Posted by Neil Paine on January 8, 2010
Recently I was reading Bill James' 1994 book The Politics of Glory, about the baseball hall of fame, and he devoted a chapter to the debate of Don Drysdale vs. Milt Pappas. If you're unfamiliar with the players, all you need to know is that Pappas and Drysdale had virtually identical won-loss records (209-166 for Drysdale, 209-164 for Pappas), Drysdale had an ERA+ of 121 vs. Pappas' 110, and Drysdale was inducted into the Hall of Fame with 78.4% of the vote in 1984 after 10 years on the ballot (never receiving less than 21% of the vote), while Pappas received 1.2% of the vote in 1979 and dropped off the ballot forever.
Knowing what sabermetricians know now, there's no doubt Drysdale was a better pitcher than Pappas -- his ERA, FIP, and other component elements of skill and performance were superior to Pappas'. But in the dark ages of the late seventies, the BBWAA was still pretty obsessed with pitcher W-L records when it came to HoF voting, and Pappas actually had the superior record. Besides, even looking back now, was Drysdale that much better than Pappas, to the point that the former starts with 20 times as much HoF support and builds to induction a decade later while the latter drops off the ballot as an afterthought? Why was the perception so markedly different between the two pitchers?
James claims that the difference lies in Drysdale's "peakiness" -- that although Drysdale and Pappas had equal W-L records, Drysdale was able to cluster a larger portion of that production in a handful of seasons (25 wins in 1962, 23 in '65, 6 straight shutouts in '68) which proved far more memorable to the writers than Pappas' 15-ish wins year in and year out were. On the face, this seems unfair, a case of popular perception being dictated more by flash than substance. But James wanted to know if "peaky" pitchers like Drysdale actually contributed more to their teams' pennant chances than a consistent pitcher like Pappas, so he built a simulation that ran fictional careers for two pitchers like Drysdale/Pappas, one of whom had a high peak but little production aside from those seasons, the other with a moderate peak but the same win total because of a longer, more consistent career. In James' study, it turned out that the peaky pitchers, the Drysdales, actually led their team to more pennants than the Pappases, validating the BBWAA's perception that Drysdale was better despite equal career W-L records.
What does any of this have to do with basketball, you ask? Well, I wanted to answer the same question, except instead of a star pitcher, we'll look at an all-star player. I took every player in the Win Shares era who was named to at least 1 All-Star team in their career, and split them into two groups based on the "peakedness" of their career Win Shares distribution. The two groups looked like this after I made sure they both had the same # of career WS:
As you can see, both players have the same career value, but the consistent player's career rises, peaks, and winds down gradually, while the peaky player is hot and cold, mixing big years with abruptly mediocre ones, culminating in a big drop-off at the end of his career. Which player do you think would add more to his team's chances of winning an NBA title if he was the lead dog on a team surrounded by average teammates?
Well, like James, I put together a Monte Carlo simulation in which the both players would play out 20-year careers according to the distributions above (with some random fluctuation built into both playing time and per-minute performance) and tracked how many times their teams won the NBA championship. I did this 10,000 times and then tallied the results for both our peaky player and our consistent one. Which one led their team to more titles in 10,000 careers?
|% of Years||0.88%||0.93%|
Contrary to James' findings in baseball, my simulation found that building around a consistent player whose peaks are not as strong as some of his fellow All-Stars, but whose valleys are not as low, actually led to a greater # of championships than a peaky player whose career was up and down and who couldn't be counted on to be an All-Star caliber player in any given season, especially late in his career. This makes sense if you remember that The Politics of Glory was written before there were wild card teams and an extra round of the playoffs -- in baseball prior to the strike, peak performances often meant the difference between making the playoffs and sitting at home, because teams had very little margin for error (only 4 out of 28 teams made the playoffs in 1993, and for the majority of MLB history only 2 teams advanced beyond the regular season -- you either won the pennant or nothing).
In today's NBA, however, 16 of 30 teams make the playoffs; even when the league only had 23 teams in 1984, 16 made the postseason. So flukish peak performances are not really necessary to carry a team into the postseason, and once there, every team has at least some chance of going all the way. That's why consistency is key -- when surrounded by average teammates (as our All-Stars were in this simulation), you're going to make the playoffs a great deal of the time no matter what, so the ability to not hurt your team once you get to the playoffs is as important as carrying them there in the first place.
Then again, the difference between the two groups was incredibly small, 110 titles in 200,000 seasons. In a 20-year career, that's a difference of just 0.011 championships, essentially negligible. And if you look back at NBA history, maybe the real moral of the story is that it's best to have a player who's "consistently peaky" -- that is, a superstar who plays at an insanely high level every year. Probably the only titles in my lifetime that were led specifically by a Drysdale or a Pappas, and not a freakish megastar hybrid, were a trio of Detroit Pistons championships: 2004 by Chauncey Billups (Pappas), and 1989-90 by Isiah Thomas (Drysdale). Thomas is in the Hall of Fame already, while Billups has just a 25% chance of making it according to our HoF Probability metric. So what do you think, will Mr. Big Shot suffer the same fate as Milt Pappas?