Tango on Baseball Archives

Aaron's Baseball Blog - Basketball (February 9, 2004)

Aaron goes beyond field goal %, and does field goal points per field goal attempt. So far so good. But then he adds:

The above numbers ignore what each player does at the free throw line, which is an entirely different issue.

However, taking a guess, I'd say at least 90% of the foul shots were from the 2-pt area. To ignore the foul shooting is to ignore that you had almost guaranteed points lost.

You have 2 ways to handle this:
1 - If you do NOT want to look at a player's foul shooting ability, you can do:
pts = field goal points + 0.75 * free throws
opps = field goal attempts + fouls (note, if you get 2 free throws for 1 foul, that's 1 opp)
pts/opps

The 0.75 if the average number of points for a free throw, more or less.

2 - Including foul shooting:
pts = field goal points + free throw points
opps = (same as previous example)

Note, I would remove any technical fouls and points.

You might be tempted to go with #1, but you might be more likely to foul a player if you know he's a bad thrower (and vice-versa).

Either way works, but I would definitely include fouls in the mix.
--posted by TangoTiger at 11:37 AM EDT

Posted 12:14 p.m., February 9, 2004 (#1) - J Cross (homepage)
Points (including free throws) per Shot (homepage).

Posted 12:22 p.m., February 9, 2004 (#2) - Jim P
You'd have to make two more adjustments:
1. Some free throws come from non-shooting fouls.
2. Some free throws come after successful shots (the old-fashioned 3 point play).

Posted 12:31 p.m., February 9, 2004 (#3) - tangotiger
1. Yes, anything like technicals, or I suppose those late non-shooting fouls, should be removed.

2. Yes, that would count as "1" opp.

***
J. Cross: If I read that measure properly, there are no opps included for free throws in there, and the numerator will also include technicals.

***

Can I assume that a missed field goal on a foul does not count as a field goal attempt if missed, but it does if they get the basket?

Posted 12:53 p.m., February 9, 2004 (#4) - FJM
Jim: I understand why you want to exclude non-shooting fouls, just like technical fouls. But why would you want to exclude fouls completing a 3-point play?

Posted 1:30 p.m., February 9, 2004 (#5) - Rob H
Can I assume that a missed field goal on a foul does not count as a field goal attempt if missed, but it does if they get the basket?

Yes. I do not understand the use of 0.75 in #1 above.

Looking at some other basketball stats this afternoon, I am convinced RPI stands for "We'll Screw Gonzaga" in some unknown foreign language. College basketball has a way's to go.

Posted 1:40 p.m., February 9, 2004 (#6) - tangotiger
If you don't want to measure how well Shaq does on free throws, but you want to account for the fact that he gets alot of free throws, then you need to add in a term:
.75*free throws
which will give you the number of points an avg NBA player would have gotten, if he got the number of free throws that Shaq earned by being in a position to take a shot, but was denied by the opponent's foul.

Posted 1:42 p.m., February 9, 2004 (#7) - Jim P (homepage)
Jim: I understand why you want to exclude non-shooting fouls, just like technical fouls. But why would you want to exclude fouls completing a 3-point play?

We shouldn't exclude the "and one", but include it as part of the same opportunity.

And with fouls on 3 point shots, the shooter gets 3 free throws, so those 3 foul shots also count as 1 opp.

nba.com didn't have detailed stats, but espn has adjusted field goal pct and a bastardized points per shot (see homepage).

* PPS: Points Per Shot = PTS/FGA
^ AFG%: Adjusted FG Percentage = [(PTS - FTM)/FGA]/2
ADJ FG% measures shooting efficiency by taking into account the total points a player produces through his field goal attempts. The intention of this adjustment is largely to evaluate the impact of three-point shooting. For ex: If Shaquille O'Neal has 3-5 FG, all two-point shots for 6 points, then his ADJ FG% = [(6/5)]/2 = .600. Meanwhile, if Ray Allen is 2-5 FG, but his 2 FGM are both three-pointers for 6 points, then his ADJ FG% = [(6/5)]/2 = .600
---
Thinking about it some more, I'd say that we could probably ignore all these adjustments for most players and just say that opp = FGA + FTA/2.

Posted 1:57 p.m., February 9, 2004 (#8) - Craig B
Incidentally, I've been doing basketball work over the last couple of weeks. Aaron is going to run a piece I did on the top MVP candidates sometime soon (maybe tomorrow) in which I introduce a new Wins Above Replacement metric.

I also have some new work on defense, including two new defensive metrics, up on Batter's Box. Craig Names His NBA All-Defensive Team For Some Reason.

Anyway, Points per Shot Attempt.

According to John Hollinger, the proper denominator to use is FGA + (0.44*FTA). Since there are 0.44 possessions lost per free throw.

That doesn't seem quite right to me... 0.44 possessions, but not 0.44 shots. However, I use it, but for scratchpad calculations you'd be just as good, if not better, using 0.5*FTA.

I sent the following e-mail to Aaron...

Aaron, nice piece on adjusted FG%.

Now I have another stat that will blow your freakin' MIND. Points per shot attempt, or PSA, which is another stat from the prodigious basketball brain of John Hollinger.

Adjusted FG% is great, but what it doesn't tell you is that the Shaqs and Zach Randolphs are shooting in traffic all the time - so they are constantly being fouled, getting to the line, and getting more points.

Now, a stand-around jumpshooter like Walter McCarty, who takes two-thirds of his shots from three-point land, looks great on adjusted FG% because he hits 40% from behind the arc. But Walter never gets to the line and when he does, he shoots 65% or so.

Andre Miller, on the other hand, shoots about one three-pointer a game and is lousy at them, so on adjusted FG% he doesn't look all that hot. (He also shoots under 50% from the field) But in figuring Miller's performance, you have to take into account that he makes his living posting up other point guards; he gets about six free throws a game, and hits 85% of them. Three times a game, Miller takes a shot that isn't counted as a "field goal attempt", much less a make, but which nets him around 1.7 points.

Miller gets more points per shot attempt than Waltuh, but his adjusted FG% is lower. What we need to demonstrate this, is PSA. You calculate PSA as follows:

Total Points
--------------------------------------------------
Field Goal Attempts + (0.44 * Free Throw Attempts)

The reason for the 0.44 is that historically, each free throw attempt is equal to 0.44 possessions. You could use 0.5 if you wanted to, and the numbers would come out practically identical.

To use our two examples, McCarty had an adjusted FG% of .535, Miller of .470. But Miller has a PSA of 1.13, McCarty 1.10.

Posted 2:13 p.m., February 9, 2004 (#9) - Rally Monkey
Take PSA and divide by 2, and you get a better measure of effective FG%.

I'll guess Kevin Garnett will rank at the top of MVP candidates this year.

Posted 2:58 p.m., February 9, 2004 (#10) - J Cross
hmmm... Maybe we also want a shooting value above replacement metric:

(PSA - .9)*opps

where opps = FGA+.44*FTA

Posted 3:11 p.m., February 9, 2004 (#11) - David Smyth
Points per FGA? Don't you also have to take into account the missed shots.

I mean, a 50% 2ptFG rate and a 33.3% 3ptFG rate will both produce the same pts/att, but the missed shots have a marginal impact, pretty much like the out in baseball. The 3 pt guy is missing more shots (IOW, losing more possessions). If an avg possession is worth 1 pt, then the 3 pt "equivalent" to a 50% 2pt rate is 37.5%, not 33.3%.

Posted 3:28 p.m., February 9, 2004 (#12) - J Cross
I'm not following you David. You use a possesion on each attempt whether you hit or miss. 2 for 6 on 3's is just as good as 3 for 6 on 2's...

or maybe it's a little better. Let's say that on each missed shot your team has a 30% chance of getting an offensive rebound. If you go 2 for 6 on 3's you've missed 4 shots and gets 1.2 rebounds. So, you've scored 6 points and used 4.8 possesions. 1.25 points per possesion. If you shot 3 for 6 on 3's you missed three shots and got .9 rebounds. 6 points on 5.1 possesions or 1.17 points per possesion. Ofcourse we'd have to weigh in the chances of a team getting rebounds from different types of shots.

Posted 4:46 p.m., February 9, 2004 (#13) - Patriot
What David mentions strikes me as somewhat similar to what I wanted to do with QB rating. Everyone(this was on FH a while ago) was looking at a yards/attempt, and I wanted to factor in completions somehow. I felt that 5/10 for 50 yards was better then 1/10 for 50 yards. But I was the only one who felt that way, and now I can see why. My personal sympathies in both cases are with higher rates of success and equal results. Whether my personal sympathies have anything to do with reality is another matter.

Posted 4:55 p.m., February 9, 2004 (#14) - David Smyth
If player A takes 100 2pt shots and hits 50, he has scored 100 points. If player B takes 100 3pt shots and makes a third of them, he has also scored 100 pts. Each player has scored 1 pt per att.

But B has missed 67 shots, while A has only missed 50 shots. A team scores about 1 pt per possession. (This is analogous to the baseball expectancy grid of a team scoring .5 runs per inning.) So those extra 17 possessions lost are worth about -1 pt each, relative to average.

This is similar to a pair of baseball players who both create 100 runs in 600 PAs. If one has made 300 outs and the other 400 outs, their values are quite different.

Posted 5:02 p.m., February 9, 2004 (#15) - David Smyth
Patriot, I don't see the QB situation as being analogous.

Posted 5:05 p.m., February 9, 2004 (#16) - Patriot
Yes, but as J Cross pointed out, the other team gets the ball either way. And if you miss more shots, you theoretically should get more offensive rebounds.

And above, someone said that if a free throw is made, it counts as a field goal? That would be news to me.

Posted 5:09 p.m., February 9, 2004 (#17) - tangotiger
Player A takes 100 2pt shots, hits on 50, and the other 50 are rebounded. Of those 50, the opposition grabs 40 of them, and the off gets another 10 possessions.

So, that's 90 possessions turned over, with 100 points earned.

Player B take 100 3pt shots, hits on 33, and the other 67 are rebounded. Of those 67, the opposition grabs 63 of them, and the off gets another 4 possessions.

So, that's 96 possessions turned over, with 99 points scored.

To equate them to say both with 1 point per possession, you need:
player A: gets 90 points per 100 shots, or 45% from 2pt
player B: gets 96 points per 100 shots, or 32% from 3pt

(All numbers just educated guesses.)

Posted 5:10 p.m., February 9, 2004 (#18) - Patriot
Well, I said it was "somewhat similar", not completely analogous :)

I probably won't convince you otherwise, but my reasoning was that both of our players have produced the same thing(50 yards, 100 points), in the same number of attempts(10, 100). The only difference between the two statistically, in both cases, is their production per successful attempt. So the issue is, is production the only thing that counts(probably IMO), or is success rate the only thing that counts(no way), or is it some combination of the two.

In the QB case, I was saying "combination". In this case, you are saying "combination". I see them as similar situations. That doesn't mean that they have to have the same answer.

Posted 5:13 p.m., February 9, 2004 (#19) - Patriot
Edit: the difference is not only their production per successful attempt, but their success rate. The only thing that IS THE SAME is their overall production. Completely reversed what I should have said.

Posted 5:17 p.m., February 9, 2004 (#20) - tangotiger
To expand my example more to include fouls:

Player A:
100 2pt "opps"
41 hits
10 fouls (of which he makes 80%, and which the def gets back all of them)
10 missed and off rebound
39 missed and def rebound

So, that's a total of 82+8 = 90 points earned for a price of 90 possessions turnover. His shooting percentage is 41/(41+10+39)=45.6%

Player B:
100 3pt shots
31 hits
4 fouls (of which he makes 80%, and which the def gets back all of them)
3.8 missed and off rebound
61.2 missed and def rebound

So, that's a total of 93+3.2 = 96.2 points earned for a price of 96.2 possessions turnover. His shooting percentage is 32.3%.

So, a player with a 46% 2pt FG is equivalent to a 32% 3pt FG player.

(Again, just plug in the appropriate numbers.)

Posted 5:29 p.m., February 9, 2004 (#21) - DavidSmyth
If 2 QBs have the same yds/att (say 8.0), but different completion % (say 50% vs 60%), then for the 60% guy to be more valuable, you have to make the case, I think, that the individual play in football is an opportunity unit. It might be true that the 60% guy will make more 1st downs, and keep drives going, but it also may be the case that the 50% guy will throw for more TDs. That is, is 10 passes of 10 yds each really the same as 1 pass of 100 yds (guaranteeing a TD)? And if so, or if not, how much credit goes to the QB, and how much to the receiver and his yds after catch? I have no idea if, after you have yds/att, the completion% has an impact. There simply hasn't been enough research by the stat-savvy crowd.

Posted 5:37 p.m., February 9, 2004 (#22) - David Smyth
Tango's point seems to be, relative to the analysis I posted, that you should take into account how easy it is to get an offensive rebound on a missed 2pt att vs a missed 3pt att (plus perhaps how easy it is to convert the off. rebound into a basket). I agree.....

Posted 5:41 p.m., February 9, 2004 (#23) - Jim P
Player A takes 100 2pt shots, hits on 50, and the other 50 are rebounded. Of those 50, the opposition grabs 40 of them, and the off gets another 10 possessions.

Player B take 100 3pt shots, hits on 33, and the other 67 are rebounded. Of those 67, the opposition grabs 63 of them, and the off gets another 4 possessions.

Did you mean to estimate that the defense grabs 80% of the rebounds off missed 2 pointers, but 93% of rebounds off 3 pointers? I would have guessed that 3 pointers would have led to more offensive rebounds from the ball bouncing hard off the rim, maybe something like:
99% of free throws are rebounded by the defense
80% of 2 pointers
70% of 3 pointers.

This would lead to an equivalent shooter B with:

100 3pt shots
25 hits
4 fouls (of which he makes 80%, and which the def gets back all of them)
21.3 missed and off rebound
49.7 missed and def rebound

So, that's a total of 75+3.2 = 78.2 points earned for a price of 78.7 possessions turnover. His shooting percentage is 26.0%.

Posted 5:58 p.m., February 9, 2004 (#24) - Patriot
David...I agree that C% is probably mostly irrelevant.

And to get back to basketball, I know we can't do any better, because the stats are what the stats are, but there is, at least in my observation, a big difference between rebound rates on different types of 2 pt shots. A center who is standing under the basket and puts it up has a much better chance of getting the board then someone taking a baseline jumper. Or at least it seems to me.

Posted 6:34 p.m., February 9, 2004 (#25) - Rob H
hmmm... Maybe we also want a shooting value above replacement metric:

Very much so, since it is easier for a player to shoot 50% on 10 FGA than on 20 FGA. And "replacement level" would be a team-influenced stat, since the worse your teammates are, the less discriminating you should be.

Posted 7:51 p.m., February 9, 2004 (#26) - Rally Monkey
Patriot, I'm pretty sure the 5/10 QB is more valuable. If he's getting 10 yards every other attempt, he's getting first downs and holding onto the ball. The other QB has one big play, but in the meantime he's forcing his team to punt a few times and giving the opponent chances to score.

Posted 9:45 p.m., February 9, 2004 (#27) - dr feelgood
Forget which qb is more valuable, which qb is better. I would think that the 5/10 qb would be much much more likely to repeat his 50 yard performance than the 1/10 guy. Of course, I might be wrong, but if I had to bet, I'd bet on the small play guy repeating his performance.

Posted 9:57 p.m., February 9, 2004 (#28) - Patriot
Rally Monkey and dr feelgood, as I sort of indicated, I tend to agree with you. But I can't give a good "proof" of why...it just seems right to me. And I agree that the 5/10 guy is more likely to repeat his performance, but that doesn't necessarily make it more valuable.

Posted 10:43 p.m., February 9, 2004 (#29) - Rally Monkey
The reason the 5/10 QB is more valuable is because pass attempts are not the context here. They are more analogous to at bats in baseball than outs. The 5/10 QB will be able to continue his drive, and make more pass (or run) plays than the other guy will, compared to his opponents scoring attempts.

Posted 10:49 p.m., February 9, 2004 (#30) - Rally Monkey (homepage)
I posted NBA win shares about 2 years ago on this site. I apologize for dropping the ball on the project, I've lost a lot of my interest in pro basketball in recent years. Anyway, today I finally felt motivated to get the page back up. The above link has an explanation of the system, I've made one change, subtracting 0.35 for each foul committed.

Tango, you have my blessing if you want to move this page into your library for better safekeeping.

Posted 10:49 p.m., February 9, 2004 (#31) - Patriot
But a QB is not the entire offense. The running game will have a lot to do with whether the drive continues or not. And the other QB already ended his drive with a TD. And not all completions are good. A 3rd and 9 completion for 6 yards...a -1 yard screen...etc. I think the 5/10 QB is better, as I said before, but I don't think it's a slam dunk.

Posted 10:52 p.m., February 9, 2004 (#32) - Rally Monkey (homepage)
And here are win shares for the current season, through games of last saturday.

Posted 9:14 a.m., February 10, 2004 (#33) - kevin
There's a ton of problems with rating basketball players based on stats, especially shooting stats.

First, unlike baseball, there is an elective element to shooting stats. The coach can call out a play for his "best" scorer. It penalizes the rest of the guys, who get to set picks and get no credit for the play.

Second, again unlike baseball, defense accounts for half of the players value. When a defender deflects a pass sufficiently to disrupt the flow of a play, he gets no credit but he should. Defensive statistics in basketball might be worse than the ones for baseball. If you cut your man off trying to go baseline and he steps on the line, the player gets no credit for that because those kinds of stops are not recorded.

I could go on and on with this but you get my drift. The way basketball games are scored needs a major overhaul before the statistics become meaningful enough to start accurately rating players.

Posted 10:05 a.m., February 10, 2004 (#34) - Craig B
Second, again unlike baseball, defense accounts for half of the players value.

Is this true? What evidence do we have for this? My own instinct is that the existence of the shot clock makes offense more important. There is, in my view, some anecdotal evidence, particularly based on salary structure, that defense is not as important as offense. At the very least, offensive ability appears to be much more variable than defensive ability - meaning that in an analysis of the player pool, offensive ability has more "value" than defensive ability.

Furthermore, lots of teams have been able to have good defenses despite putting together teams full of garbage... last year's Nuggets were an excellent example.

The way basketball games are scored needs a major overhaul before the statistics become meaningful enough to start accurately rating players.

Well, what do you mean by "accurate"? It's certainly true that ratings aren't going to capture 100% of the facts about players. What we can do, though, is try to accurately represent what data we have - and it is rather a lot.

Posted 10:21 a.m., February 10, 2004 (#35) - Rally Monkey
I agree with you Craig. From my observation, getting a guy like Larry Brown on the sideline has as much impact on defense as a huge shotblocking center.

Its in no way perfect, but I think my win shares does a decent job rewarding defenders. Only 5 players have more WS this year than Ben Wallace, and he doesn't score at all.

Posted 10:43 a.m., February 10, 2004 (#36) - kevin
Craig B,

Bill Russell didn't score all that much. He was a pretty bad free throw shooter. He never took a three-pointer in his life. But he was the most dominant player of his generation, maybe of every generation.

If you go back and start adding up the stats (I know, they didn't keep track of blocs and steals and turnovers then), Russell comes out below Wilt and Robertson and Pettit and Baylor. But he was better than those guys because he could prevent the other team from scoring and he allowed his own team to get easy baskets. What he was doing to effect those outcomes wasn't kept track of. A lot of them are still not kept track of.

How can offense be more important than defense? Name one team that played mediocre defense that ever won the title?

George Gervin has the reputation of being a great player because he scored a lot of points and his shooting percentages were pretty good. I went to a Celtics game against the Spurs and the Celtics. I was sitting right behind the Spurs bench and the Celtics on the floor were laughing at Gervin. They took turns posting on him. M.L. Carr was scoring at will on him. The other 4 guys could have been playing terrific defense and it wouldn't have mattered one bit. Because of Gervin, they had a massive hole in the defense. If they tried to switch, then the Celtics would just go to the guy Gervin switched off on. If they tried to double, then the Celtics would just find the open man.

Gervin was grotesquely overrated because you could not structure a consistent defense as long as he was on the floor. The Spurs had other good defenders like Johnny Moore, Mike Mitchell and Artis Gilmore but they always lost to the Lakers because they could not stop them from scoring. The only time they could win is if the other team was cold and played a bad game. That isn't going to get you very far.

Posted 10:53 a.m., February 10, 2004 (#37) - Rally Monkey
Kevin,

Check out career win shares. I've got Russell ahead of Gervin, Pettit, and Baylor. Slightly behind Robertson, well behind Chamberlain. Russell would rank better if we had block and steal data for him, maybe around 16 per 82 games.

I can understand if you want to argue he was better than Wilt. (The never ending debate for basketball historians), but even with the data limitations, Russell looks pretty good by the numbers. I've done some preliminary work on playoff win shares, and Russell looks even better there, while most players drop off in playoff games.

Posted 10:59 a.m., February 10, 2004 (#38) - kevin
There is, in my view, some anecdotal evidence, particularly based on salary structure, that defense is not as important as offense. At the very least, offensive ability appears to be much more variable than defensive ability - meaning that in an analysis of the player pool, offensive ability has more "value" than defensive ability.

The are two problems with this statement. Salary should not be used as a synonym for value because players are not strictly paid for how good they are. Even if you eliminate the errors in judgement, there is still the marketability angle that goes into paying players. There's a lot of irrationality that goes into how players are paid. Draft order, for instance.

Second, even if there is more variability in offensive ability than defensive ability (and I question this assumption), all that means is the the standard deviation of defense is tighter than offense, it doesn't change the balance of value, which is logically 50/50. Every time one team is on offense, the other is on defense and vice versa.

Posted 11:01 a.m., February 10, 2004 (#39) - Craig B
Russell comes out below Wilt and Robertson and Pettit and Baylor. But he was better than those guys because he could prevent the other team from scoring and he allowed his own team to get easy baskets.

Well, whether Russell was better than Wilt will get you a hell of an argument. I can't rightly say, as this was before my time, but it's hardly a foregone conclusion.

How can offense be more important than defense? Name one team that played mediocre defense that ever won the title?

I'm sorry, that's a stupid argument. Name one team that had a medicore offense that ever won the title. See? It doesn't tell you anything. Why the hell would a championship team ever be mediocre at anything?

You're also committing the player vs. team fallacy. Just because half of a team's success comes from defense, does not mean that half of the value of the players comes from defense.

Gervin was grotesquely overrated because you could not structure a consistent defense as long as he was on the floor.

Look, I'm not disagreeing with you. The stuff you are pointing to is all important. (It does, by the way, show up in team statistics).

What you're doing, though, is closing your eyes and screaming "YOU CAN'T RATE PLAYERS ON STATISTICS!!!!". That's not helpful. We *do* have a very good statistical record, that we can use to tell us about what goes on on the floor. Or we can choose to ignore it. Feel free to ignore it; but don't tell me that the car I'm building isn't any good because it doesn't fly. Deal with it on its own merits.

Posted 11:16 a.m., February 10, 2004 (#40) - kevin
Rally, you have Pettit ahead of Russell per 82. You also have Olajawon and Robertson and Jabbar ahead of him. If your numbers are telling you that Bob Pettit and Kareem are better than Russell, then you have a problem with your metric.

No way is Jabbar better than Russell. He did two things better than Russell. he could shoot the hook better and he was better free throw shooter. Every thing else, Russell did better, some things a lot better.

For instance, centers get no credit for outlet passes that lead to baskets or free throws, usually because the ball hits the floor or another pass is made before the shot is taken. But the outlet pass is incredibly important because it leads to easy baskets. Russell did that as well as any center. I am willing to wager that, if that information were available, Russell would lap the field. Kareem couldn't throw an outlet pass if his life depended on it. The Lakers had a fsst break because Magic rebounded the ball himself and ran it down the floor. It would have been a lot easier for Magic if Kareem could have thrown him an outlet every now and then.

So what's the result? The Lakers post and Kareem gets credit for a basket that should have been scored by a guard or forward, if only Kareem could find the wherewithall to throw it to him. Because of this, and the fact that Kareem was such an effective post scorer, Kareems' teams were locked into a low-post offense. They had no other choice. It's a credit to Magic that his superior talents could override Kareems' limitations.

Posted 11:21 a.m., February 10, 2004 (#41) - kevin
What you're doing, though, is closing your eyes and screaming "YOU CAN'T RATE PLAYERS ON STATISTICS!!!!". That's not helpful.

I'm not saying this at all. I'm saying the current statistics stink so bad they give erroneous valuations. I think it's very helpful to point out that alterntative statistics need to be recorded.

Posted 11:27 a.m., February 10, 2004 (#42) - Stephen
Kevin,

Not to say that your opinions are incorrect, but your whining sounds very much like unreasonable Jeter fans. How much of your worship of Russell stems from the fact the he played for your favorite team and won a bunch of team championships?

The only things Frank Thomas did better than Ichiro! are hit for power and take walks. Ichiro does everything else better, some things a lot better. No way is Thomas better than Ichiro.

Again, I'm not saying your subjective opinions are necessarily wrong, but you seem unable to concede that recalling memories of your favorite player is probably not the most objective way to rank him.

Posted 12:04 p.m., February 10, 2004 (#43) - kevin
Stephen, your statement would make more sense if the Celtics won before Russell got there and won after he left. They were a good team but not a championship contender before he got there and they they tanked after he left. That is very telling as to his value. It's not whining. I just think you have a problem if you are using metrics to list the greatest and Russells doesn't come out in the top 5.

I'm going to say it. I think Russell was the best player of all time. He won in college, he won in the pros. He almost never lost. His college team had the record for consecutive wins before UCLA broke it. There is nothing subjective about counting championship trophies. And he was by far the most dominant player on every team he played for.

Let me redo RallyMonkeys list to WS/82:

o'neal,shaquille--675 165 20.0 C
Chamberlain,Wilt-1045 254 19.9 C
jordan,michael----990 234 19.4 G
duncan,tim--------370--83 18.3 F
robinson,david----923 196 17.4 C
johnson,magic-----906 190 17.2 G
bird,larry--------897 183 16.7 F
malone,karl------1353 265 16.0 F
Abdul-Jabbar,K---1560 300 15.8 C
barkley,charles--1073 203 15.5 F
robertson,oscar--1040 180 14.2 G
pettit,bob--------792 135 14.0 F
olajuwon,hakeem--1238 210 13.9 C
russell,bill------960 160 13.7 C

You see the problem you have? Russell comes in at #14. Do you really think Charles Barkley is a better player than Bill Russell? Charles Barkley, of the Erving/Malone/Cheeks/Toney/Barkley Sixers, who couldn't win 55 games? If you can't see something wrong with this list, then I don't know what to say.

OK, you think I'm biased for Russell. Let's choose another player. How about David Robinson. Are you trying to tell me that you honestly think David Robinson is one of the 5 best players of all time? Olajawon used to kick his ass left and right but the list indicates Robinson is better than Olajawon by a large margin.

And where is Bob Cousy? Cousy routinely led the league in assists by a whopping margin and was by far the best rebounding guard of his era, Cousy or Andy Phillip. Cousy's rebounding totals were as good as Magics. But your list has Kevin Johnson over Cousy. Kevin Johnson was a crappy rebounder and a mediocre defender. But he took a lot of shots so your metric likes him.

Are you beginning to see the problem?

Posted 12:05 p.m., February 10, 2004 (#44) - kevin
Stephen, your statement would make more sense if the Celtics won before Russell got there and won after he left. They were a good team but not a championship contender before he got there and they they tanked after he left. That is very telling as to his value. It's not whining. I just think you have a problem if you are using metrics to list the greatest and Russells doesn't come out in the top 5.

Let me redo RallyMonkeys list to WS/82:

Are you beginning to see the problem?

Posted 12:14 p.m., February 10, 2004 (#45) - kevin
Sorry for the double.

I do not wish to disparage the work being done. I just wish we would acknowledge the giant gaping holes in the available data and not take these lists too seriously. If anything, they highlight problems in what statistics are recorded.

Posted 1:21 p.m., February 10, 2004 (#46) - Rally Monkey
I never claimed to take these lists too seriously. I recognize the data limitations, I've already posted that Russell should be closer to 16 wins per 82 if he had his blocks and steal recorded, like the other centers (excluding Wilt) on this list.

Are you trying to tell me that you honestly think David Robinson is one of the 5 best players of all time?

I never tried to, you're the one who resorted the list.

To address the question of which one was better I'd need to do more work here. First of all, include Robinson's final season. The career list is only updated to 2002. I'd need to acknowledge that Olajuwon played longer past his prime, bringing his per game average down. I'd look at some measure of peak performance, maybe best 3 consecutive years. And of course playoff performance. Off the top of my head, Robinson lost more in the playoffs than most other star players. I'm not sure which of the two was actually better.

Posted 2:00 p.m., February 10, 2004 (#47) - kevin
One other thing. I'm old enough to have seen Russell play. It was near the end of his career but he was still a great player.

In a regular season game against the Knicks in Russells last year, Willis Reed outplayed Russell and he looked like he was sort of drifting through the game. Russell was notorious for this. In unimportant games, he would go on cruise control. Dennis Johnson did the same kind of thing. But when the playoffs came around, he would really turn it on. He kicked Reeds ass in the playoffs that year. That's why his playoffs stats are better than his regular season ones. I don't know what to make of this other than the fact that it is hard to argue with the results.

Another interesting case is Wilt. Wilt was a horrible FT shooter. One year, in 1968, he shot .380 from the stripe. This obviously invited the hack-a-shack defense, especially since, in those days, they would shoot a one-and-one on non-shooting fouls when they were in the penalty. That is, before the 5 team foul penalty, a non-shooting foul would earn 1 FT. After 5 team fouls, the shooter would shoot the second FT only if the first one was made. If the first was missed, the miss was in play. So, basically, you could not let Wilt touch the ball with less than 5 minutes to go in a close game, or even if the game wasn't close because he would miss all his freebies. Russell wasn't very good either but he was a lot better than Wilt and he didn't need to score to be effective so he wouldn't touch the ball much in crunch time. That's one big reason why Russells' team always seemed to beat Wilts' team in the playoffs.

Plus, I thought the never fouling out thing was pretty weird on Wilts part. If Wilt never fouled out once after over 1000 games, then he wasn't challenging the shooters enough.

RallyMonkey, hwo do you know how much Russells' WS should be adjusted if you don't have the blocks and steals data? And even if you did, you would also need the number of times he made someone miss just by intimidation and know the difference in intimidation misses that Russell earned against the number of the other players in the league.

Let's face it. the statistics are horrible for basketball. So horrible that giant mistakes can be made in valuing past performance.

Posted 7:05 p.m., February 10, 2004 (#48) - Cooper
I like Hollinger's stuff...it's well thought out and well conceived. What i like most is his use of points per possession. That equals out the differences between the generations. Is Russell snapping up 30 rebounds that impressive when you factor in that the typical game had so many more possessions and they shot so much poorer (FG%). Btw, wasn't Russell's FG% terrible for center? Dint he shoot in the high .300's? Hollinger does a nice job of weighting all these different factors.

If you get the chance check out his Alley Oop site or read Basketball Prospectus.

Kevin , I seem to remember Kareem throwing that over hand outlet pass to Magic an awful lot of the time.

Posted 7:15 p.m., February 10, 2004 (#49) - Rally Monkey
RallyMonkey, hwo do you know how much Russells' WS should be adjusted if you don't have the blocks and steals data?

Dave Heeren in the 90-91 Basketball Abstract estimates Russell would have added 3 points to his tendex, Chamberlain 2. So I took his estimates and plugged it into the win share formula. I was going by memory before. I rechecked it and get 2-3 WS for Russell, 1-2 for Wilt.

Sure, there are plenty of limitations in basketball stats. Its annoying that, more than a decade after the first 'sabrmetric' basketball books, I still don't know who draws the most offensive fouls. But there's enough here that you can understand the players better by using stats than ignoring them. Especially if you understand where the greatest weaknesses are (defensive specialists).

Posted 10:30 p.m., February 10, 2004 (#50) - kevin
Rally,

I don't know how Heeren can just stab at a number like that and plug it in. If you don't have the data, you shouldn't just make something up that seems to make sense to you. You will almost certainly be wrong. Nobody has a clue as to how many blocks Russell would get per night. I have heard some wild numbers thrown about. Russell himself said he blocked Walt Bellamys' shot 15(!) times in one game. Can you imagine someone doing that today?

I got one of Heeren's books and thought Tendex was a terrible metric. He doesn't really weight anything. His player evaluations are a joke. He says Dennis Johnson is an average player. Johnson was a superb player, one of the half dozen best guards in the league. Johnson gets underrated because his game was real subtle and weighted to the defensive end.

Heeren doesn't stop to consider that the defensive statistics he is using are incredibly superficial. All there are are blocks and steals and rebounds. Those things only measure a fraction of defense. It's kind of like trying to evaluate pitchers by how many triples they give up. Deflections aren't counted. Forced turnovers aren't counted, at least as a individual stat. Forced passes aren't counted. Opponent FG% isn't counted, on an individual defender basis. There is no metric to measure how good a guy is at lending help defense (Jordan would really shine there, as would Russell). When a shot blocker intimidates someone into a three-second call, he isn't given credit for that. If you had some of this stuff, you might start to make some headway. No distinction is made between good and bad fouls.

Cooper,

The problem with Kareem is that he had trouble finding the outlet man after he got a rebound. He wasn't all that strong and could be stripped unless he took pains to secure the ball. By the time he secured it, the defense was back in position. Russell and Walton were the two best at the outlet. The ball was on it's way to the outlet man almost before their feet hit the ground. That's what made the Celtics fastbreak so effective. They had 2 on 1s and 3 on 2s on a huge percentage of possessions.

Wilt's rebound numbers are a little better than Russell's but not much, until you calculate for rebounds/minute rather than game. Russell got more rebounds/minute in both the regular season and the the playoffs than Wilt did. The only competition either of them had for the rebound title was each other. No one else was even close until the late sixties.

Russells FG% was actually quite good. His first few years in the league, he was a top 5 guy. It gradually drifted down as he got older but it remained solid until his couple of years, when it wasn't very good. But his game was never really about scoring anyway.

I want to apologize for being too critical. I would actually like to stimulate discussion on this: if you could start all over, what statistics would you like to have? Once you decide what is missing, then you can adjust your valuations with consideration of the data gaps.

Posted 11:15 p.m., February 10, 2004 (#51) - Hi, my name is Kevin
... and I'm a Celtics fan.

Posted 1:27 a.m., February 11, 2004 (#52) - Rob H
Kevin , I seem to remember Kareem throwing that over hand outlet pass to Magic an awful lot of the time.

Me too. I also remember Kareem was a much better player during his ten years in the league before Magic arrived than in the ten years after.

Posted 7:17 a.m., February 11, 2004 (#53) - kevin
Really, Rob H? His numbers were much better before Magic but they didn't seem to translate to wins, especially playoff wins. And isn't that what being good is all about? The best players are the ones who contribute most to helping their team win?

Kareem wasn't a leader. In fact, he was disliked. His inability to get a job in the NBA speaks volumes about what his peers thought of him (did you see the HBO piece on him?). for better or worse, the best player on the team is the de facto leader. Kareem was so aloof and detached, he neglected his leadership responsibilities. If Magic hadn't come along, he would have gone down as the biggest enigma in the history of the NBA.

Plus, he didn't do the little things. He never hit the floor for a loose ball. He wasn't a good enough shooter to be able to draw the opposing shotblocker away for the basket. You were locked into a low-post offense with Kareem. That thing in Airplane was funny but it was also true. The kid was just summarizing the opinion of a lot of people about Kareem at that time.

Posted 7:22 a.m., February 11, 2004 (#54) - ntr Hi, my name is Kevin
... and I'm an idiot.

Posted 7:41 a.m., February 11, 2004 (#55) - Cooper
Kareem went to the playoffs and won pretty dang often. Kareem was smart and he took care of his body, thus never missing games. There's alot of value in being in the lineup every game.

Kevin, you are wearing green colored glasses and you have brought in to everything Tommy Heinson has ever said. I thought you'd have more to offer beyond "every celtic good -everybody else bad". You might get more of the response you are looking for on a Celtic's message board. Good Luck.

Posted 9:06 a.m., February 11, 2004 (#56) - Rally Monkey
Heeren's number is an estimate, who knows if its right but its better than assuming Russell didn't block any shots at all.

For Dennis Johnson, I saw him play the last few years of his career, and at that time I think he was no better than average. He didn't have the speed to keep up with quick PG's (especially Isiah) and if the Celtics played a team I was rooting for, I was happy any time Dennis took a 20 footer. It sure beat watching Bird, McHale, and Parish score their automatic points.

Still, Johnson was a great player early in his career. Win shares (I've updated the career page to 2002-03 and any technical changes) puts him at 6.6 per 82, over a 1000 game career. 5 is an average player who plays every minute of every game, by definition. That puts his career total in the top 100 all time. I don't think anyone but an extreme Celtic fan would put him in the top 50.

Posted 9:10 a.m., February 11, 2004 (#57) - kevin
Cooper, I'm not saying Kareem wasn't a great player. I'm using him as an example of how he might be overrated because the stats that are recorded are the ones that reward him the most and might underrate his superstar peers when he is compared to them. There's a bit of a Joe Carter element to Kareem. Same with Gervin, Theus etc. And his career pre-Magic was a big disappointment. He had been ouplayed by Walton in the '77 playoffs, he played poorly when the Warriors eliminated the Bucks in the '73 playoffs. Sure he went to the playoffs a lot. All the great players went to the playoffs a lot. But his teams always seemed to be getting upset for one reason or another.

You may or may not like the Celtics but you can't argue with their success. You may resent that but you have to respect it too. Auerbach constructed 3 different championship teams (4 if you consider that the Celtics teams at the end of Russells' career were completely different than the squad he entered the league with). I would think as an analyst that you would be curious as to how he was going about it, what types of players he was looking for, that caused that much success. You can do this with all the great teams, the Lakers in the early '50's, the Jordan Bulls, the late '80's Pistons etc.

I'm using the Celtics players to criticize the valuations because they're the ones I'm most familiar with. I could just as easily do it with players from other teams as well (Robinson-Olajawon for example).

Posted 9:23 a.m., February 11, 2004 (#58) - kevin
For Dennis Johnson, I saw him play the last few years of his career, and at that time I think he was no better than average. He didn't have the speed to keep up with quick PG's (especially Isiah) and if the Celtics played a team I was rooting for, I was happy any time Dennis took a 20 footer. It sure beat watching Bird, McHale, and Parish score their automatic points.

I agree with this totally, Rally. The Celtics got one really good year and two pretty good years out of him. After that he was average or a bit above average.

But prior to Jordan, he was the best defensive guard in the league, Johnson or Moncrief. And he gets a big bump because he could play three positions. I never saw him do it with Seattle or Phoenix but the Celtics would sometimes use him at small forward to cover a 6:5 shooter and he would take the guy right out of the game. His play in the playoffs was damned impressive with Seattle in the late '70's. He was the key to that team (he deservedly got the playoff MVP in '79).

You don't have to be a Celtics fan to be a Johnson fan since most of his best years were with other teams. So for Heeren to say Johnson was average is just absurd (IIRC, it was a career valuation, not a yearly one). Again, a lot of Johnsons' game was about defense and not making mistakes (not making dumb fouls, for example). Most of the things associated with that element of the game are not recorded.

Posted 9:38 a.m., February 11, 2004 (#59) - Cooper
Kevin -with all the talent the Celtics had on the team from 80-89 -they never should have lost a game.

Walton better than Kareem. The Big 3 have to be top 20 players in your estimation (i'm assuming). Dennis Johnson "take guys right out of a game"-what's next "Greg Kite was better than it appeared.... he really knew how to foul a guy and those kind of things aren't listed in the stats".

The Celtics were a great team, but if they were as good as you claim then they greatly under performed and fell below expectations. Couldn't it be possible that your perception is clouded a bit?

Posted 9:59 a.m., February 11, 2004 (#60) - kevin
Cooper,

When Walton was healthy, he was better than Kareem. Maybe you didn't see that series in '77 but I did. The Lakers had a better regular season record than the Blazers in '77 but got swept. One of the principal reason was that Walton outplayed Kareem. His numbers don't look so good now because of his multiple injuries and his minutes were never very high but, when he was healthy and playing, nobody was ever better. By the time he got to the Celtics, he had lost a lot but could still play. And he only really played the one year for the Celtics. His best moments were with other teams, except for the trophy in '86.

The Celtics were a great team, but if they were as good as you claim then they greatly under performed and fell below expectations. Couldn't it be possible that your perception is clouded a bit?

Where the hell did this come from? The Celtics in the '80's were as good as their record. They were a consistent 60 win team until age and injuries degraded their effectiveness.

No, the Big Three are not top 20. Parish was very good but he was never an elite center. Rally's numbers have him around guys like Gilmore and Sikma and sounds about right. Bird is certainly top 20, McHale probably top 50.

Posted 10:08 a.m., February 11, 2004 (#61) - kevin
Let's talk about fouling. There's a great deal of strategy in fouling. You can stop a team from scoring by fouling. You can work the clock by fouling.

Alternatively, you can give away easy points by fouling too. The good players make smart fouls. The dumb players allow themselves to be manipulated intoa situation where they have to give up a bad foul. Fouling is a major strategic option in basketball but distinctions between good and bad fouls are not made in the record. What does a coach tell his team last as they leave a timeout in a close game? "And whatever you do, don't foul!" Bad defenders get forced into making fouls that hurt the team.

Posted 1:15 p.m., February 11, 2004 (#62) - kevin
The inadequacy of the available statistics can also be reflected in the similarity scores on basketball reference. Has anyone looked at those? The third most similar to Russell is Dick Barnett. I can say with absolute certainty that the only similarity that Russell had with Dick Barnett as a basketball player is that they both shot lefthanded.

Posted 1:36 p.m., February 11, 2004 (#63) - Craig B
It's certainly a problem with the similarity scores method.

I don't think it's a problem with basketball statistics.

(Anyway, Dick Barnett's similarity with Russell came out at 845, that's not remotely similar. Dick Barnett has dozens of players more similar to him than Russell.)

Posted 1:53 p.m., February 11, 2004 (#64) - Rally Monkey
What that means is there is no similar player to Bill Russell. I thought Pervis Ellison was, but not since 1986 at Louisville.

Craig, nice article on Gleeman's site. Its very similar to win shares, even though you are subtracting a replacement level and I subtract a 0 win level (as James did for his WS). In Basketball there's not much difference, 0.733 to 0.600, which seems intuitively correct since a replacement level team will win only 9-12 games in a season.

Posted 2:23 p.m., February 11, 2004 (#65) - kevin
It's a problem with the statistics too. There is no way that Dick Barnett should be anywhere near Russells list, low score or no. No matter how badly you muck the formula up, there is no way you should come up with shooting guards to match Russell. I think David Robinson is a pretty good match for Russell but he is not on the list.

Russell list

Similar Players (Career)
Nate Thurmond* (860)
Jerry Lucas* (845)
Dick Barnett (845)
Dave Debusschere* (840)
Tom Vanarsdale (830)
Johnny Kerr (827)
Bill Bridges (822)
Richie Guerin (816)
Bob Boozer (814)
Dick Vanarsdale (814)

Barnett, the VanArsdales and Guerin shouldn't be anywhere near this list. They are on it because they scored a similar amount of points and assists. That's the problem with measuring scoring. It's a very context dependent statistic, like RBIs. There have been lousy players who have been good scorers and good players who have been lousy scorers.

Barnett has Jack Twyman as his number 2 man at 926. that's a fairly high score. Twyman was a scorer/rebounder forward. He wasn't remotely like Barnett.

I don't know how the scores are done but there needs to be some serious weighting of the numbers according to position, for starters.

Posted 2:34 p.m., February 12, 2004 (#66) - Rex
Among the better no-nonsense threads seen in quite awhile. Please keep it up.

Posted 5:37 p.m., February 12, 2004 (#67) - kevin
I'll take you up on your request, Rex.

Let's just chuck scoring average right out the window in terms of value for starters. Scoring average is a about as valuable for rating a player as ABs is for a baseball player. If you have a guy with a high scoring average but a low shooting percentage, that guy shold get a double whammy penalty because not only is he being selfish and not getting the ball to his more talented teammates, he is shooting his team out of wins. Lloyd Free comes to mind here. The difference between a good shooting percentage and a bad one is about .080, or 16%. But scoring averages, if we just count guys with similar min/game numbers can be up to 200-300% (the difference between Kobe and Ben Wallace say). So there's very little relationship between total points and effective shooting.

Second, versatility is important. A player that can play more than one position is very valuable because the coach can mix and match matchups and substitute more effectively. Joe Dumars was better than Byron Scott because Dumars could play the point while Isiah rested while Scott couldn't give Magic a blow. Versatility is defensive too. A guy that can cover players from different positions is extremely valuable, not only because they give the coach more strategic options but the defense doesn't get hurt when players are forced to switch off. That's why Auerbach loved big guards. Havlicek, Chaney, Johnson, Siegfried and Sam Jones could all do the job at more than one position. So if the defensive situation calls for a switch, so what? The big guards can handle it if they have to switch off on a small forward.

Points/possession is important but limited. For instance, if a team draws a lot of fouls, that is good but it doesn't really show up in the pts/possession column, at least not directly. Drawing fouls is one of the most important statistics that is not recorded. You can infer it by how many FT a player shoots but the numbers will probably be off by a bit because only a fraction of drawn fouls result in FTs. You can do a lot by drawing fouls. You get the opposing team in the penalty, eliminating strategic options, you can get their best players out of the game (a huge and underrated and unrecorded statistic) and you get to shoot easy shots to put points up on the board and freeze the clock. One might argue that the best teams are the ones that most effectively foul (or not foul)and draw fouls. Some players are much better at drawing fouls than others. Jordan, Bird and Magic all drew tons of fouls. Factor that in and those guys move up the list.

So the subject of fouling is neglected, important and very accessible to analysis, since the information is avaialble from game logs you can lift off of ESPN. It's just a matter of deciding how to weight them. Certainly, it would be easy to start just by counting drawn fouls.

Posted 7:21 p.m., February 12, 2004 (#68) - Rally Monkey
My guess is some teams have access to superior stats. I think there was an article a while back about the Dallas Mavericks paying for stat service, probably to help them evaluate defense.

For the public, we don't know a thing we didn't know since the mid 70's. Turnovers forced would be a huge improvement on steals. Fouls drawn is a good suggestion. Shots altered, defensive breakdowns (how often do you force your man to take an off balance shot, vs drive for a layup), block shots recovered vs blocked shots out of bounds. All of this would help quite a bit.

Not all high scoring/low shooting % players are selfish, however. If that was the case the Sixers wouldn't miss Iverson when he's hurt. There is value in a player who can get his shot off without help, especially when the rest of the team is made up of extremely limited offensive players. There are high scorers who actually suck, but its better to identify them on a case by case basis, instead of focusing on low shooting percentages.

Posted 9:21 p.m., February 12, 2004 (#69) - kevin
Rally,

Iverson takes a ton of bad shots. The degree of difficulty quotient on his shots are probably the highest in the league. It really hurts the Sixers when he does that. But he makes up for it by drawing fouls and getting to the line a lot. He shoots a ton of freebies. That helps his teammates by giving them a blow so they can conserve their energy for defense. That's why the Sixers suffer when he is out.

He also gets a lot of points off of steals. And his passing has gotten better. His assist totals are very good for a 2.