March 13, 2003
Success Cycles Revisited
A little over a year ago, Jonah Keri wrote an article for BP on the concept of "success cycles." In that article he described the apparent cyclical process of team-building--that organizations rebuild, contend, get old, and rebuild once more--while making the point that successful organizations recognize where they are on the success cycle, and make decisions accordingly. I was excited by this concept, at first, as it was both logical and intuitive, and began to do some charting of it. In the end, however, I have come to the conclusion that success cycles do not exist.
Now, don't get me wrong. I agree with Keri that a team must know the state of its organization, and those it competes with, in order to do the best job it can. For instance, it's of no use to acquire expensive stop-gap players if your team has no realistic chance at competing. But if a couple of expensive free agents are the difference between a playoff berth and sitting at home, those acquisitions are worth their weight in gold. In other words, I agree completely that intelligent and successful organizations must have a realistic idea of how good they are now, how good they expect to be in the immediate future, and make all their decisions based from there.
I disagree, however, that team success is--or should be, in any way--a regular cycle. There are many historical examples of teams contending for far longer than they "should," while organizations like the Brewers waver between mediocrity and awfulness for over a decade without ever pushing forward.
After an examination of overall team records since 1973 (which excludes post-1977 expansion), I don't see any evidence that teams go through predictable success cycles: that is, given a record of X and a previous record of Y, there's no reason to expect that increase or decrease in team wins to continue. Plotted on a graph, a true "cycle" would look more or less like this:
Or, given a 10% random variation--from eight games below expectation to eight games above expectation--something similar, like this:
Whereas a team playing .500 ball all the time with a +8/-8 swing in luck might look like this:
You can see where there's going to be some confusion here. You can still pick up the larger swings in the first graph, though, where the second doesn't sustain much of anything. That randomness is a little overblown, too: I just tossed a spreadsheet function in there, but in real life I think teams are much more likely to have a +1 season than they are a -8, which would bring out the cyclic nature (if it exists) and make the .500 team look more .500-ish.
There are 25 teams with a substantial history since the advent of free agency. Twenty of those 25 teams don't show any cyclical behavior when eyeballed. For purposes of the graphs reproduced here, I'm here graphing their post-1972 records with the 1981 season dropped, and the '85 and '94/'95 seasons pro-rated to a 162-game full season with a rough adjustment. If I'd thought there was anything to be gained by looking at this further, I'd have gone back and used win percentages.
Boston Red Sox
As you can see, Boston bounces around--mostly over .500--for 30 years. And Boston fans think they have it bad? They should see the Cubs' graph.
New York Yankees
Like Boston, almost entirely bouncing above the .500 line except for the 1985-1990 decline (from 97 wins to 67) and the 1990-1993 rise (67 to 88 wins), which forms a nice little bowl in the middle.
Kansas City (random)
As you might expect, Kansas City checks in with a few spikes up and down off the long, steady descent.
And on the other side, Seattle, with random deviations off a franchise-long crawl to respectability:
St. Louis has a nice up-down pattern with no cycles:
Other random-looking teams in a boring rundown section:
What about the others, though? Do the other teams show cyclical behavior?
Some do. Oakland, for instance:
Spend some time below .500, spend a couple years over .500, go back under .500
Atlanta shows some of this:
The Braves spend some time being bad, then climb up for four years, return to being bad, get good, and then stay good for over a decade. Side note for the very observant: this chart doesn't include the 1985 adjustment.
And then there's Baltimore, which doesn't look entirely random, but...
It starts out good, gets worse and worse, rises, stays over .500 for a while, but then gets worse. I don't think it's cyclical, but I can see where there's a top to start, a drop, then an ascension, and then a drop--so I didn't count it as obviously random.
The other charts that don't immediately look like they're random:
That there are five teams of 25 that show any kind of cyclical behavior is enough for me to discard the concept. Of course, with enough squinting you can see patterns in anything. I could have counted Baltimore as random, or invented a "steady decline or improvement" category, for that matter. But strong success cycles would have made these choices easy: we'd see the same wave form over and over.
There is a significantly higher tendency for a team to move toward .500 than there is for it to follow-up on a one-year improvement or drop. Using post-1973 team-wins here, I had about 750 team-years to work with, and the correlation coefficient between a team's record and their regression to .500 (as the average of last year's record and a .500 record) was .446. While looking toward momentum (as last year's record plus last year's gain/loss in wins) only got .300, which is much weaker. At the same time, that does tell us a little about the way teams get better and worse from year-to-year.
That there are not regular success cycles makes sense to me. Organizations vary in quality, which means that while it's hard to build success, sustaining that success is relatively easy in comparison. Franchises that had great farm systems three years ago might be barren today, and stocked again in two years. Prospects flame out or challenge for jobs on different schedules: the Indians might give Einar Diaz a multi-year contract one year because they're bereft of catching prospects, but the next year they might have a system led by Victor Martinez, with Josh Bard already playing well for the big club. At the same time, free agency carries its own variability: one-year stopgaps like Bret Boone or John Burkett can have historically great or awful seasons, late into their 30s. Veterans don't decline in straight lines.
We can see, too, that teams with advantages can do great things. A lot of money can mean the that Red Sox or Yankees can stay above .500 almost indefinitely. I think this is because both teams can afford expensive stop-gaps. If the Yankees need to pay a third baseman to cover until Drew Henson is ready, they can go pick up a quality player like Robin Ventura, rather than settling for David Bell. Front office management can be the difference: the A's aren't going to drop to 70 wins anytime soon, barring radical changes in their financial or front-office situation; and for all the shots we've taken at Atlanta, they've sustained a long run of success that almost every franchise must envy. One of the ways we might measure organizational health in the future would be to look at their ability to defy the regression. Organization Above Replacement, perhaps?
The idea of the Success Cycle is seductive, like the concept of biorhythms. It appeals to our sense of larger order, and intuitively we think that teams go up and come down. But as I look more closely for actual evidence of its existence in baseball, I find only that if there is such a thing, it is lost in the much larger forces of organizational quality, team strength, and luck (uh, I mean, random variance).