Recalling Jim Johnson

On April 12, the Orioles made arguably their most productive move of the season, regardless of whether or not they had any idea it would work out that way. They optioned infielder Scott Moore to the minors and recalled 24-yr old pitcher Jim Johnson. Johnson, a starter in the minor leagues, had made two “audition” starts for the Orioles in 2006 and 2007, racking up the undesirable line of 5 IP, 12 H, 10 ER, 5 BB, 1 K. His minor league numbers suggested he was better than that and, since making his 2008 debut on April 13, he has been a key component on an its-surprising-they-aren’t-thirty-games-under-.500 team.

Johnson also stakes claim as the highest-ranked reliever with both a 1.85 WPA/LI and 2.15 REW. Relative to context-neutral wins and wins based on shifts in run expectancy, there has not been a more productive relief pitcher to this date. His overall numbers this year: 43 G, 57.1 IP, 34 H, 24 BB, 30 K. He has also surrendered 12 earned runs on the year, four of which came in one outing in the last couple of weeks.

His 3.30 FIP suggests an ERA of 1.88 has not necessarily been born out of his controllable skillset and, amongst relievers with at least 40.0 IP, his ERA-FIP differential ranks tenth in baseball. The higher FIP is also a direct result of his low strikeout rates and relatively high walk rates. While giving out 3.77 free passes every nine innings won’t lose a reliever his job, especially when comparing him to his peers, the 4.71 K/9 is, at least right now, a bit of a red flag. Looking at the same group of relievers with forty or more innings tossed, his K/9 is the fifth lowest, as is his K/BB. Due to only 34 hits allowed, though, his WHIP is currently a tremendous 1.01.

The question should then become, well, if he doesn’t strike anyone out and his walk rate isn’t that impressive, how has he managed to produce an extremely respectable 3.30 FIP? The answer: he has not given up a home run yet this year. In 43 games, not one ball has left the yard in fair territory after sailing from his hand to the batter’s box. In fact, he is the only reliever with forty or more innings not to serve up a gopher ball this year.

In the minors he had decent but not overwhelming numbers, but his 93-94 mph fastball and 78 mph curveball combination has definitely more than gotten the job done to this point in his first full year in the big leagues. He has only given up, as mentioned, 34 hits this year but his BABIP is a ridiculously low .205. Either Jim Johnson is going to emerge as the elitest of relievers or this is going to regress from here on out. His strand rate of 78% is high but nowhere near the likes of Joe Nathan and his companions atop the leaderboard.

He isn’t allowing a ton of baserunners primarily because balls put in play haven’t been falling in for hits as much. If/when that regresses we can expect the BABIP and WHIP to rise. Though his strand rate isn’t ridiculously high, it is well above average and, if sustainable, will help prevent some of these “new” baserunners from scoring. Regardless, even if or when his numbers do worsen, it won’t mask how effective he has been this year or make us forget that at the end of July he was one of baseball’s most effective relievers. It will take another year or two to know his true talent level but that does not take anything away from his productivity.



Print This Post



Eric is an accountant and statistical analyst from Philadelphia. He also covers the Phillies at Phillies Nation and can be found here on Twitter.


Sort by:   newest | oldest | most voted
Rick
Guest
Rick
8 years 1 month ago

Given the inherent variability of certain statistics, I wonder if it’s possible to construct a confidence interval around FIP.

For example, we don’t expect Johnson to continue to not allow a HR all year. And once he does allow a HR or three, that FIP will increase. Yet, as he accrues IP, that HR rate increasingly represents a skill instead of “luck”.

So how can we account for that, given the small sample size for relievers and the inherent variability of HR rate in particular. Do we look at a Marcel projected FIP?

David Appelman
Guest
8 years 1 month ago

I would argue that we do have some idea of how many home runs he should give up.

-He is an extreme groundball pitcher and induces them at about a 58% pace. His LD% is a little low, so for the sake of argument lets say he’s actually a 55% GB pitcher and a 30% FB pitcher.

-Your average reliever allows about 9.5% of his fly balls to become home runs.

-He’s allowed 164 balls in play (not including bunts) and if he allows 35% of them to be fly balls, and on average 9.5% of them become home runs, he should have given up between 4 and 5 home runs this season.

Even if there’s something about his pitching style which for some reason allows him to have a lower than average HR/FB rate, he should still have given up between 2 and 4 home runs.

David Appelman
Guest
8 years 1 month ago

I think 50+ innings worth of data is enough for GB/FB rates for a pitcher. They tend to stabilize quickly. The rest is based off league averages and you can even give him the benefit of the doubt if you want.

The actual % of balls in play he allows overall you might have more trouble with, i.e. are his k9/bb9 really 4.71/3.77. But I don’t think they’re that far off what they should be considering his minor league numbers, and those were as a starter.

I think it’s quite reasonable to say he should finish the season with an additional 2-4 home runs, but he certainly wouldn’t be the first pitcher to defy the odds and finish up a season without allowing a home run.

Rick
Guest
Rick
8 years 1 month ago

I guess what I’m getting at is that I often find FIP virtually useless because the nature of HR allowance is so different than that of BB and SO. It takes so much longer for HR rates to stabilize – to become predictive.

I suppose this issue is addressed using xFIP (Johnson’s is 4.34). However, it doesn’t strike me right to standardize HR rate as that’s essentially the same as rewriting the formula with just SO and BB.

So I guess what I’m asking is if we can come up with a version of FIP that is a better predictor of itself by adjusting the HR rate. Would using GB/FB instead be more robust? Though perhaps at that point, you’ve violated the simplicity of FIP and might as well craft a new metric entirely.

Scappy
Guest
Scappy
8 years 1 month ago

I think what Rick may be onto would be something like xERA. Use the frequency of GB/FB/LD with SO and BB to determine how many runs a pitcher should have given up. How one would do that is beyond me.

Rick
Guest
Rick
8 years 1 month ago

An aside, is there a decent way to quantify the amount of variability? That is, whether we’re talking ERA, FIP or what have you, can we also report a confidence interval intended to capture true skill. Perhaps this is just a function of IP, but it would be nice to see. At 50 IP, your 95% CI as +/- 2.00; at 100 it’s +/- 1.00, etc.

If the purpose of a stat like FIP is to attempt to convey performance rather than result, it would seem logical to convey the amount of confidence we have that the number being reported is an accurate measure. I suppose we all do this intuitively based on IP, but I’m curious what it looks like in the reported units.

David Appelman
Guest
8 years 1 month ago

Rick, xFIP I believe replaces the actual HR a player has given up with “0.105 * Total Fly Balls”, so I think it’s more or less what you’d be looking for?

Eric, I think we’re definitely on the same page that his numbers are more or less, out of whack, and due for some correction. ;)

Rick
Guest
Rick
8 years 1 month ago

Thanks David. I was under the mistaken notion that xFIP used a league average HR/9 figure with no player specific modifier. That makes sense.

Still interested in the confidence interval concept, though that applies to any range of predictive stats.

wpDiscuz