Monday, February 13, 2012

Playoff Performance

Introduction


One of the most common refrains about an NBA star is how the player either performed when it mattered or folded in the playoffs. Usually championships are used to measure guys against one another, but basketball is a team sport and your teammates ultimately will decide whether you win a title or not. However, individual performance can be tracked, but it must be adjusted for the fact that on average a player's percentages and statistics on a per minute basis go down. Only the best teams in the league are allowed, and the further one goes into the playoffs the harder the competition gets. Offensive numbers plummet as squads tighten up their defense, while having more days off only intensifies the series where the same players match up for as many as seven games in a row. Even with the gauntlet that is the playoffs, do any players truly get better? And whose performance is the worst?

Methodology

I devised this idea when I was trying to compare Tim Duncan's playoffs to Kevin Garnett's. In his prime, Garnett had terrible teammates while Duncan enjoyed the likes of David Robinson, Tony Parker, Manu Ginobili, and a nice cast of role players employed by Popovich. Using just per game averages is troublesome because it can mask poor shooting percentages and minutes played will skew the results. There are certainly a lot of problems with the advanced metric PER, but one of its disadvantages is actually helpful here -- it rewards players for shooting more often, meaning a guy scared of the playoffs limelight can't just stand in the corner and only take shots when he's open for a better PER number. You can also compare players with little in common because every box score statistic is used.

When you look at the two aforementioned big men's playoff stats, however, you'll notice that Duncan got to play heavy minutes in his best years while Garnett only went to the finals when he was in his thirties after he lost most of his explosiveness. Only using the difference between career regular season and playoff PER would be problematic in that respect. A better method is to compare the difference in regular season and playoff performance on a season by season basis. A number is calculated for an average of career regular season PER weighted each season by playoff minutes. Then the difference between the weighted regular season and playoffs PER will show how performance dipped or increased.

For example, imagine a player, Moses Monroe, with two seasons in the league where his rookie year he had a 12 PER and his sophomore year 19. He went to the finals in his rookie year as a bench player, but the next year their star player suffered a devastating injury and the team missed the playoffs entirely. His playoffs PER is around 12 while his career PER around 16. Simply comparing those two numbers one would conclude he's terrible in the playoffs, but using the method outlined above one would see his performance remained steady. His playoffs PER was low because his only appearance was as a rookie when he struggled throughout the whole year.

Players included in this study needed a minimum of 4000 minutes in the playoffs. This was chosen to trim the list of candidates, but also with more appearances random chance or circumstance have less of an effect on the results. I regretfully couldn't include guys from the pre-80's basketball eras because basic statistics like turnovers weren't even tracked until the 1977-78 season, and consequently a true PER can't be calculated. I also included a few players I thought were interesting who were under the minutes requirement, although most of the important players in the last 30 years had 4000 or more.

Results

I broke up the results in two tables. One is for the stars, shown directly below, who are defined as having a regular season PER of 18 and above. While this was somewhat arbitrary, I needed some systematic way of separating Michael Jordan from, say, Tayshaun Prince. It's much more of an accomplishment to raise your statistics when your output is already amazing. For people who don't know PER or need a refresher, 15 is the league average, 20 is an all-star candidate, 25 is a weak MVP candidate, and 30 is a historic season. Averaging over 25 for a career is amazing, and from the table below only Olajuwon, Duncan, Jordan, O'Neal, and Lebron have done it for the playoffs. I also included basic stats on a per 40 minutes basis (per 40 so the comparison is more apples to apples.)

In terms of PER differences, there are a few surprises. One is that Michael Jordan doesn't dominate the list, but given that his performance hardly even dipped from his lofty level I'd say it's actually a compliment. A difference of -0.11 is essentially zero, and he has the highest playoffs PER and points per 40 minutes at an astounding 32. The only stars who improved were Isiah Thomas, Hakeem Olajuwon, Reggie Miller, Tim Duncan, and James Worthy. Those men are generally highly regarded for delivering in the playoffs, and remember Michael Jordan is right behind them in sixth place.

Star Players and Playoffs

Name
Rank
PER difference
Regular season PER
Playoff PER
Points/
40 mins
Rebounds/40 mins
Assists/40 mins
Isiah Thomas
1
1.46
18.3
19.8
21.5
5.0
9.4
Hakeem Olajuwon
2
1.22
24.5
25.7
26.1
11.3
3.2
Reggie Miller
3
1.09
18.5
19.5
22.4
3.1
2.7
Tim Duncan
4
0.27
25.2
25.4
23.0
12.5
3.5
James Worthy
5
0.13
18.2
18.3
22.8
5.6
3.5
Michael Jordan
6
-0.11
28.7
28.6
32.0
6.2
5.5
Charles Barkley
7
-0.14
24.3
24.2
23.4
13.1
4.0
Ray Allen
8
-0.30
18.1
17.8
19.7
4.1
3.4
Dirk Nowitzki
9
-0.32
25.1
24.8
25.1
10.1
2.5
Dwyane Wade*
10
-0.69
25.3
24.6
25.5
5.7
5.5
Manu Ginobili*
11
-0.75
21.3
20.5
21.0
5.9
4.8
Scottie Pippen
12
-0.88
19.2
18.3
18.0
7.8
5.2
Magic Johnson
13
-0.91
23.9
23.0
19.6
7.8
12.4
Chauncey Billups
14
-0.92
20.3
19.4
19.1
3.7
6.3
Jason Kidd
15
-1.02
18.6
17.6
13.9
7.0
8.5
Kobe Bryant
16
-1.07
23.4
22.3
25.9
5.2
4.8
Kevin McHale
17
-1.25
20.6
19.4
22.3
8.8
1.9
Shaquille O'Neal
18
-1.30
27.4
26.1
25.9
12.4
2.9
Paul Pierce
19
-1.39
19.9
18.5
21.5
6.5
4.1
Jeff Hornacek
20
-1.81
18.3
16.5
17.6
4.4
4.4
Lebron James
21
-1.82
28.1
26.3
25.7
7.7
6.4
Tony Parker
22
-1.83
18.4
16.6
20.7
3.5
5.4
Clyde Drexler
23
-1.83
21.5
19.7
21.3
7.2
6.4
Vince Carter*
24
-1.98
21.7
19.7
23.1
6.2
4.5
Patrick Ewing
25
-2.00
21.6
19.6
21.6
11.0
2.1
Steve Nash
26
-2.03
21.9
19.9
19.4
3.9
10.0
Chris Webber*
27
-2.03
20.9
18.9
20.6
9.6
4.0
John Stockton
28
-2.16
22.0
19.8
15.2
3.8
11.5
Kevin Garnett
29
-2.26
23.9
21.7
20.2
11.5
4.0
David Robinson
30
-2.36
25.4
23.0
21.0
12.3
2.7
Gary Payton
31
-2.74
18.1
15.4
15.7
4.2
5.9
Larry Bird
32
-2.79
24.2
21.4
22.6
9.8
6.2
Karl Malone
33
-3.18
24.3
21.1
24.1
10.4
3.1
Dominique Wilkins*
34
-3.45
22.2
18.7
26.2
6.9
2.6
*Less than 4000 career playoff minutes

There's actually two groups near the top of the list. Guys with a difference of 1 or more like Reggie Miller can say with confidence they improved, but there's a large group in the -0.32 to +.27 range who remained as effective in the playoffs as in the regular season, which is definitely an accomplishment given the competition and stage. Barkley doesn't have a title, but he shouldn't be blamed for disappearing when it matters because he was still a force as the round mound of rebound. Nowitzki and Ray Allen won titles in recent years, and given how well they play when the pressure is on they definitely deserve one.

Looking through the middle of the table, it's apparent how common it is for a player's statistics to go south in the playoffs. The average reduction was -1.21, where you'll see a host of stars with titles to their names like Kevin McHale and Paul Pierce. Shaq and Kobe had differences worse then -1, but they also have a few titles to their names. Magic Johnson deserves his nickname, but he also didn't literally get better in the playoffs. A drop-off over one PER isn't the difference between a bust and a star, but it's also enough to suggest there is a definite disparity between the regular season and the playoffs.

The worst performer was Dominique Wilkins, a flashy dunker who never won a title and subsequently was known as a big scorer who wilted at important moments in the playoffs. Conventional wisdom was correct, as it was for Karl Malone. Both these superstars had falloffs over -3 on their PER scores, which is such a significant drop they have little to argue in defense. Others at the bottom of the table won't shock most NBA fans -- David Robinson, Kevin Garnett, Chris Webber, and Patrick Ewing. Guys like Stockton and Payton are surprising, but the former never won a title in a long career and the latter was a bit player on the Miami Heat led by Wade. I'm surprised to see Nash on the list, but since he also doesn't possess a title it's not entirely a shock.

The biggest surprise of the results is definitely Larry Bird with a PER difference of -2.79, and I understand why people will want my head for even including him in the table. He was hobbled in the playoffs a couple times, but this is over his entire career, and he's known as a player who will run through a wall if it means a win for his team. In one of his MVP seasons, he lost to the Lakers in '85 with a poor shooting slump. Back troubles were one explanation, but others said he had trouble against Michael Cooper's defense. Maybe Bird is a case where bad timing on injuries unfairly impacts his playoff stats, but wouldn't every player say that? If normal probabilistic variation is the mechanism behind playoff success and victory, then isn't there no meaning behind the clutchness ethos we hear in sports? I don't think I'd go quite so far.

LeBron James receives more negative press than anyone else right now for his playoff production, but he is far from the worst and even at a reduced level is better than almost anyone else. He also had one of the best single season playoff runs in history sporting a godly 37.4 PER with game averages of 35, 9, and 7 before he was vanquished by the Orlando Magic. However, he hasn't always delivered, and last year in particular he was underwhelming.

Further into the Numbers

Perhaps it's better to visualize the data in a graph so the players are grouped more intuitively. The one included below displays a player's PER difference along with his playoff PER to separate players based on their overall performance. The winners in this are the players in the top right hand corner, while the guys in the bottom left corner suffer the indignity of both functioning poorly compared to everyone else and compared to their regular season play. You can also see that while Isiah Thomas should be commended for his improvement, but Karl Malone still had a better PER. This comments on the exaggerations of the media -- even a disappointment like David Robinson is still fantastic, as the differences are fairly small. Isiah Thomas certainly didn't will his team to victories on his own, and he also didn't turn into Michael Jordan or even regular season Lebron.



I'd also like to reiterate that I am in no way endorsing PER as the be-all, end-all of player rating measures. This method does not with 100 percent accuracy tell you who achieved and who didn't. The innate problem with PER is that it relies on box score stats, which, for instance, don't track individual defense well. Gary Payton's forte was tough man to man defending, as the Glove was arguably the best antidote (i.e. could slightly slow him down) to Michael Jordan. That doesn't show up in PER, and any increase in ability at that end in the court wouldn't be detected. It doesn't, however, excuse the fact that his accuracy submarined while shooting less the playoffs than the regular season. I will note that he was stellar in 1996 when he lost to the Bulls on one of the best teams to never win a title.

One cause for a negative PER difference is relativistic. It's Kevin Garnett syndrome -- if you play at a high level of energy in every game, or in every single moment of your life, when the playoffs roll around you can't up your game because the dial is already cranked to 11, and as a result it looks as though you're going backwards when it's actually everyone else moving forward while you stay still. This could also explain Larry Bird, who no one will confuse with playoff disappointment.

But how important is that average drop of 1 PER? Hollinger, the man behind the metric, now publishes estimated wins added derived from PER. Essentially, you find the replacement PER difference of the player (normally you take the difference between the player's PER and a replacement level guy of around 11), multiply it by minutes, and convert it to wins. A difference in 1 PER over 4000 minutes equates to roughly 2 wins given away. It's not a disaster, but in a close playoff series that dip in performance could swing one team to victory. The most extreme player in this regard is Karl Malone, who strangely never missed the playoffs, and whose drop in PER amounted to an estimated 12.5 wins "lost." If you adjust his PER reduction in terms of everyone else (the average loss is 1.21 PER) it's still 7.8 wins. By contrast, Olajuwon added an estimated 7.0 wins throughout his career in the playoffs, although he was trumped by a lesser player who leads the next group at 7.3 wins.

Role Players

The next table is comprised of "role players," whose regular season PER numbers were below 18. Again, I'm not saying that's the sole factor in determining whether or not you're a role player, but it was an easy way to separate Derek Fisher from Dwyane Wade. There were 19 players with 4000 playoff minutes and over who qualified, and I included one more role player, Steve Kerr, who was almost like a good luck charm in winning a title. The same patterns are observed with this group as the last: few actually get better, the average drop was -0.89, title winners occupy both the top and bottom of the table, no one had a mythic rise in PER, and conversely no one had a mythic reduction.

Name
Rank
PER Difference
Regular season PER
Playoff PER
Points/
40 mins
Rebounds/40 mins
Assists/40 mins
Robert Horry
1
0.94
13.5
14.5
11.4
8.0
3.4
Derek Fisher
2
0.39
11.9
12.3
12.7
3.4
3.8
Michael Cooper
3
0.15
13.0
13.2
13.3
4.8
5.9
Dennis Johnson
4
-0.08
14.5
14.4
17.8
4.5
5.8
Maurice Cheeks
5
-0.12
16.7
16.6
15.8
3.7
7.6
Ben Wallace
6
-0.43
16.7
16.3
8.3
12.9
1.5
Charles Oakley
7
-0.43
14.0
13.6
12.1
11.3
2.2
Tayshaun Prince
8
-0.54
14.9
14.4
13.4
6.0
3.0
Dan Majerle
9
-0.75
13.9
13.1
13.6
5.7
2.9
Bruce Bowen
10
-0.75
8.2
7.5
7.7
3.5
1.7
Richard Hamilton
11
-0.93
17.4
16.5
20.6
4.0
3.7
Sam Perkins
12
-0.99
15.2
14.3
15.6
7.8
2.1
Joe Dumars
13
-1.00
15.4
14.4
17.2
2.5
5.0
Horace Grant
14
-1.08
16.8
15.7
12.4
9.4
2.3
AC Green
15
-1.39
14.9
13.5
12.8
10.5
1.3
Rasheed Wallace
16
-1.44
16.9
15.5
16.3
7.5
1.8
Byron Scott
17
-2.14
16.0
13.8
18.3
4.0
2.9
Danny Ainge
18
-2.17
14.3
12.2
15.1
3.5
5.2
Dennis Rodman
19
-2.53
14.8
12.3
9.0
14.0
1.7
Steve Kerr*
20
-2.58
13.7
11.2
11.1
2.2
3.1
*Less than 4000 career playoff minutes

The fact that Robert Horry is the king of this group in a way validates the study. His nickname is Big Shot Rob, and he had a number of key moments for several different teams in the playoffs. He also has more championship titles than anyone except for Celtics players from the 50's and 60's. Another rarity is that he was a part of three different multi-title teams: the Olajuwon-led Rockets of the mid 90's, Shaq and Kobe's Lakers at the turn of the century, and two of the Spurs' championships from the last decade. He undoubtedly upped his game, but let's not get ahead of ourselves here -- even with his improvement he's miles away from having the same impact as Karl Malone, LeBron James, or David Robinson.

Inversely, Steve Kerr's five titles with the Bulls and Spurs are definitely more happenstance and luck than his superb "winning" skills. He's had his moments, sure, but he's also had both his shooting percentages and shots per minute rate go down. Kerr has the worst PER difference, and the next two guys also had success. Rodman is not the type of guy to dominate the box score stats except for rebounds; however, even his prodigious talent at crashing the boards was lessened in the playoffs. Danny Ainge has the moniker of a clutch player because of his playoff glories, but the evidence points otherwise.

A couple Lakers -- Michael Cooper and Derek Fisher -- sport some of the top marks, but in Fisher's case especially I have to preface that statement with how that still didn't turn him into a star. It's also much easier raising your game when there's a lot to improve. It is, however, interesting to see what championship pieces improved. For example, Ben Wallace and Prince barely saw a dip in their numbers while Rasheed Wallace and Rip Hamilton could not say the same.

Conclusion


While there are problems with the statistical metric PER, the comparisons here are limited to a player versus his own stats in the same season. The results mainly agree with conventional wisdom, but there are also a few surprises. Charles Barkley has no ring, but it's certainly not for how he did in the playoffs. Larry Bird's stats were worse even though that's contrary to the NBA's lore. Two others were even worse though; Malone and Wilkins lost over 3 PER from their regular season scores to the playoffs. That kind of difference is large enough to distinguish between an MVP winner and a fourth place finisher in the voting -- significant, yes, but Malone was still a good player.

Overall, stats go down in the playoffs, but not at a monstrous rate. The average star player will lose about 1 PER between the regular season and playoffs, and few are able to buck that trend. Many barely had a change in their numbers, which is accomplishment in itself due to the higher competition and especially with someone performing at such a high level like Michael Jordan, and most others received a sizable reduction in play. There's a short list of guys who can legitimately claim they got better when it mattered, as they experienced a significant boost in PER at 4000 minutes and more -- Isiah Thomas, Hakeem Olajuwon, Reggie Miller, and Robert Horry.

No comments:

Post a Comment

Post a Comment