What Happens When Madison Bumgarner is Really On?

On 7/18, Madison Bumgarner changed his positioning on the mound. Eno noted it and Madison Bumgarner confirmed it. Bumgarner has worked to make his pitches (and I presume his release point) very similar through video and in front of a mirror “making sure he sets up in in the right places.”

Eno summed it up: “Bumgarner is ready to make the most of his old playbook. Throw lots of fastballs, cutters, and curves, all from the same release point, all with similar spin, and all exploding out of a slow, deliberate delivery.”

Bumgarner has done just that confirmed by RMSE (Root-Mean-Square-Error/Deviations) on release point consistency. The RMSE represents the standard deviation of the differences between each pitch’s release point (observed) and the mean, within-game release point (model/estimator): the lower the value, the better. I am considering this a skill/benefit for two main reasons: general command and deception. It’s hard to pickup different pitches form an extensive repertoire if they are released from the same point.

First, a real life example: my adult baseball league. It is real baseball (not softball). Most pitchers through between 65 and 80 MPH so nothing too intense, but not easy either. I did strike out 6 times in 29 AB for the season. Of the 6 strike outs, only one was a swinging-strikeout – the rest were called strikes from umpires who basically wanted to get the games over with. I had little trouble making contact as most breaking balls were released at completely different arm slots. It was easy to distinguish. However, I actually struck out three times in one game and the guys wasn’t even hitting 70 MPH. He had 3-4 different pitches, and it all released from what looked like the same point. I don’t want to just toot my own horn. I only batted .241 based on a .000 groundball BABIP (I’m a chubby 30 year old: think a Caucasian, bearded Pablo Sandoval). I even got thrown out by the right fielder once!

So that’s the quick real-life verification. Onto the actual value…

Release point consistency will have varying levels of importance/effect for different pitchers. Think of it like MPG for a car: the science is there, but actual MPG will vary based on the driver. I drive like a New York manic still. My wife drives like a Minnesotan. Either way, we have a Prius, and the science is there.

In addition, a sundry of pitchers purposely work from different slots as a different source of deception, but I will attempt to verify my opinion on the subject in this post and then on a grander scale down the road. You can check out the 2010 leaders in release point consistency for fastballs within games in Mike Fast’s BP post. For reference, to compare with Bumgarner’s RMSE’s below: R.A. Dickey led the league with 1.3 inches (change). Trevor Cahill was on the other side of the spectrum at 4.5 inches. One exclusion from the list based on multiple distinct release points was Bronson Arroyo.

Mike Fast referenced that it might be fruitful to look at release point consistency between pitch types. Here is some Madison Bumgarner… fruitfullness? Fruitfillery?

I pulled three CSV files from Baseball Savant:

• All 2014 regular season Madison Bumgarner pitches up until and including 7/13 (prior to his significant mound move that Eno highlighted)
• All Madison Bumgarner pitches after and including 7/18 (mound move)
• The three 2014 Madison Bumgarner playoff games

Here are the RMSE results:

Prior to 7/18, Bumgarner’s general RMSE (pitch-to-pitch for the season/not from game to game), was 2.91. If we treat each game as the same number of pitches, his average within-game RMSE was 1.80. He had two games that approached 3.0.

As you can see from the gray line drop-off above, from 7/18-onward, Bumgarner’s general RMSE was 1.85 – impressively better. His average within-game RMSE was 1.70. The worst RMSE was a 2.26.

In the playoffs so far, he has an overall 1.58 RMSE and a 1.56 average. His within-game RMSE’s are 1.38 (10/6), 1.59 (10/1), and 1.7 (10/11). Appropriately, the outcomes have been impressive.

Madison has become more and more deceptive through release point consistency. In the following content, when I say “on,” I mean consistent (better than his RMSE i.e. a smaller value). When i say “off,” I mean less consistent (worse than his RMSE average i.e. a larger value).

Let’s look at the outcomes when better, around or worse than his average release point consistency prior to and from 7/18: 

7/18 and onward is the left cluster. 7/13 and prior is on the right. Within each cluster, you will find the better than average release consistency (signaled by “<“), near-average release consistency (~) and worse than average release consistency (>) in that order from left to right.

Attending to the left cluster (7/18 and onward), if we treat each game as an equal number of total pitches, since 7/18, Bumgy had a 16.1% swinging-strike rate (71.7% contact-rate) and a 1.07 ERA (1.70 FIP) when he was on. That compares to a 8.2% swinging-strike rate (83.8% contact rate) and a 4.81 ERA (5.47 FIP) when he was off. From a swinging-strike perspective, he was better than Clayton Kershaw when he was on; he was basically Kyle Lohse when he was off.

*According to the above, perhaps even more interesting, it seems that release point consistency became an increasingly important part of Bumgarner’s success. Prior to 7/18, when he was on, his swinging-strike rate was 11.1% compared to 10.3% when he was off. That’s barely distinguishable. His ERA (2.82 vs. 3.64) and FIP (2.82 vs. 5.17) was still better, but again less distinguishable (2.35 ERA differential) than the 7/18-onward cluster (3.74 ERA differential).

Further verifying that his release point consistency became more important, look at these correlations (despite the uber-small samples): 

Depicted, are the general correlations to RMSE for the season in gray; the correlations up to and including 7/13 in orange; and the correlations for 7/18 and after in blue.

From 7/18-onward, Bumgarner’s RMSE had a significant correlation to swinging-strike (.725), contact (.673), and K/9 (.876) rates in the right directions – same goes for FIP (.731) and even ERA (.436).

Per his conversation with Eno, he works on his deception. RMSE clearly confirms the glorious effect.

After 7/18, when he was at his average release point consistency, he was good old Madison Bumgarner. When he was off, he was basically Franklin Morales (5.4 FIP). Don’t get me wrong: he can remain effective when he’s off based on his per-pitch effect (four pitches with a swinging-strike rate greater than 20%), but release point consistency is probably a big part of that, and when he is on, he is elite.