Royals continue to defy projections

Zimmermann, Kennedy set to duel on WNB (1:24)

The Baseball Tonight crew previews the Wednesday Night Baseball matchup between the Tigers and the Royals. Jordan Zimmermann is set to take the mound for Detroit against Ian Kennedy and Kansas City. (1:24)

Computers. How would we survive without them? They can find you a hotel room in Des Moines, Iowa. They can connect you with your cousin in Alaska. They can compute the square root of 1,786,443,956.

In other words, computers can do everything, right? Well, except for one thing, that is: They just can't figure out the Kansas City Royals.

This is a story about computers and the world's most unprojectable baseball team. A team that feels as if it's the only thing in the entire universe that computers can't measure.

If computers messed up your tax returns as badly as they've messed up on the Royals, you'd be audited by the IRS for the next 50 years. If they were as far off on your checking-account balance as they've been on the Royals, your entire living room would have been repossessed.

The Royals have existed in two very different worlds for the past three years.

There is the real world, where life for this team has been pretty darned awesome. And there is the computer world, represented by baseball's three most esteemed projection systems. One is the model devised by Baseball Prospectus co-founder Clay Davenport. A second is the PECOTA system currently utilized at Baseball Prospectus. The third is Dan Szymborski's ZIPS model, which we use here at ESPN.

In Kansas City, it's become a thing to make fun of those projections. So it's up to us to defend them, apparently. These are projection models designed by brilliant people. People who love baseball. People whose passion in life has been devoted to following, analyzing and quantifying baseball. People who have been right way, way, waaayyyy more than they've been wrong.

Nevertheless, before the 2013, 2014 and 2015 seasons, those folks assigned their trusty computers to whir and churn and process zillions of data points ... and then, somehow, those computers appear to have mistaken the Royals for the Padres in their preseason projections.

So there is math. There is science. And there are the Royals, who are now, happily, hard at work trying to make their modest 2016 projections look just as ridiculous as those 2013, 2014 and 2015 projections.

The Royals come into Wednesday night's ESPN game against the Detroit Tigers with an 9-4 record after 13 games. That puts them on pace to win 112 games. Which would be 29 more than the 83 that ZIPS projected. And 36 more than the 76 PECOTA projected. And 38 more than the 74 Davenport projected. And yes, the really smart people behind those projections are as confounded by this as we are.

"You know that there are mysteries about baseball that we and our little human brains can't grasp," Baseball Prospectus' Sam Miller wrote recently. "And you want to believe that those mysteries can be solved by the power of science and math. If we as a species can't project Wade Davis, how will we as a species ever fix global warming or colonize Mars?"

Hmmmm. Maybe because baseball is more unpredictable than rudimentary problems like that? But whatever. We're so fascinated by this crazy disconnect that we began searching for the real-life reason, or reasons, that the Royals keep defying those projections. And we're not alone.

"If there is a real cause, I would like to know what that is," Davenport told us. "I'm just not positive there is a real cause."

So is there? We went looking for an explanation, and we started with Royals general manager Dayton Moore.

"I think our team has the ability to win in multiple ways," Moore said. "We can win with our defense. We can win with our legs. We can win, obviously, with our bullpen. Our starting pitching, I believe, is more prepared to give us innings. And then there's the fact that these guys love to play baseball with one another and for our fans. I believe our environment is very, very strong. And because our environment is so strong, players enjoy playing here. I can't necessarily quantify how many wins that is. But I think it all has a lot to do with it."

OK, that's not just one explanation, clearly. That's about six explanations. So we decided to enlist Davenport, Szymborski and Miller in a detailed journey into pretty much every conceivable theory. It turned into quite the fun ride.


You don't need to relive the life and times of Wade Davis to know that, by almost any measure, the Royals have had the best bullpen in baseball over the past four seasons. So if you're looking for the easiest rationale for the great projection crisis, you've come to the right place.

"I really think the bullpen can explain at least half of that overperformance during the year," Davenport said. "Their bullpen is that good."

But how can just one aspect of a team be so disproportionately responsible for underestimating the entire team? Without turning this into a Ph.D. dissertation, we could sum it up a couple of ways.

Many stats define how great this bullpen has been. But it all comes back to this: When the Royals have led after eight innings over the past four seasons, they've gone 241-4. When they've led after seven, they've gone 211-10. So when you win pretty much every game in which you take a lead, over that many years, you're defying all probability and all norms of bullpen volatility.

Not to mention, Davenport said, that only in the bullpen does a team have "the ability to selectively use [its] best players at key times." So where most projections assume that players find themselves in a wide variety of random situations, "Wade Davis is not used in a random fashion," Davenport said, "but at very precise, high-impact times."

And one more reason consistently great bullpens are projection busters? "The team with the best bullpen is virtually always going to outperform its projections," Miller said. "But that team with the best bullpen is extremely hard to identify in advance."

Except, of course, in Kansas City, where the Royals apparently are allergic to the bullpen roller-coaster ride that afflicts every other franchise on earth.


We know the Royals catch the ball. But they're more than just a Web Gem reel. They've led the major leagues with an incredible 196 defensive runs saved over the past four seasons, according to FanGraphs. No other team in the American League is within 132 of them! (Baltimore is second, with 64.)

So is that a skill that's simply "not stat-able," to steal a term Kansas City manager Ned Yost tossed out there to reporters over the weekend? Well, let's just say it's more nuanced than that.

For one thing, publicly available defensive metrics just aren't as reliable as other numbers. So that at least makes it theoretically possible, Miller theorized, that "everybody is underrating how much the Royals' defense matters."

Except ... eh, wait a minute. They're constantly getting credit for their defense, "So nobody is ignoring it," Miller said. And if you look at metrics like defensive efficiency, he said, "The Royals don't actually turn a huge number of balls in play into outs."

So "underrated" is a word that doesn't seem to apply to this team's defense. What's more accurate, said Szymborski, is that projection models are "naturally conservative" in the way they factor in defense. Too much year-to-year fluctuation in the way we measure defense. Too much potential for injuries to wreak havoc on an overly optimistic projection.

In other words, the projections can't help but give slightly too little weight to all the different ways the Royals impact games with their defense. But is that why these projections are winding up being off by 12 to 23 wins in one year? No way. Our projectionists are pretty much convinced this is a relatively minor glitch. And that's all.


Think back to Game 5 of the World Series. Think back to Eric Hosmer's mad stampede for home plate, on a simple ground ball to third, to score the run that changed the World Series.

Think back to Lorenzo Cain's scintillating sprint around the bases nine days earlier. Think back to the seemingly impossible run he scored from first base -- on a single -- to turn Game 6 of the ALCS upside-down.

Those are freeze frames in baserunning brilliance that seem to sum up the reasons the Royals won the World Series. But this just in: The reasons the Royals won the World Series and the reasons they defy the projection machines during the season are two very different things.

"If we're talking about the Royals in October, all bets are off," Miller said. "The Royals have been way better in October than we would ever project any team, no matter how good, to be."

But the Royals of April through September are not a team whose baserunning exploits are being misread by the computers. If they were, Davenport said, "You would expect them to score more runs than expected, based on the number of singles, doubles and walks" they accumulate. Instead, they're actually scoring almost exactly as many runs as their expected runs calculations say they should. So whatever the computers are missing, it isn't their baserunning.

Well, what is it then? Let's take on some quick hits.


Szymborski has a theory that if there's one consistent reason the Royals beat the projections every year, "Dave Eiland may have something to do with it." After all, year after year, the Royals find pitchers (Edinson Volquez, Chris Young, Ryan Madson and others) who outperform their projections, right? There's only one problem with that theory: If there really is an Eiland Effect on everyone he touches, the projection model would pick up on it after one year, in theory. So that can't be all of it.


Moore talks proudly of the "environment" his front office and coaching staff worked to establish "from Day 1," which allows players to get the most out of their talents. That sound you just heard was the modern-metric fans out there cringing. But wait, Miller isn't totally ruling it out. "None of us acts like baseball isn't loaded to the gills with mystery that we can't measure," he said. So if that's why the Royals are outperforming their projections, he's "thrilled." He just wonders: Why is it affecting some players more than others? And "why aren't the Royals, as individuals, more consistently outperforming their projections?" If they were, you see, the computers would begin factoring that in too. So is there something there? Sure. But maybe we should really ask ...


Let's go back to Hosmer's mad dash home. We definitely underestimate how many elements went into that play. Scouting. Preparation. The ability of Hosmer to absorb information he was given and use it in the moment. But there's also something else, and that's the "latitude," Moore says, that Yost gives his players to use their instincts to make big plays in big spots. So even a sabermetrician such as Miller admits, "There's potentially something very significant to Ned Yost letting his players play and develop their own decision-making skills."

When you watch great players make one game-changing play after another in huge situations, you should never underestimate both the coaching and the culture that made that possible. But could any projection model capture that? How exactly? It's just "a very hard thing to measure," Davenport said.

There is more, naturally. We could delve into a dozen other areas, actually. But it's time to reveal the secret. Why is there no great overarching explanation for what the Royals have done and why the computers haven't seen it coming?

It's because, over the past two seasons, "The Royals didn't do the same thing twice," Miller said. "They did different things, once apiece."

In 2014, their individual players didn't put up numbers all that different from what the projections expected. But as a team, they found ways to win more games than they had any right winning, based on the metrics, because, well, it's baseball. And because human beings play it.

But in 2015, they vastly outperformed their projections as a team because their players vastly outperformed their individual projections -- because, well, see above.

If those players, as individuals, had just beaten the computer models year after year, the computers would have been smart enough to catch on to that. But the Royals are so sneaky about foiling those computers, they keep finding new ways to do it. And that's just not fair.

By the way, Szymborski said, you should know that these Royals are not unique. About a decade ago, the Chicago White Sox of the early 2000s were pulling the same tricks, particularly in 2005, when they were projected by ZIPS to win 79 games and wound up winning 99 games, followed by the World Series.

But what does that tell us? The same thing these Royals tell us: that no projection system will ever get everything right. And for the best reason possible. Because, as Davenport so beautifully put it, "That's part of being human."

"If I had the ability to make perfect projections, I promise you I wouldn't be applying that to baseball," Davenport said, laughing softly at the imperfection of humankind. "I'd be raking it in on Wall Street somewhere."