Guru's stat gives M's inside pitch
Numbers are like an old friend to baseball fans, always around to brighten up even the dullest of days on the field. But they can also be...
Seattle Times staff reporter
PEORIA, Ariz. — Numbers are like an old friend to baseball fans, always around to brighten up even the dullest of days on the field.
But they can also be a mistrusted enemy, deceiving about their true face. That's why it's so imperative for baseball teams to work with numbers that spell things out quickly and honestly.
The Mariners use such a statistic — Predicted Earned Run Average — to help evaluate pitchers. Better yet, they hired the guy who invented it.
Longtime baseball statistician and consultant Mat Olkin created the stat 15 years ago, but it's still a relative secret outside of baseball's hard-core numbers community. It's also not the only stat he has used since his 2005 hiring, though it provides fascinating insight.
"I'd see a guy who didn't get a good ERA but he still seemed to be getting hitters out," Olkin, 37, said by phone from his Connecticut, where his home office has floor-to-ceiling shelves overflowing with baseball books. "What Predicted ERA gives you is a truer indication of how effective a pitcher really was."
Olkin says the research behind Predicted ERA is mostly the product of baseball statistics guru Bill James. But what matters for Mariners fans is that Miguel Batista, Jeff Weaver, Horacio Ramirez and other new pitchers here this spring all got vetted by Olkin before being acquired.
While stats-based research makes some fans' eyes glaze over, this is no yawning matter.
The regular ERA stat is based on runs crossing the plate, often because of random factors a pitcher can't control. That makes it tough to accurately assess quality, a crucial need when determining whether that $10 million free-agent will see his ERA balloon from 3.00 to 5.00.
Predicted ERA provides something closer to crystal-ball insight.
It takes factors within a pitcher's control — his on-base and slugging percentages against — and multiplies them together. The sum is then multiplied by 31, a figure Olkin arrived at through calculations involving the number of hitters typically faced over nine innings and the ratio between earned runs and total runs allowed.
Put it together and you get something that looks like a pitcher's regular ERA, only higher or lower.
A Predicted ERA that is vastly higher than a pitcher's regular ERA is a warning sign of a future ERA spike. If it's much lower, a hidden gem may be at hand.
Mariners relief pitcher George Sherrill posted a high 5.21 ERA in 2005, but his Predicted ERA was an impressive 3.06. Sure enough, his regular ERA dropped to 4.28 last season. Better still, his Predicted ERA fell all the way to 2.66, a sign his regular ERA could drop again this year.
Sherrill admitted he'd never heard of Olkin's stat.
"Certain numbers can be skewed one way or the other," Sherrill agreed, adding that regular ERA doesn't accurately portray relievers because one bad inning can ruin a month of stats. "The only thing you can really look at is something like walks. I try to keep those as low as possible."
Predicted ERA takes walks into consideration through its use of on-base-percentage.
But M's starter Batista, also unaware of Olkin's stat, notes that on-base percentage won't tell the whole story when a skilled hurler selectively pitches around hitters. That leads to more walks, but pitchers with a lower slugging percentage against — the other main stat incorporated by Predicted ERA — will avoid the ensuing extra-base hits that lead to runs.
"Some pitchers will walk guys, but that doesn't mean the guy's going to score," Batista said.
Olkin says Predicted ERA is only one stat he uses.
"I'd rely on it, sure," Olkin said. "But by no means is it the only thing I'd rely on."
Nor is Olkin the final word.
"We use a variety of factors to evaluate players," Mariners general manager Bill Bavasi said. "Mat and the numbers he uses are one of those factors. But there are a lot of other things."
Geoff Baker: 206-464-8286 or firstname.lastname@example.org