The Orioles’ Secret Sauce

Last week, both Jeff and I wrote about the Orioles and BaseRuns. Jeff said this towards the end of his piece:

In four of the last five years, the Orioles’ BaseRuns record has been better than the projected record by at least six wins. In the fifth year, they were the same. The point being, the Orioles have knocked their projections out of the park, and they’ve done it far more than anyone else.

I ended my article with the following quote:

Haven’t [the Orioles] overperformed their BaseRuns wins for many years now? Yes, they have. But they’re overperforming at run prevention, not run scoring.

That night, Buck Showalter thumbed his nose at us both. Specifically, he mocked and derided the concept of run prevention by refusing to use his best run preventer in a tied elimination game with one out and runners on the corners. That refusal hurt the team’s chances of winning in a high-profile way. And thus another Orioles team bit the dust.

Given what happened, the prospect of talking about the Orioles and run prevention makes me twitch. But I’ll suppress it because there’s some interesting analysis here. Onward!

Before I jump into run prevention, I want to clarify what the Orioles are doing:

orioles-baseruns-overperformance-1

As Jeff said, the team’s overperformed its BaseRuns for seven years now. And as I said, it can’t be the offense. The Orioles rarely overperform their batting BaseRuns:

orioles-batting-baseruns-overperformance-1

The secret lies in run prevention. The Orioles consistently allow fewer runs than BaseRuns predicts:

orioles-pitching-runs-overperformance

Note the sharp drop, indicating a large overperformance, in 2014 when the team won the AL East.

Based on these graphs I’d guess that, when it comes to BaseRuns overperformance, run prevention matters more than run scoring. Luckily I don’t have to guess; I have data. The following table supports this claim:

Effects of Run Prevention vs. Run Scoring on BaseRuns Overperformance
Area of Performance Probability of Overperforming BaseRuns Given Worse-than-Average
Performance (%)
Probability of Overperforming BaseRuns Given Better-than-Average
Performance (%)
Increase in Probability of Overperforming in BaseRuns (percentage points)
Run Prevention 27.8 63.5 35.7
Run Scoring 34.2 55.2 21.0
SOURCE: 2002-2016 Team Data

The logic: if a team moves its pitching BaseRuns from underperforming to overperforming, the chance they’ll overperform their BaseRuns wins improves by 35.7 points. Teams that underperform their pitching BaseRuns have just a 27.8% of winning more games than BaseRuns predicts. But teams that overperform their pitching BaseRuns have a 63.5% chance of overperforming their BaseRuns.

However, if a team moves its batting BaseRuns from underperforming to overperforming, the chance they’ll overperform their BaseRuns wins improves by only 21 points. That’s still good, but less good.

Effects of Run-Prevention Metrics on BaseRuns

Which aspects of run prevention matter the most? The following table shows how improving the major run-prevention metrics, each indexed to league average, affects a team’s chances of overperforming its BaseRuns:

Effects of Run-Prevention Metrics on BaseRuns Overperformance
Metric Probability of Overperforming BaseRuns Given Worse-than-Average* Performance (%) Probability of Overperforming BaseRuns Given Better-than-Average* Performance (%) Increase in Probability of Overperforming in BaseRuns (percentage points)
K+ 43.3 46.9 3.6
BB+ 37.6 53.4 14.8
BABIP+ 42.2 45.7 3.5
HR/FB+ 42.1 47.6 5.5
GB/FB+ 48.3 41.4 -6.9
LOB+ 28.8 62.0 33.2
SOURCE: 2002-2016 Team Data
*For K+ and GB/FB+, “worse than average” means values below 100. For the other metrics, values above 100 are worse than average.

Relative strand rate (LOB+) has the largest effect on whether a team will overperform its BaseRuns. Moving it from below average to above average boosts a team’s chances of overperforming its BaseRuns by a whopping 33.2 points. That’s nearly 10 times the improvement you’ll get from improving strikeouts.

Unfortunately, this knowledge has limited practical applications. For starting pitchers, high strand rates typically regress to the player’s true-talent mean the following year. And players’ true-talent means aren’t dramatically different. Sure, a lots of strikeouts and a great defense will help. And in rare cases a great pitcher can consistently strand runners. But don’t count on it.

The Orioles are doing what they can. Beginning in 2012, they’ve been above average at stranding runners:

orioles-relative-strand-rate

Have the Orioles found a mix of pitching and defense that somehow allows them to strand more runners than their competition? The graph above says they have, for five years running. Note the spikes in 2012 and 2014, both years that the Orioles reached the playoffs. No spike occurred in 2016, though.

The next-most important pitching factor is walk rate. This isn’t as impactful as strand rate, but pitchers also have much more control over walk rate. General managers can select for pitchers with more control as well as catchers who frame strikes better than others. Pitchers can also learn control more easily than they can learn, say, how to strand runners.

You wouldn’t know it given the recent performances of Ubaldo Jimenez and Yovani Gallardo, but the Orioles have tried to take this lesson to heart:

orioles-walk-rate

The team’s walk rate reached its nadir in 2012, when they surprised everyone with a 93-win season. Coincidence? I don’t think so. That year they were better than average in both LOB+ and BB+. Teams that surpass both marks have a 65.4% chance of overperforming their BaseRuns wins.

Effects of Batting Metrics on BaseRuns

To make sure I’m not missing anything, I analyzed a few batting metrics in the same way:

Effects of Run-Scoring Metrics on BaseRuns Overperformance
Metric Probability of Overperforming BaseRuns Given Worse-than-Average*
Performance (%)
Probability of Overperforming BaseRuns Given Better-than-Average*
Performance (%)
Increase in Probability of Overperforming in BaseRuns (percentage points)
K+ 17.6 22.9 5.3
BB+ 20.3 20.5 0.2
ISO+ 21.0 20.0 -1.0
BABIP+ 23.8 18.4 -5.4
SOURCE: 2002-2016 Team Data
*For K+, “worse than average” means values above 100. For the other metrics, values below 100 are worse than average.

The first line provides more evidence that offensive strikeout rate affects BaseRuns, if only by a small amount. Becoming a below-average strikeout team boosts your chances of overperforming your BaseRuns by 5.4 percentage points. Curiously, increasing offensive power (ISO) or BABIP decreases your chances of overperforming your BaseRuns. This logic doesn’t make sense to me, but it’s what the data says. I’m open to explanations.

More importantly: strand rate and pitchers’ walk rate are still the most impactful metrics by a long shot. Wait, let me amend that. They’re the most context-neutral metrics. Perhaps in a future article I’ll analyze the relevance of context-sensitive metrics. I also acknowledge I haven’t examined the effects of defense on BaseRuns. Defense affects strand rate. But this article is too long already. I’d also like to explore the effects of reliever run prevention vs. starter run prevention.

We learned this:

screen-shot-2016-10-12-at-7-29-20-am

Now we know a little more about the Orioles’ secret sauce. They strand runners better than most and, at least before they signed some high-risk pitchers, limit bases on balls. This formula doesn’t work for every team, but the Orioles have taken advantage of it for seven years now.

Well, for seven years minus one Wild Card game, at least.

We hoped you liked reading The Orioles’ Secret Sauce by Ryan Pollack!

Please support FanGraphs by becoming a member. We publish thousands of articles a year, host multiple podcasts, and have an ever growing database of baseball stats.

FanGraphs does not have a paywall. With your membership, we can continue to offer the content you've come to rely on and add to our unique baseball coverage.

Support FanGraphs




Ryan enjoys characterizing that elusive line between luck and skill in baseball. For more, subscribe to his articles and follow him on Twitter.

newest oldest most voted
Tct
Member
Tct

Doesnt the fact that increasing your ISO or BABIP decreases the probability you will over perform just suggest that base runs overvalues those things? I couldn’t tell you how baseruns is calculated if my life depended on it, so take it easy if I am wrong. But it seems like this data is just as interesting, if not more so, as a way to see how base runs is flawed rather than how the Orioles keep out performing it.