Edge%: It’s Baaaack

A while back, Jeff Zimmerman and I introduced the concept of Edge% — a metric that attempted to quantify the extent to which a pitcher worked the edges of the strike zone. Jeff initially looked at how this applied to Tim Lincecum and how his performance depended to some extent on his ability to pitch to the edges of the plate. I followed up with a high-level piece that compared the performance of pitchers at an aggregate level depending on how extreme their Edge% was in a given season.

While the findings were interesting, they were also a little inconsistent. That’s because Jeff and I independently created two distinct metrics. We decided to combine our efforts (as we have been known to do) and settle on a single, consistent formula.

And that’s the focus of this article.

As before, the strike zone was adjusted based on the handedness of the batter. This time, however, we also adjusted based on height (see here for Mike Fast’s original breakdown):

Edge for right-handed batters:

px > -1.03 AND px < -.43 OR px >.7 AND px <1.00 

pz > (.92 + batter_height *.136) AND pz < (2.60 + batter_height *.136)

Edge for left-handed batters:

px > -1.20 AND  px < -.9 OR px > .21 and px < .81

pz > (.35 + batter_.height *.229) AND pz < (2.0 + batter_height*.229)

Since 2007, for pitchers with at least 180 IP in a given season, here are some percentiles for the new Edge%:

Percentile Edge%
90th 19.8%
75th 18.8%
50th 17.7%
25th 16.5%

Not a huge change from the original formula, but a few differences by percentile. Where we do find some differences is in the relationship between Edge% and individual pitching metrics — both outcome and process metrics. More on that below.

In terms of the repeatability of Edge%, pitchers with >= 180 IP in a season experienced a year-to-year correlation in their Edge% of .59. For context, that puts this version on par with FIP-, K/BB, and tRA.

When we look at the average performance of pitchers in each percentile and compare them to each other, a few interesting patterns emerge*:

And, again, here are some basic correlations:

The biggest difference we see with the new formula is that the relationship between Edge% and K% is now negative. This makes sense since we’ve now restricted the vertical edges of the metric, somewhat biasing the data against pitches that generate strike outs more based on vertical chasing.

What’s more interesting is that there appears to be a somewhat “non-linear” relationship between Edge% and many metrics. So, for example, it isn’t simply that it’s better to have an extremely high versus an extremely low Edge%. If we look at K%, the 90th percentile has about a 1% advantage over the second quartile (50th to 75th percentile). However, they are half a percent worse relative to the third quartile (25th to 50th percentile). Now, K% appears to be one of the (very) few metrics where the most extreme Edge% pitchers are worse than that third quartile, but for many of the metrics the degree to which they are better dips lower for that same group.

Walk rates are still better for high Edge% pitchers — and the relationship is now even stronger (correlation of -.34 compared to -.11), and generally speaking high Edge% pitchers experience a lower adjusted ERA and FIP (again, the difference between high and low Edge% and these metrics grew in size).

This is just the beginning of a series of articles that Jeff and I will have in the coming weeks on the topic. A big area of focus will likely be the relationship between Edge% and Heart%, as it likely isn’t as cut and dry as higher Edge% = better pitcher. Certainly there is a lot to tease out, so stay tuned for more on this topic from us both.


*The percentile averages were not weighted by innings pitched. It might impact the specific numbers, but the innings were restricted to a minimum of 180, so I don’t think it will change things all that materially.

Bill leads Predictive Modeling and Data Science consulting at Gallup. In his free time, he writes for The Hardball Times, speaks about baseball research and analytics, has consulted for a Major League Baseball team, and has appeared on MLB Network's Clubhouse Confidential as well as several MLB-produced documentaries. He is also the creator of the baseballr package for the R programming language. Along with Jeff Zimmerman, he won the 2013 SABR Analytics Research Award for Contemporary Analysis. Follow him on Twitter @BillPetti.

Newest Most Voted
Inline Feedbacks
View all comments
9 years ago

Cool article. It would be good to combine Edge% with a measure of velocity which might explain some of the “non-linear” relationships (i.e higher velocity pitchers have a lower Edge%).