February 18, 2013

Overthinking It

Why There Probably Are No Next Orioles

by Ben Lindbergh

Printer-friendly

Contact Author

“I know a lot of the national reporters say we’re going to finish last and lose a lot of games again. You know what? Oakland was supposed to be last [in the division] last year, Baltimore was supposed to be last, and they both ended up making the playoffs.” —Astros GM Jeff Luhnow, February 5.

Sometime between now and Opening Day—if you haven’t already—you’ll probably hear someone speculate about the surprise team(s) of 2013. Every spring, fans and analysts attempt to predict which teams will surpass the expectations of PECOTA and the pundits. Most of those predictions, of course, don’t come to pass. It’s tough to beat the stats, the oddsmakers, and the combined predictive powers of people who spend large chunks of their lives watching and reading and writing about baseball teams. Especially since some of the people who can beat the consensus consistently start publishing their predictions, the consensus becomes a bit better and harder to beat.

So while we all have hunches or get good vibes from certain teams, our hunches and vibes probably aren’t predictive. But it’s still fun to try to outsmart the system, and being the first to see something coming can make you look like you’re better at baseball analysis than everyone else. That means we’ll see more of the same old articles, but now there’s something new. We’re no longer settling for predicting the next “surprise teams.” We’re predicting the next Orioles and A’s.

The Orioles and A’s, as everyone knows, made the playoffs last season and won well over 90 games despite not having done either of those things for some time—15 years in Baltimore’s case, and six in Oakland’s. Someone, somewhere, might have predicted their success, but it wasn’t at Baseball Prospectus, and it wasn’t Billy Beane. It was a wonderful story, and we’re wired to want more wonderful stories. So now we need new Orioles and A’s. Some people started looking for them last year; others started looking last week.

But there’s a catch: Teams like the 2012 Orioles and A’s don’t come along often. You’ve probably heard of the Availability Heuristic, one of those handy hallmarks of human psychology discovered by Daniel Kahneman and Amos Tversky. This one, in the words of Wikipedia, is “the tendency to make a judgement about the frequency of an event based on how easy it is to recall similar instances.” In other words, if we just saw something happen, particularly in a memorable way, we’re more likely to overestimate how likely it is to happen again. Last year, we saw two teams go from afterthoughts in April to playoff teams in October in about the most memorable way possible. Maybe we’re overestimating the odds of it happening again.

It’s not that rare for a team to improve by at least 20 wins from one season to the next: Before last year, it had happened 26 times in 16 seasons during the wild-card era. Last year, it happened twice, which wasn’t that unusual either. What was unusual was the way it happened, particularly for the Orioles. The O’s improved by 24 wins, which had happened only 10 times in those 16 seasons. Their 20-plus-game improvement put them over 90 wins, which had happened only six times. (It’s probably easier to take a big jump from terrible to not too bad, like the 2003-4 Tigers, 2004-5 Diamondbacks, and 2008-9 Mariners, than it is to take a same-sized jump from not too bad to 93 wins.)

Maybe more importantly, no one saw Baltimore’s success coming. You can’t say that about the 1999 Diamondbacks, who signed Randy Johnson and Steve Finley and traded for Luis Gonzalez, among other moves, after a lousy expansion season. You can’t say it about the 2008 Rays, whom PECOTA projected to be 22 wins better than they had been the year before. And you won’t be able to say it about the 2013 Blue Jays, who might enter 2013 as the consensus favorite in the AL East after winning 73 games last year. The Jays could win 93 or more this year, but between the players they’ll get back from injury and the ones they’ve acquired over the winter, no one would be especially surprised. If we want true comparables—a team that improved by at least as many wins, won at least 90, and did it all without any warning or pre-season buzz—we’re down to a handful in the past couple decades.

Trying to isolate common traits of those teams so we can see the next surprise coming is an exercise as frustrating as earthquake prediction, one of Nate Silver’s subjects in The Signal and the Noise. Seismologists can estimate the probability that there will be an earthquake of a certain magnitude, somewhere in the world, over a particular period. What they can’t predict with any accuracy is exactly where and when it will occur. Similarly, we can estimate the likelihood that some team will exceed its projection by a given number of games, but we can’t tell you which team it will be. There are too many aspects of baseball performance we either don’t yet know how to predict or will never know how to predict, because they’re not predictable. (Fortunately, when PECOTA fails to predict something, nobody dies.)

Efforts to identify short-term earthquake indicators have failed. Some earthquakes are preceded by a series of mini-quakes that, in retrospect, look like warning signs. But other clusters of mini-quakes never lead to a larger one. Seismologists make the occasional correct call, but only because they’re such prolific prognosticators. If we predicted that every team projected to finish under .500 would actually end up 20 games over, we’d have a lousy success rate but be right once in a while.

People have proposed indicators to pinpoint surprise teams, too. In The Baseball Book 1990, Bill James laid out several that he used to predict how teams would do in the upcoming season. The indicators were based on Pythagorean record and some simple principles: good teams tend to get worse and bad teams tend to get better; teams that improve in one year tend to decline the next, and vice versa; young teams tend to improve, while old teams tend to decline; teams that played well late in the previous season tend to sustain their success, and so on. Building on James’ indicators in 2011, Dave Fleming of Bill James Online came up with a list of 14 questions designed to predict whether a team would improve:

1. Are the hitters young or old? Are their best years to come, or are their best years behind them?
2. Are the pitchers young or old? Are their best years to come?
3. Did the team improve during the second half of the season?
4. Did the team do well during August/September?
5. Assuming that the team posted a losing record: Did the team do well the year before that? Is there a recent history of success?
6. Did the team’s AAA organization do well?
7. Did the team’s AA organization do well?
8. Did the team underperform their Pythagorean W-L record?
9. Does the team have good management? Do they utilize talent well?  
10. Is the team’s division open for challengers?
11. Did the team acquire new talent in the off-season?
12. Is the team getting significant players back from injuries?
13. Does this team have clear and present talent?
14. Does this team have a story for how it will be successful?

Some of these are subjective, so it’s not clear how much value the questionnaire adds. If we’d applied this test to the Orioles a year ago, about half of our answers would have been, “No.” The A’s would have done a bit better, but they wouldn’t have aced it. Fleming, for his part, picked Cleveland as his surprise team. The Indians improved by 11 wins but finished 80-82.

PECOTA doesn’t foresee any team improving by 20 wins this season—the Red Sox, currently projected to gain 17 wins, come closest—but it’s not in PECOTA’s nature to predict extremes. (PECOTA doesn’t project anyone to hit more than 40 homers, either, but someone has in every season since 1986.) But as Sam Miller, Paul Sporer, and I discussed on the Cleveland season preview podcast earlier this month—before Michael Bourn signed—the Indians are a popular pick again.

A week ago, Baseball Nation’s Grant Brisbee wrote about whether Bourn would make Cleveland a contender. “It’s probably not going to happen,” Grant said. “But at least one team will be this year’s Orioles, and the Indians are just as good a pick as any.”

When Grant, who is as wise as he is witty (note: he is very witty), says, “at least one team will be this year’s Orioles,” he doesn’t mean, “at least one team will go 29-9 in one-run games, outplay its Pythagorean record by 11 games, and finish with 93 wins.” He means at least one team will be significantly better than we expect, which will probably be the case (as it is in most seasons).

Two things. First, as Grant acknowledges, calling a team the next Orioles effectively disqualifies it from consideration. What made the Orioles the Orioles (and the A’s the A’s) is that no one conceived of the possibility that they could do what they did. Now that they’ve done it, of course, it’s easier to imagine that anyone could. But the Indians have reached the point at which they’re too popular a pick to provoke the same surprise.

Second, using the 2012 Orioles as a proxy for “surprise team” raises expectations awfully high. The second wild card makes it more likely that a team can play over its head long enough to luck into a playoff spot, so in that sense we are more likely to see a surprise team today. But the Orioles won often enough that they would have made it to a tiebreaker with Texas with only one wild card. There’s always a surprise team, but there’s not always an Orioles.

The Cleveland love hasn’t been limited to the last couple weeks. Erik Karabell expressed similar sentiments on ESPN’s now-defunct Baseball Today podcast after Nick Swisher signed.

“What if they’re the Baltimore? What if they’re the team that can contend? We don’t know that”, Karabell asked his co-host, Keith Law.

Law wasn’t buying the “They could be Baltimore” argument.

Do you want to start spending money on the thought that, ‘We’re going to do what Baltimore did,’ which has to be one of the most remarkable one-season turnarounds of the last, I don’t know, 10-20 years, in terms of what was on the roster? … I would not want to run my club thinking, ‘We could be this year’s Baltimore.’

Law was right. Expecting the improbable isn’t the best way to run a team. And it’s probably not the best way to make pre-season predictions.

Thanks to R.J. Anderson and Dave Studeman for research assistance.

Ben Lindbergh is an author of Baseball Prospectus. Follow @benlindbergh
Click here to see Ben's other articles. You can contact Ben by clicking here