Orioles vs. the stats: Why analytics are down on the Orioles

Orioles vs. the stats
(Jay Judge / Baltimore Sun Graphic)

SARASOTA, FLA. — As baseball embraces the numbers behind the game, 2016 appears to be another year in which the Orioles try to find success in the face of data that say it won't come.

The sounds of the Orioles' first glove pops and batting practice swings of spring training are often accompanied by the winds of denial, with players and fans alike rejecting projection systems and forecasts that don't match their own assessments.


Websites that host such projections, and thus hear the brunt of the backlash, say there's plenty of underlying data that support their low expectations for the Orioles, and having been wrong in the past doesn't mean that they'll be wrong again.

"The thing that we can kind of take comfort in is we know we're not biasing the projections in any way," said Dave Cameron, managing editor of FanGraphs, which has been down on the Orioles in recent years. "We're not anti-Orioles. We're not trying to say, 'How do we make Orioles fans upset today?' We have algorithms based on a decent amount of historical data, and it's generally proven to work in most cases. This is what those algorithms spit out."


FanGraphs and Baseball Prospectus are two of the foremost sites for public baseball analytics and projections, with each site sending members into Major League Baseball's front offices on both the scouting and data sides. Their sites have created and emphasized stats that even casual fans now use, and both use the extensive data history of the game to make annual projections.

Both sites take a player's history and the underlying statistics that they control, distribute those "events" throughout the course of the season based on playing time and project the number of games a team with that type of history would win in a given season.

"We're essentially just forecasting the number and type of events," Cameron said. "We think this team is going to hit X number of home runs, Y number of doubles. We're just kind of aggregating individual player projections and summing them up and trying, based on kind of proven mathematical models, if you have this many home runs and this many walks and give up this many home runs, it will generally translate into this many runs and runs allowed. Then you can turn that into an expected win total."

For the fifth straight season, comprising almost all of the Orioles' recent renaissance, they're projected to finish below .500. As of March 25, Baseball Prospectus' PECOTA (Player Empirical Comparison and Optimization Test Algorithm) system projected the Orioles to go an American League-worst 74-88, and FanGraphs had them at 79-83. But because of some recent seasons in which similar uninspiring projections were blown away, fans are reluctant to take the predictions seriously.

The Orioles themselves feel this way. Shortstop J.J. Hardy said the Orioles "know that [analysts'] opinion doesn't matter." Manager Buck Showalter said he doesn't read the predictions to begin with.

And the two sites hear frequently from Orioles fans who point out they don't believe what they're reading because of past success despite low expectations.

Specifically in the Orioles' two most recent playoff years, the forecasts didn't see them coming. From 2012 to 2014, no team outperformed its FanGraphs projections more than the Orioles. In 2012, when the Orioles broke a long playoff drought by finishing 93-69, PECOTA projected them to win 71 games. FanGraphs and PECOTA each had the 2013 Orioles, who finished 85-77, pegged to win 75 games. The 96-win Orioles of 2014 were forecast to go 75-87 by PECOTA and 76-86 by FanGraphs.

It was only last season that they were more in line, with PECOTA projecting 78 wins and FanGraphs at 79. The team had to scrape to get to 81-81. Sam Miller, editor-in-chief of Baseball Prospectus, said PECOTA has a 1-in-8 chance of underprojecting three years in a row, but that does not indicate a fourth straight year will follow. Last season was an indicator of that, he said.

"PECOTA's recent history of being 'low' on them was cited as proof that the system just can't handle their exceptional unprojectability," Miller wrote in an email. "Then, the Orioles won 81 games and weren't playoff contenders."

Miller and Cameron said the same few factors contribute to teams' overperforming, including good health, an above-average bullpen and good situational baseball — getting outs when teams are in scoring situations against them and driving in runs when they have the opportunities.

"The thing the Orioles were good at a couple of years ago, and they weren't good at last year, is essentially timing those events to be as fortuitous as possible," Cameron said. "They essentially sequenced their events in the best way possible. That's not something we can project — the timing of events is outside of the scope of our knowledge."

Many of these factors are cited as reasons that projections have recently come in so low for the Orioles and Kansas City Royals, who have reached two straight World Series and enter 2016 as reigning champions.


These factors can create big swings in actual wins, but there's hardly any consistency to them, and as a result they are difficult to predict. When it comes to forecasting how a team will fare based on expected runs scored and runs allowed, the Orioles struggle mightily in the latter. Analysts believe this team will hit, but Cameron cautions that teams that strike out substantially more than the league average have historically come in way under their projections.

FanGraphs projects the Orioles rotation to be the worst in baseball except for that of the Atlanta Braves, who are rebuilding.

"If you're bad pitching, especially guys that don't strike guys out a lot pitching in front of bad outfield defense, that can turn into a lot of runs," Cameron said. "The offense should be fine, but we think the rotation is going to be not good, and the bullpen is going to have to be amazing to carry them again."

Similarly, PECOTA projects the Orioles to allow the most runs in baseball, and projects that a rotation that replaces Wei-Yin Chen with Yovani Gallardo won't improve upon the American League's second-worst starters' ERA last season.

"The Orioles could overcome this in any number of ways — Kevin Gausman breaks out, Chris Tillman switches back to his 2014 self, some unexpected spring training invitee ends up shocking the sport, etc.," Miller wrote. "But that's what the Orioles are up against this year: a rotation that most likely isn't very good. They'll have to fix it or outslug their problems."

Both sites have had to combat the perception that there are blind spots in their systems, or perhaps that they're biased against certain teams, but overall don't waver much in their belief that they're right more often than they're wrong.

Miller said Baseball Prospectus knows "baseball is unfathomably unpredictable," and projections can both be "seen as hubristic" and "wildly inaccurate." Cameron said the exact numbers on their projected records typically hit within five games in either direction.


"These aren't intended to be prophecy," Miller said. "They're simply a way of organizing information in a way that we think is useful to a lot of people, primarily at the individual player level."



Recommended on Baltimore Sun