I DON'T LIKE SPORTS TALK RADIO. It is a betrayal to the business I'm in, but my saying it ain't gonna change the business. So I'll get it off my chest.
As much as sports talk radio hosts (whom I occasionally respect) try to keep things intelligent and progressing towards a reasonable conclusion, the conversations often get long, tedious, and highly prejudicial. Some guy who thinks Michael Jordan is the greatest because "Dude, I've seen him like 30 times in person and not one of yo phat dogs can stop him," is not going to be convinced by any argument, even if someone says that Bill Russell won 10 titles, that Wilt Chamberlain won 7 scoring titles AND 11 rebounding titles AND an assist title, or that Oscar Robertson had the same fierce competitiveness and couldn't be stopped until he had a triple double every night.
As much as I dislike the format and many of the personalities involved, sports radio does occasionally focus on that one question that we all think about: Who's the best player ever? My brother and I pursued the question all throughout our childhood until we had discussed all angles, which may be why current discussions bore me.
What I can now offer to the discussion is something that I couldn't do as a card-collecting, stat-quoting, always-right teenage sports fan (though I still hated sports talk radio): a collection of research and experience that can summarize players' contributions with numbers and established methods.
I try not to say "numbers" too much in the Journal because the term scares people. (The movie A Civil Action is a prime example of how science scares people. The movie itself mentions this tenet, but the movie also epitomizes it. The book on which the movie was based contained all sorts of science that the movie left out because you, the audience, are just too stoopid.) What numbers do, however, is provide a nice summary of thoughts that can be manipulated with rules, rules that, amazingly, have been generally agreed to. For example, if 1 championship is good (the "thought", in this case), 1+1+1+1+1+1 = 6 must be better (addition being the "rule"). We don't have to go into the details of each and every one of those championships to say that a team that wins 6 titles is probably better than one that wins 1 title. 6 > 1 is a pretty good summary of the info. It isn't the whole story, but the details would have to be extremely convincing to change our conclusion. That is the premise of the comparisons made below and something to keep in mind as you see the numbers for the three greatest players of the modern era: Magic Johnson, Larry Bird, and Michael Jordan.
In about ten separate articles in JoBS, I have started to write down the ideal way to evaluate individual players. To this point, however, I don't think that method has ever made it to the final cut. So here it is:
An ideal method for evaluating the ability of an individual basketball player would be to simulate his/her performance on every possible team vs. every possible opposing team. The winning percentage of all the teams that the individual plays on represents his/her context-insensitive winning percentage and accounts for all of his/her skills and deficiencies.
For example, in our recent reality, we know that Michael Jordan and his 1998 cast of former Bulls won 62 out of 82 games and the NBA title against a supposedly watered-down league. That's one simulation. Simulating another reality would have 1998 Jordan playing alongside Mike Woodson, Michael Cage, Quintin Dailey, Benoit Benjamin and the rest of the 1988 Clippers; they probably wouldn't win half of their games unless we pitted them against the 1948 NBA in some sort of Pleasantville take-off (and there is still some doubt that they'd win half). Of course, we could simulate what so many New Yawkah's have been dreaming about, too: Jordan with Ewing and Bernard King and Charles Oakley and, oh, let's say, Maurice Cheeks against all comers from history.
With those simulations and about 50,000 more (at a minimum), we could then get a sense for Jordan's context-insensitive winning percentage. Then we'd have to do it for Magic and Larry to get their winning percentages. With those numbers, I would actually agree that the comparison of these players is accurate. The argument over who is the better player could be put to rest. Sports talk radio would die!
Unfortunately, this is impossible. Jim Rome lives on.
There is one key aspect of the above theory to carry through in a practical look at Magic, Larry, and Michael. That is the idea of different contexts. So many of the arguments about the three players focus on Magic having better teammates or Michael playing against a watered-down league. It makes sense to try to account for these different contexts in any practical evaluation of their abilities, try being the operative word.
The basic evaluation method will be the individual win-loss algorithm outlined in the previous two articles (The Need for Individual Win-Loss Records and What Do Individual Win-Loss Records Mean?), specifically, Method 3 of the second article. Briefly, this method compares a player's offensive rating to his defensive rating and estimates a winning percentage based on the Pythagorean Relationship.
Fortunately, this method requires the input of team statistics, which provides a route for simply looking at the influence of different contexts.
The first way to look at the three players is what they actually did in the contexts they actually had. Magic played with the Lakers, Bird with the Celtics, and Jordan with the Bulls. In these contexts, the estimated win-loss records for the three players are:
Player | W-L | Win% |
Johnson | 149-17 | 89.7% |
Bird | 157-21 | 88.2% |
Jordan | 193-19 | 91.2% |
Jordan won the most and had the highest winning percentage. Johnson and Bird had similar totals, though, as you can see from the charts below, Johnson didn't show the drop-off that Bird did in the last few years of his career due to injuries.
The seasonal progressions (shown behind the above links) that define these players' careers are fascinating in themselves. All three players tended to show a general increase in offensive ability through time, even as their raw skills declined, but a decrease in defensive ability as indicated by their stops per possession (Jordan didn't show this as much). Johnson picked up his 3-point shot in 1989 to raise his offensive rating while his floor % remained fairly steady or slightly dropped. If you take out Bird's last four seasons, he has a win-loss record of 128-13 for a winning percentage of 90.5%, slightly better than Magic's. Jordan's win-loss record reflects quite well what people said about him -- that his early years were great, but not quite as efficient as his later years.
A couple of other relevant notes:
As mentioned above, one of the goals of this work is to look at the three players in different contexts. Each player's success was due in part to the teams that they played on, so it is logical to check to see what would happen if you put their numbers in the context of other teams. In this case, let's look at what happens if each player is placed in the context of the other two players' teams.
Magic Johnson all of a sudden has the highest winning percentage of the three players when the team context is either the Lakers or the Celtics. With the Bulls as context, a team very unlike the Lakers and Celtics in terms of the number of Hall of Famers as teammates, Jordan has the best winning percentage:
Lakers | Celtics | Bulls | ||||
Player | Win-Loss | Win% | Win-Loss | Win% | Win-Loss | Win% |
Johnson | 149-17 | 0.897 | 150-17 | 0.901 | 156-16 | 0.904 |
Bird | 153-23 | 0.867 | 156-21 | 0.881 | 157-26 | 0.860 |
Jordan | 180-24 | 0.884 | 181-24 | 0.884 | 194-19 | 0.912 |
For whatever reason, Bird and Jordan had their best winning percentages with their own teams. Johnson had his worst with his own team, though his numbers are the most consistent from team to team.
Whether this is "real" or not, it's interesting to come up with reasons for why it might be real. For example, it is difficult to dispute that Johnson was the best passer of the three, followed by Bird, then by Jordan. By being a good passer, Johnson would have consistently brought his teams up to a comparable level. Jordan being on the other extreme would have had to make more jumpers early in his career in order to stay out of the middle where Kareem Abdul-Jabaar or Kevin McHale or Robert Parish roamed. Bird fit somewhere in the middle, needing his shot, but also having that great interior passing ability.
It may be all wrong, but it's interesting that the numbers suggest such a story.
Another good context in which to evaluate the players is the average team. Specifically, place each player in the context of the average team for each of their years in the league. So, if they all played on 0.500 teams every year, how would they look? Part of their greatness, their ability to make great teams, is not shown here, but it is a good unifying context that seems "fair" to all.
Player | Win-Loss | Win% |
Johnson | 145-21 | 0.875 |
Bird | 149-28 | 0.841 |
Jordan | 183-27 | 0.871 |
As before, Johnson has the best winning percentage and Jordan has the most number of net wins (wins minus losses).
Finally, let's look at what happens if you put these players on two different teams with nearly identical offensive and defensive ratings. Specifically, the 1995 Seattle SuperSonics and the 1995 Utah Jazz were both very good teams, winning 57 and 60 games, respectively. But the Sonics' offense shot under 50%, shot a lot of threes, and did well on the boards, whereas the Jazz shot 51%, made nearly 200 fewer threes, and were among the worst offensive rebounding teams. The Sonics' defense was frantic and occasionally pressed in the George Karl style, averaging more than 10 steals per game but being fairly weak on the defensive glass. The Jazz defense was only slightly better than average in forcing steals but was one of the best on the glass.
Sonics | Jazz | |||
Player | Win-Loss | Win% | Win-Loss | Win% |
Johnson | 160-12 | 0.932 | 161-16 | 0.909 |
Bird | 162-21 | 0.887 | 162-27 | 0.855 |
Jordan | 190-22 | 0.897 | 190-28 | 0.870 |
It seems a little strange that Utah had the better record that year since Seattle's context makes each of these three look better than Utah's context does. Both of these teams were upset in the first round of the 1995 playoffs and probably could have used any of these guys.
Rather than changing context here, let's change the method slightly. Method 2 of What Do Individual Win-Loss Records Mean? calculates a win-loss record by counting the number of games in which a player's offensive rating is better and worse than his defensive rating. For no clear reason, it is my sense that this statistic is a closer representation of the ideal "context-insensitive" win-loss record, even though it still has the context within it.
Since game-by-game stats aren't available throughout these players' careers, the results of this method will have to be estimated, which they fortunately can be using certain statistical rules associated with binomial and normal distributions and I'm sure the mathphobes just left the room. I will move that explanation to a later time.
What this method does is reward players who are not only efficient but also efficient over a large number of possessions because that implies consistency. And what it says about the three greats is this:
Method 2 | ||
Player | Win-Loss | Win% |
Johnson | 678-228 | 0.748 |
Bird | 657-240 | 0.733 |
Jordan | 744-186 | 0.800 |
For the record, this estimate makes two assumptions.
This analysis is far from the ideal method for evaluating players. Simulating a team made up of players who, in reality, have never played together is a major goal of JoBS and a very reasonable goal. This 1999 NBA season, with all its free agents and swapping of players, will actually be an invaluable experiment for revealing how to do such a simulation. Maybe in a year or two, we'll have some good rules for how to do it.
In the meantime, the methods presented here for comparing the win-loss contributions of Johnson, Bird, and Jordan are the best available. To some degree, they do isolate the individual contributions from the context in which they play. The lengths of these players' careers and the fact that they played with several different sets of teammates helps make the long term evaluations more comparable.
I expect that when simulations of players on teams is truly possible (and verifiable), we will see similar relative results as seen here. By "relative", I mean that we won't see Johnson's teams winning 89.7% of their games in the ideal simulation, but we probably would see Jordan's and Johnson's team simulations winning slightly more than Bird's teams.
Don't argue with me unless you can show me good simulations.
(This look at the three players is somewhat unfair to Larry Bird, however. His last four years in the league were hampered by injury and clearly not what made him the Legend. Removing those last four years would place him on equal standing, if not slightly better standing than Johnson. But career longevity is one of the bases for evaluating players and, due to the convenient simplicity of equal 13 season careers for all 3, I kept those last four in there. Now, don't argue with me.)
It is not a complete study and it is definitely not conclusive, but I believe it does point to Jordan being the best of the three players. He appeared to contribute more net wins (wins minus losses) on a consistent basis from season to season, even if his winning percentage wasn't always the best. Johnson often had a comparable winning percentage and had a more consistent winning percentage from context to context, which strengthens his claim, but Jordan's enormous advantage in net wins and in the Method 2 calculation holds a little more weight with me.
Stepping back and thinking about the ideal ranking method, I can see the hypothetical melange of Jordan's teams having the highest winning percentage. His raw physical skills and competitiveness were expressed in so many different ways -- shooting, passing, rebounding, blocking shots, stealing passes, and just shutting down an offensive player -- ways that Johnson and Bird almost, but not quite, matched. Jordan scraped 0.500 ballclubs out of his early years when the other talent was probably worth 0.350. He turned very good defensive clubs assembled by Phil Jackson into outstanding offensive clubs that thoroughly dominated and set records for wins. He never lost in the Finals.
All three players can be seen as making bad teams decent and good teams great. Any simulation that doesn't show that is wrong. The problem with greatness is that it is so rare that it is difficult to define. It is a problem that confronts science of all sorts: rare events like solar eclipses or twins separated at birth provide us with exciting looks at truths that we can see no other way. When they are gone, we miss the inspiration that they bring. We cannot fully understand everything they bring because the circumstances under which they occur are so special -- we cannot generalize from them, one of the fundamental duties of science, because they are so unique. We can only hope that we are around long enough to see more rare events that inspire us even further.
Next time you listen to sports talk radio and the topic is the best player of all time, think a little bit. Try to frame those players in different contexts. Try to imagine what it really means to be the best player of all time. Maybe I'll start listening.
As almost always, thanks to Doug Steele for his valuable statistics. Though not explicitly used here, they played an instrumental role in the Method 2 estimation made above.