by HartfordWhalers » Mon Jun 26, 2017 3:32 am

Curmudgeon wrote:How can you determine rebound rates and steal rates from a box score? I suppose you could take number of offensive rebounds divided by your own team's missed shots, and number of defensive rebounds over the number of shots the other team missed, and come up with a weighted average of some sort. But that ignores positional issues: some players will get fewer rebounds depending on where the are positioned in the offense his team is playing, and the defense his team is playing. The same for steals. A player's steal rate will be lower if the man he is guarding doesn't have the ball very often.

Also, as the article concedes, defense is simply not measured adequately by a box score. It isn't in baseball either.


I'm not really sure the confusion or question on rebounding rate and steal numbers.

If you are trying to argue that no stat is perfect, that is a point so basic that no one argues otherwise.

In terms of measuring defense, I would agree that BPM doesn't really have the tools to measure it directly extremely well. Steals and blocks can be a good stat, or they can be the sign of someone overreaching and taking bad risks that hurt their team. What BPM can do, is say that the players who match a certain profile tend to be very good defenders historically, and thus a player who matches that profile is likely to be a positive defender. If high steals are good 90% of the time, BPM will find steals good and be right 90% of the time, but also miss on some players.

Thats why in my examples I contrasted the results of BPM defensively with something like RPM, which works off the scoring margin that occurs with a player on the court. For an even more thorough assessment of a player, those two could be combined with player tracking defensive numbers for instance.

Blindly relying on any one measure isn't a way to get a great picture of anything. But dismissing information isn't a good idea either.

BPM finds Smart pretty awful at offense. BPM finds Bradley was bad at defense last year. Both those positions have a lot of other evidence that support them.
KqWIN
Head Coach
Posts: 6,214
And1: 1,505
Joined: May 15, 2014
 

Re: Please Explain VORP 

Post#5 » by KqWIN » Mon Aug 28, 2017 4:14 am

I don't think BPM uses steals and blocks to determine DBPM, at least not in the way we think. Two regressions are done to determine DBPM. One onto RAPM, and another onto ORAPM. We calculate the amount of BPM that is due to offense in the latter and the rest comes from defense. In other words, DBPM is BPM that is not accounted for in offense.
User avatar
bondom34
Retired Mod
Retired Mod
Posts: 44,464
And1: 26,491
Joined: Mar 01, 2013
       

Re: Please Explain VORP 

Post#6 » by bondom34 » Mon Aug 28, 2017 8:47 am

KqWIN wrote:I don't think BPM uses steals and blocks to determine DBPM, at least not in the way we think. Two regressions are done to determine DBPM. One onto RAPM, and another onto ORAPM. We calculate the amount of BPM that is due to offense in the latter and the rest comes from defense. In other words, DBPM is BPM that is not accounted for in offense.

Yep, from the link:

Defensive BPM is simply overall BPM minus offensive BPM. The offensive BPM regression was tuned to minimize weighted squared error on both offensive and defensive RAPM simultaneously.

Dion Waiters wrote: What’s the analytics on that?

E: Pacers for Reggie
W: Thunder for Russ
KqWIN
Head Coach
Posts: 6,214
And1: 1,505
Joined: May 15, 2014
 

Re: Please Explain VORP 

Post#7 » by KqWIN » Tue Aug 29, 2017 3:45 am

These discussions are dead, and I'm going off on a tangent here, but I'm curious as to what people think. There was some talk about how Avery Bradley is underrated on defense by the numbers. Fair, Bradley may provide better ball pressure than anyone in the league. However, I do think our eyes aren't trained to look a defense correctly. We can easily see Bradley get up in a guy's shorts because our eyes are attracted to the ball. But how valuable is that on it's own? More importantly, how does that compare to other aspects of defense?

We have a good idea of how good offense is generated. Teams will use some kind of motion, usually a form of PnR, to create an advantage that makes the defense collapse and hopefully leads to a shot around the rim or behind the three point line. Obviously it's more complex than that, but that's what most good offenses are trying to accomplish.

The ability to rotate, cut off driving lanes, and remove the advantage from on/off ball screens ect. aren't getting enough love, and quite frankly it's much harder to notice someone doing a good or bad job. Now I'm not saying that Bradley's intense ball pressure and man to man defense isn't important, but I feel as though it get's too much weight in comparison to those other things. The mono y mono defense is what we recognize as great defense, but I feel as though it might be the equivalent of hitting a 16 foot jumper in a guy's face. Extremely valuable on it's own right, but not the pillar of great efficiency.

HartfordWhalers wrote:.

bondom34 wrote:.
User avatar
bondom34
Retired Mod
Retired Mod
Posts: 44,464
And1: 26,491
Joined: Mar 01, 2013
       

Re: Please Explain VORP 

Post#8 » by bondom34 » Tue Aug 29, 2017 3:56 am

KqWIN wrote:These discussions are dead, and I'm going off on a tangent here, but I'm curious as to what people think. There was some talk about how Avery Bradley is underrated on defense by the numbers. Fair, Bradley may provide better ball pressure than anyone in the league. However, I do think our eyes aren't trained to look a defense correctly. We can easily see Bradley get up in a guy's shorts because our eyes are attracted to the ball. But how valuable is that on it's own? More importantly, how does that compare to other aspects of defense?

We have a good idea of how good offense is generated. Teams will use some kind of motion, usually a form of PnR, to create an advantage that makes the defense collapse and hopefully leads to a shot around the rim or behind the three point line. Obviously it's more complex than that, but that's what most good offenses are trying to accomplish.

The ability to rotate, cut off driving lanes, and remove the advantage from on/off ball screens ect. aren't getting enough love, and quite frankly it's much harder to notice someone doing a good or bad job. Now I'm not saying that Bradley's intense ball pressure and man to man defense isn't important, but I feel as though it get's too much weight in comparison to those other things. The mono y mono defense is what we recognize as great defense, but I feel as though it might be the equivalent of hitting a 16 foot jumper in a guy's face. Extremely valuable on it's own right, but not the pillar of great efficiency.

HartfordWhalers wrote:.

bondom34 wrote:.


Agree to a tee with what you're saying. People who claim and use only the eye test defensively mainly ball watch. When you ball watch, you obviously see on ball defense where Bradley excels. To add to that any current player doesn't care about off ball defense when they're asked who is a good defender, they're answering who stops them. Neither of these are measured in any APM since it's essentially a team based metric.

Now is Bradley underrated by RPM/RAPM? Possibly, there's likely some middle ground. But I'm also not sold he was ever an amazing defender for Boston, I tend to think that was Crowder (or a mix of Crowder/AB most likely). If this IT trade goes through I'm going to be interested in their defense this year. Hayward's not bad but not Crowder either.
Dion Waiters wrote: What’s the analytics on that?

E: Pacers for Reggie
W: Thunder for Russ
KqWIN
Head Coach
Posts: 6,214
And1: 1,505
Joined: May 15, 2014
 

Re: Please Explain VORP 

Post#9 » by KqWIN » Tue Aug 29, 2017 4:20 am

bondom34 wrote:
KqWIN wrote:These discussions are dead, and I'm going off on a tangent here, but I'm curious as to what people think. There was some talk about how Avery Bradley is underrated on defense by the numbers. Fair, Bradley may provide better ball pressure than anyone in the league. However, I do think our eyes aren't trained to look a defense correctly. We can easily see Bradley get up in a guy's shorts because our eyes are attracted to the ball. But how valuable is that on it's own? More importantly, how does that compare to other aspects of defense?

We have a good idea of how good offense is generated. Teams will use some kind of motion, usually a form of PnR, to create an advantage that makes the defense collapse and hopefully leads to a shot around the rim or behind the three point line. Obviously it's more complex than that, but that's what most good offenses are trying to accomplish.

The ability to rotate, cut off driving lanes, and remove the advantage from on/off ball screens ect. aren't getting enough love, and quite frankly it's much harder to notice someone doing a good or bad job. Now I'm not saying that Bradley's intense ball pressure and man to man defense isn't important, but I feel as though it get's too much weight in comparison to those other things. The mono y mono defense is what we recognize as great defense, but I feel as though it might be the equivalent of hitting a 16 foot jumper in a guy's face. Extremely valuable on it's own right, but not the pillar of great efficiency.

HartfordWhalers wrote:.

bondom34 wrote:.


Agree to a tee with what you're saying. People who claim and use only the eye test defensively mainly ball watch. When you ball watch, you obviously see on ball defense where Bradley excels. To add to that any current player doesn't care about off ball defense when they're asked who is a good defender, they're answering who stops them. Neither of these are measured in any APM since it's essentially a team based metric.

Now is Bradley underrated by RPM/RAPM? Possibly, there's likely some middle ground. But I'm also not sold he was ever an amazing defender for Boston, I tend to think that was Crowder (or a mix of Crowder/AB most likely). If this IT trade goes through I'm going to be interested in their defense this year. Hayward's not bad but not Crowder either.


It's safe to say that Bradley is underrated by those metrics, but at the same time it can't be a total accident that he's rated as low as he is. You're totally right about the way players themselves thing. They're going to think about the guy who is breathing down there neck, but they aren't going to think about the guy who makes the rotation a split second faster.
blabla
Sophomore
Posts: 116
And1: 54
Joined: May 23, 2012

Re: Please Explain VORP 

Post#10 » by blabla » Wed Sep 6, 2017 8:17 am

KqWIN wrote:I don't think BPM uses steals and blocks to determine DBPM, at least not in the way we think. Two regressions are done to determine DBPM. One onto RAPM, and another onto ORAPM. We calculate the amount of BPM that is due to offense in the latter and the rest comes from defense. In other words, DBPM is BPM that is not accounted for in offense.

This is irrelevant. Getting DBPM from regressing onto DRAPM, or just creating it as difference from BPM and OBPM (regressed against RAPM and ORAPM, respectively) leads to the exact same results.
KqWIN
Head Coach
Posts: 6,214
And1: 1,505
Joined: May 15, 2014
 

Re: Please Explain VORP 

Post#11 » by KqWIN » Thu Sep 7, 2017 2:37 am

blabla wrote:
KqWIN wrote:I don't think BPM uses steals and blocks to determine DBPM, at least not in the way we think. Two regressions are done to determine DBPM. One onto RAPM, and another onto ORAPM. We calculate the amount of BPM that is due to offense in the latter and the rest comes from defense. In other words, DBPM is BPM that is not accounted for in offense.

This is irrelevant. Getting DBPM from regressing onto DRAPM, or just creating it as difference from BPM and OBPM (regressed against RAPM and ORAPM, respectively) leads to the exact same results.


You're right, I guess I should read some of my old notes before I pretend to know what I'm talking about :D A few months out of school and I've already forgotten everything.
Dr Positivity
RealGM
Posts: 36,557
And1: 4,813
Joined: Apr 29, 2009
       

Re: Please Explain VORP 

Post#12 » by Dr Positivity » Wed Sep 13, 2017 6:36 am

I'm guessing Bradley excels at guarding PGs but is low key bad against SGs. He is basically a SG on offense but a PG on defense. I would be interested to see him playing PG beside Lebron or Harden
asubstituteforwar.wordpress.com

Return to Statistical Analysis