[Note: I'm just as anxious to get to the draft and next season as everyone else, but it's a long way off. Here's something different to think about in the meantime ...]
In many of the discussions we have here at Depressed Fan and other "hardcore" NBA blogs, people quote and then debate the meaning of different statistics. Defensive statistics, in particular, are often thought to be misleading, and with good reason. For example, if Lou's man blows by him and Brand goes over to help but Sammy doesn't rotate to Brand's man, does the resulting dunk by Brand's man count against Brand? Or on offense, if Jrue beats his man, draws a defender, and dishes to Jason Smith who fumbles it out of bounds, is the turnover solely Jrue's responsibility? Somehow, there has to be a better way of assigning credit and blame. I thought about this for a while and came up with a system that I'll describe after the break ...
The system is based on points per possession, with the initial observation being that teams in the NBA average around 1 point per possession (PPP). Actually, the average number among all NBA teams is closer to 1.05, but my system converts easily to PPP and its relatives, offensive/defensive efficiency (which is points per 100 possessions). So the idea then is to score each possession relative to the average 1 PPP. An empty possession is worth -1, a 1-point possession is worth 0, a 2-point possession is worth +1, and a 3-point possession is worth +2. But how are the points assigned?
On
offense:
- A made 2-point basket is +1 and might be divided into +0.5 for the passer and +0.5 for the shooter if the pass was responsible for the shot.
- A made 3-point basket is +2 and might be divided +1.5/+0.5 or +1/+1 for the passer/shooter if the pass resulted in an open shot.
- 2 made free throws is +1 but might be divided +0.5/+0.5 for the passer and shooter if the pass set up the foul shots. 1 of 2 free throws is a 0.
- A turnover is -1 and might be divided -0.5/-0.5 for the passer and catcher if the responsibility is partly the fault of the catcher.
- A missed field goal attempt resulting in an opponent defense rebound is -1 for the shooter. The one exception would be if someone else dribbled the clock down and gave the shooter only a desperation attempt (in which case it would be divided -0.5/-0.5).
- A missed field goal attempt resulting in an offensive rebound is -0.5 for the shooter and +0.5 for the offensive rebounder.
On
defense:
- A made basket is either -1 or -2 (for a 3-pointer), with blame assigned partly to the defender nearest the shooter and partly to the defender whose lack of defense resulted in the shot.
- 2 made free throws is -1 and is assigned to the defender who fouled the shooter, unless another defender's poor defense resulted in the foul (in which case blame is divided -0.5/-0.5).
- A turnover is +1 and is credited to the defender who forced the turnover. An exception is if the turnover is "unforced" (e.g., an opponent dribbles the ball off his foot) in which case the team is credited with a +1.
- A missed basket resulting in a defensive rebound is usually a +0.5 for the defender and +0.5 for the rebounder. If exceptional defense results in a poor shot and easy rebound, it might be +1 for the defender.
- A missed basket resulting in an offensive rebound is +0.5 for the defender and -0.5 for the person who gave up the offensive rebound.
Currently, I don't divide up partial credit into anything smaller than halves, but in theory partial credit can be assigned to whatever precision the scorer wants. At the end of a game, the individual numbers on offense and defense can be summed to get the total relative to 1 PPP. For example, if the Sixers win 90-82 and there were 80 possessions for each team, the sum on offense should be +10 and the sum on defense should be -2. The interesting part would be how the sums come about.
As an example, I went back to the 4th quarter of the Sixers' win over the Lakers on 3/17/09. Now, before the Iguodala Haters start complaining that it isn't a representative game, please note that this is one of the few Sixer games I still have on my DVR and
it's just an example. The Sixers came into the quarter trailing 73-62 and outscored the Lakers 32-20. Each team had 20 possessions, so the Sixers were +12 on offense and 0 on defense. Here's the play-by-play and how I "scored" each play according to the above system. Laker possessions are in
purple, Sixer possession are in
red.
- Lou caught in traffic, Farmer hits open 3 (Lou -2)
- AI9 penetration and score (AI9 +1)
- AI9 steals pass (AI9 +1)
- AI9 pass to Lou for lay-up (AI9 +0.5, Lou +0.5)
- KB penetrates, passes to Gasol for open 16-ft jump made (AI9 -0.5, Evans -0.5)
- AI9 missed 23-ft 3-pt, Sam off. reb. (AI9 -0.5, Sam +0.5)
- AI9 draws shooting foul, 1 of 2 FT (0)
- [Ivey and Marshall in for Sam and Miller]
- Marshall steals pass (Marshall +1)
- Lou pass to AI9 for dunk (Lou +0.5, AI9 +0.5)
- KB misses jumper, loose ball foul on Marshall rebound (AI9 +0.5, Marshall +0.5)
- Lou pass to Marshall for open 3-pt made (Lou +0.5, Marshall +1.5)
- Gasol penetrates, passes to JPowell for lay-up (Evans -1, overplay set up score)
- AI9 penetrates for layup plus foul, 1 of 1 FT made (AI9 +2)
- KB misses 3-pt, Gasol off. rebound (AI9 +0.5, Evans -0.5)
- Farmar missed 3-pt, Gasol off. rebound (Lou +0.5, Evans -0.5)
- Gasol missed hook, Evans def. rebound (Evans +1)
- Ivey scoop pass to Evans for lay-up (Ivey +0.5, Evans +0.5)
- JPowell missed 5-ft hook, JPowell off. reb. (Marshall +0.5-0.5=0)
- JPowell traveling (team +1)
- Ivey penetration score (Ivey +1)
- Marshall steals from KB (Marshall +1)
- AI9 transition lay-up (AI9 +1)
- KB missed jumper, Evans rebound (AI9 +0.5, Evans +0.5)
- Lou missed jumper, Evans off. rebound (Lou -0.5, Evans +0.5)
- Evans rebound basket (Evans +1)
- Odom missed wide-open 3-pt, Evans def. rebound (team +0.5, Evans +0.5)
- Lou passes to Marshall for 3-pt made (Marshall +1.5, Lou +0.5)
- Ariza 3-pt made off Farmar penetration (Ivey -1, Lou -1)
- Ivey 10-ft jump blocked, JPowell rebound (Ivey -1)
- Farmar missed 3-pt after good AI9 defense on KB, Laker team off. reb (AI9 +0.5, team -0.5)
- Ivey steals KB pass (Ivey +1)
- Ivey pass to Lou for lay-up (Ivey +0.5, Lou +0.5)
- KB missed 3-pt, Gasol off. rebound (AI9 +0.5, Evans -0.5)
- Gasol rebound basket (Marshall -1)
- Ivey missed 3-pt, KB rebound (Ivey -1)
- Gasol fouled by Marshall, 2 of 2 FT (Marshall -1)
- [Miller returns for Ivey]
- Miller missed 3-pt, KB rebound (Miller -1)
- Odom 7-ft hook made (Marshall -1)
- Lou draws 2 defenders, passes to Marshall for open 3-pt made (Lou +1, Marshall +1)
- Fisher missed 3-pt after KB bad pass, Odom off. rebound (team +0.5, Evans -0.5)
- Gasol fouled by Marshall, 2 of 2 FT (Marshall -1)
- Lou missed penetration, Ariza reb. (Lou -1)
- KB missed 3, Evans defending on switch, Lou rebound (Evans +0.5, Lou +0.5)
- Miller pass stolen by Gasol (Miller -1)
- Gasol missed lay-up, Evans rebound (Evans +1)
- Lou lay-up blocked, Gasol rebound (Lou -1)
- KB 22-ft jumper made (AI9 -1)
- AI9 3-pt made (AI9 +2)
Sixer totals for the quarter were as follows. On offense: Iguodala +6.5 (+7-0.5), Marshall +4 (+4), Evans +2 (+2), Lou +1 (+3.5-2.5), Sam +0.5 (+0.5), Ivey 0 (+2-2), Miller -2 (-2), Total=+12. On defense: Iguodala +2 (+3.5-1.5), Evans 0 (+3.5-3.5), Ivey 0 (+1-1), Marshall -1.5 (+3-4.5), Lou -2 (+1-3), Team +1.5 (+2-0.5), Total=0.
A couple of comments. First, this quarter was relatively easy to score defensively because the Lakers attempted zero transition shots. It's harder to assign credit or blame in transition. It was also relatively easy to score because the Lakers rarely made more than one pass after a switch. As we all know, the Sixers this past year switched a lot, and it was sometimes hard to keep track of who was supposed to be guarding whom. Second, there is some subjectivity to this scoring method, but I claim that all NBA defensive statistics involve some subjectivity. Finally, it should be noted that 0 is actually a good defensive team score and a poor offensive score. If Iguodala, guarding primarily Kobe, put up a 0 on defense, that would be an excellent result, because in most Laker games, Kobe contributes heavily to their positive team offensive score (i.e., their offensive efficiency comes in large part from him).
So, let me know what you think, especially if you have any suggested tweaks to the scoring system (I just came up with it, so there certainly could be improvements). If it's an interesting idea, I might try it on selected games this upcoming year.
Great job, and I really like it.
It's funny, you look at the quarter and it was one of those times where watching the game, it seems like Evans has such a huge impact (If I'm remembering correctly), but the numbers in the micro pretty much tell you exactly what numbers in the macro tell you. Even when he's on, he does just as much harm as good. Look at all the offensive boards his man got. That's the tradeoff of playing that crazy defense where he was doubling the ball 30 feet away from the hoop.
Compared to baseball, basketball is really like the wild west in terms of using statistics to understand the game. The box score stats tell you so little about what happened in the game. Part of me loves that, but a bigger part of me wants to look closer and figure out why certain things work, why certain players are so much better than their stats. What affect coaches and systems really have on the outcome of a game. Really great work here.
The most amazing thing, though, is that you still have that Lakers game on your DVR. If I don't watch a game in a day or two it's gone forever, replaced by Sesame Street.
The most amazing thing, though, is that you still have that Lakers game on your DVR. If I don't watch a game in a day or two it's gone forever, replaced by Sesame Street.
Ha ha, it helps that (a) I'm the only in the house that knows how to record with the DVR and (b) I only have one kid. It also helped to figure out that cartoons look almost as good in Standard-Def as High-Def (HD is what really takes up the storage space, about 5% per hour for me).
Anyhow, this Lakers game was recorded in Standard-Def (before I got an HD TV) and takes up almost no space. I also have the Bulls last Spectrum game and the Magic Game 1 playoff game still on my DVR.
Great stuff. You're right that it can get a little subjective when evaluating defense, but if common sense is used, it shouldn't have a huge impact (as long as the same person does the evaluation each game). For instance, if the offensive player dribbles by his man and is challenged by a defensive big but scores anyway, how would this be marked? It seems that the original defender would receive a -1, but help defense from interior defenders is very important in evaluating their worth. Perhaps a sub-score could be added for help-defense effectiveness?
Anyway, this is a minor detail. Again, great work.
In particular, I'd like to see how Speights performs in this category. Watching the games, it seems that he is a very poor help defender, but I'd like to see some stats to back that up.
This is a good point about help defense. The question for me would be: did the help defender do his job? If he didn't, there's a good case that the initial defender and help defender would split the -1. Another example would be the Elton Brand play where a defender doesn't show on a pick-and-roll and the dribbler goes in for a basket. Blame could be divided between the initial defender and the help defender, depending on how good the pick was and how good the help was.
09-10 76ers games would have been interesting to score this way, but I deleted most of them right after I watched them!
I think the main problem with basketball analysis is that unlike baseball, you can't just plug in a number. There needs to be some type of evaluation and common sense. That's why I like what you are trying to do here. Statheads sometimes aren't willing to look at the play and see what happened because they want a number. In basketball, that's not possible. It's too much of a team game, which is dependent on circumstance. A couple of things that I saw:
1. It's hard to determine how much a pass in worth. Personally I feel a dump off for a dunk or an alley-oop should be waited more heavily than passing it to an open three point shooter. The guy still has to make a shot, and it does get hazy. I think Rajon Rondo is great, but when Ray Allen starts hitting absurd contested jumpers off screens that Rondo passes to him, Rondo's numbers are inflated. Giving him +0.5 for some of those shots seems like too much. I'd rather give the shooter closer to full credit.
2. Defensive credit to the guy nearest the shot might be unfair as well, especially for Eddie Jordan coached teams. Take a play where Willie would get beat off the dribble and the defense would scramble (Thad would undoubtedly be a part of this) and cover up 2 more passes until someone was left wide open from three. Iguodala would contest it very late by coming from out of his rotation and they'd hit a three. I'm not sure this would be Iguodala's place to get a -1.
3. I think this system would be awesome for stuff like offensive boards. Sam does have games where he gets a bunch of offensive boards and misses shots right after, which really inflates his value. Let's say he gets an offensive board and goes back up with a bad shot and misses. He does get an 0-1, but the offensive boards may take precedence if he piles them up. In this system, while he's getting a +0.5 for the rebound, he gets a -1 for the miss. That takes away what I feel are "hollow" offensive boards.
Alright, that's all I have for now. Great post.
Your method's strength and weakness is it's subjectivity. Rich's post convinces me to suggest changing one thing: all misses should be -1 and offensive rebounds +1. I acknowledge that jacking up a shot with no one underneath is less likely to result in an offensive rebound, but in general there's no reason the shooter should get half the credit for an offensive board. There's no magical way to shoot to make it clank off the rim and into a teammate's hands.
That's definitely a fair point Bob. There could be some good misses like if Jameer Nelson drives the lane and draws a couple of guys and his shot is cleaned up easily by Dwight Howard. That's a pretty good miss. In most cases as you say, why would more credit be given to the shooter if someone cleans up his mess. I can only guess Statman's thinking, but if you make it -1 and +1 on a miss and rebound, then a guy like Sam won't really be held accountable for what he does after a rebound. This way, even if he keeps the possession alive but then misses, it's still a -0.5 net for him instead of 0 if we made all missed shots one. That's a matter of opinion on how valuable the offensive rebound is. Should it be +1 because you are creating a possession that wasn't there without the rebound or 0.5 because you are simply keeping it alive?
Could an offensive rebound be +0.5 with the miss still being -1 without the math and system being compromised?
I'm actually leaning the other way, an o-board should be +1.
Essentially, what we're talking about here is relative team production. So with the baseline of 1 point per possession, Dalembert is salvaging that point by extending the possession with the offensive rebound. He's penalized if he misses the put-back or turns it over from that point.
This is what I mean...
- Possession is worth one point.
- When the shot is missed, that possession is worth zero points.
- When Dalembert gets the offensive rebound, it's worth one point again (a gain of one point).
- When Dalembert misses the shot it's back to zero, or the exact same as if he hadn't grabbed the offensive rebound.
So it's a zero sum in that case, +1 for the board, -1 for the shot. It's not logical to me for Dalembert to get a -.5 for grabbing a board and then missing a shot/turning the ball over.
I could see splitting the credit on a tap out offensive board, that would make sense. But when we're talking about value, if you think about it, an offensive rebound and put back is really worth just as much as a made three, relatively speaking. The three is 2 more points than expected (1 is expected on a typical possession). The o-board put-back is worth 2 points (0 is expected on a missed shot).
This is exactly the point I was making but expressed more fully and clearly.
Okay, so here was my thinking on the -0.5 for misses that result in an offensive rebound. I began by assigning -1 for a miss and +1 for an offensive rebound. As Brian noted, there is something elegant about -1 for losing a possession and +1 for re-establishing possession. But here was the problem: on defense, I assign a +1 for a defensive stop, and I divide it into +0.5 for the defender on the ball and +0.5 for the rebounder (or -0.5 for someone who gives up an offensive rebound to the other team). So if on offense I give -1 for a missed shot and +1 for an offensive rebound, I would be weighting misses and rebounds differently on offense and defense.
As it turns out, there is a good way to think about the system as is: every shot has two components, whether it goes in or not and who gets the rebound. I assign equal weighting to those two components (0.5 each). For a missed shot/defensive rebound, both components get assigned to the shooter (-1); for a missed shot/offensive rebound, the rebounder negates the missed shot (-0.5/+0.5). The only problem, as you note, is that missed shots are weighted differently depending on who gets the rebound. But is this really a problem? It's true, as you note, that the shooter has no idea who is getting the rebound when he misses. But it's not the intention of the shooter I'm scoring, it's the result of the play. And I would argue that a miss that results in an offensive rebound is not as "harmful" as a miss that results in a defensive rebound. In fact, possession isn't truly lost if someone gets an offensive rebound (I'm pretty sure in offensive/defensive efficiency ratings, a shot+offensive rebound+follow shot counts as one possession, not two).
So, guys (Rich, Bob, Brian), think it over and let me know what you think about the above reasoning. I think giving +1 for an offensive rebound and -1 for all misses could work, but it would weight offensive and defensive rebounds (and offensive/defensive misses) differently. Thanks.
I regret that i don't have the time right now to fully think this through. but i will suggest a case for why defensive rebounds might count less than offensive rebounds: the expectation is that a miss will be rebounded by the defending team. (what's the ratio? 65/35?) so, an offensive rebound results in a whole new possession for free. A defensive rebound is only expected. The numbers get screwy because it isn't reasonable or fair to expect 100% defensive rebounds.
I made a case but haven't convinced myself. Still, I don't see the point of penalizing someone for a misses put back.
Regarding defensive rebounding: would it make sense to penalize a defender for allowing an offensive rebound? the net result is something like 65% of a turnover.
Hmmn, this is really interesting.
So here's the question I'm left with. Are offensive and defensive rebounds equal in terms of ppp? I'm not sure it really matters, in the grand scheme of things. The important part is that defensive rebounds are worth the same on both ends of the floor, same with offensive rebounds. Both methodologies accomplish that.
I think I still lean toward +1 for o-boards and -1 for all missed shots, simply because that keeps the shooting portion of the equation in line. Since it's a zero-sum in terms of PPP, I don't think it's fair for a guy to wind up -0.5 if he grabs and o-board and misses the putback. Essentially, his missed shot + o-board is worth the same as the guy who missed the first shot and did nothing else to further the possession, if that makes sense.
The more I think about it, I like the -1 and +1 values being distributed. I know that you are scoring the result and in that case -0.5 makes sense, but it can misrepresent a shooter's game. Let's say Kapono goes 0-5 and has none of his shots go for offensive boards. His total is -5. Let's say four of them get rebounded between Sammy and Brand. Then his total is -3. I still think he's just as useless, but the team did a better job cleaning up his mess. I think they should be rewarded.
The one problem I see with this system is that it may place too high of a premium of guys who grab offensive rebounds. Reggie Evans may have scored a little too well if you go at it this way.
Evans would probably wind up being a zero anyway:
Lou misses jumper (-1), Reggie Evans o-board (+1), Reggie Evans misses layup (-1), Reggie Evans o-board (+1), Reggie Evans misses layup (-1).
Defensive rebound Gasol, Reggie Evans fouls Gasol (-1) :)
Hey guys, thanks for the feedback. Looks like it's unanimous that offensive rebounds should be scored +1 and all misses -1. In my example quarter, the changes would have been minimal: Iguodala down from +6.5 to +6, Lou from +1 o +0.5, Sam up from +0.5 to +1, and Evans up from +2 to +2.5 (only two Sixer offensive rebounds).
There's still a problem in my mind with this revision, because offense and defense lose symmetry. For example, say Kobe plays good defense on Iguodala and forces a missed jumper, then Sam rebounds it over Gasol and puts it back in. Defensively, Kobe gets +0.5 and Gasol -1.5 (-0.5 for the rebound allowed, -1 for the basket allowed). Under the original scoring, Iguodala would have gotten -0.5 and Sam +1.5 (symmetrically the opposite of Kobe and Gasol). Under the revision, Iguodala would get -1 and Sam +2. That doesn't seem right on the face of it. On the other hand, a miss/defensive rebound is never symmetrical, so maybe it doesn't matter.
I guess there's no perfect solution. I'll try it out both ways when the season starts and see which one seems to be working better.
I see what you're saying about the defensive end, I think. Does the math still work? If you give Kobe a +0.5 for causing the miss, then does Gasol get a -0.5 for not getting the defensive board?
Right, that's how it was originally designed. And that in turn was based on a +1 for a defensive stop being divided into +0.5 for the on-ball defender and +0.5 for the defensive rebounder. So a guy giving up an o-board would get a -0.5, instead. In turn, that scoring dictated how I originally thought of scoring a miss/o-board on the offensive end. But there's something to be said for the idea that all misses are scored the same; I can see both points of view ...
(The team math still works for both methods, to answer your question. As long as the net for a possession is correct, the team math works out. There are more points assigned on offense, both positive and negative, in the revision.)
What if a guy like Chauncey makes a 3 pointer 5 feet away from the 3-point line? Wouldn't it be irresponsible to contest that shot?
Rich, thanks for the thoughts. Here are some responses ...
Personally I feel a dump off for a dunk or an alley-oop should be waited more heavily than passing it to an open three point shooter.
That's true, and 82games' Passing Rating takes into account the type of basket in weighting their assists. In the case of my system, I do take into account (somewhat subjectively) how much the pass helps the shot. On one play early in the quarter, Miller passed cross-court to Iguodala, who dribbled in from 25 feet pass two defenders to lay it in. Miller was credited with an assist officially, but I gave full credit (+1) to Iguodala. On a couple of Marshall's 3's, Lou made a decent pass but I gave Marshall +1.5 and Lou +0.5. On the third one, Lou did a good job of drawing two defenders, and his pass set up a wide-open shot for Marshall, so I split it, +1/+1. Really, there should be greater precision involved, but that would also allow for more subjectivity.
Defensive credit to the guy nearest the shot might be unfair as well
Yes, and I did try to take this into account. There was one play when Evans overplayed Gasol, Gasol drove, Marshall helped, and Gasol passed to Josh Powell (Marshall's man) for a lay-up. I assigned Evans the full -1 for that play because Marshall did his job and Evans was the one who "caused" the basket. On the switch play you described, I would think about who deserved the most blame for the play and try to assign "blame" accordingly.
In this system, while he's getting a +0.5 for the rebound, he gets a -1 for the miss. That takes away what I feel are "hollow" offensive boards.
Well, actually a miss that results in an offensive rebound is scored at -0.5. There's a reason for that, although the reason may not hold water after examination. I'll give the reasoning in my response to Bob's post (which will be later because I've got to go).
Sorry, a bit O.T., but I wrote part 2 in my open letter to the Sixers.
Pt 2:
http://forums.realgm.com/boards/viewtopic.php?f=26&t=1016469
The First post basically says the team needs to go back to trying to win with a captivating Franchise player and stop with all of the quick fix band-aid moves.
art 2 is basically a simple 3 year plan:
Draft Turner and try and win ASAP, but without selling low on the current young players or committing to new long term contracts for vets. Then decide if in 3 years to either try and contend (using Brand's expiring + a young player to add the final star piece) or be in great position to make wholesale changes without being under the burden of bad longterm contracts.
Here was pt 1 from before the lottery:
http://forums.realgm.com/boards/viewtopic.php?f=26&t=1007143
Hello Webmaster,
I am webmaster of Houston Tickets Broker and Astros Tickets websites; I've found your website information (http://www.depressedfan.com/) and advice to be a very good fit for our visitors so could you please give us the best price for a small summery to add your esteemed website site wide for a period of Monthly,half Yearly and Year? We will make payments Via PayPal so if interested, please mention your PayPal id.
If we are happy with your price, then we will send you the Link details that you can place on your website and we will make the payments to the PayPal id provided by you.
Regards,
Peter Freeman
[email protected]