The Astros’ Sign-Stealing Left a Fingerprint in the Audio Data

Houston Astros equipment lays on the ground at Globe Life Park on July 11.
Houston Astros equipment sits on the ground at Globe Life Park in Arlington, Texas, on July 11. Brandon Wade/Getty Images

This article originally appeared in Baseball Prospectus.

This week, we learned that the Houston Astros used a crude system of monitors and audible signals to steal signs and communicated them to hitters at the plate. A camera in the outfield relayed a video feed to players in the tunnel, who whacked a trash can lid in order to relay what pitch was coming. In addition to being blatantly illegal—despite apologists’ counterarguments, the rules are very clear that no video technology can be used in real time—the Astros’ method left a plethora of evidence behind. The abundant traces of cheating allow us to go back and analyze when and how effectively they cheated.

The slamming sound of the trash can is so clear and distinct that it’s easily visible in the audio data. Here’s what a sample crash looks like in terms of volume.

Astros sound wave
Rob Arthur

The particular audio sample comes from a September 2017 game, which you can listen to below by fast-forwarding to the 2:58:30 mark.

Note the background noise level from immediately before and the two peaks, corresponding to two whacks on the lid: the signal for a breaking ball.

The slamming sound was so loud that it made a significant difference in the noise levels preceding almost all of the breaking balls the White Sox threw in that September game. (That is, until Danny Farquhar noticed the sound and switched up his signs.) Averaging over all the times the Astros were at bat and about to receive a non-fastball, the disparity in noise is clear.

Graph showing the Astros and White Sox fastball batting and noise levels preceding pitch
Rob Arthur

When the Astros were at the plate and a non-fastball was on its way, the trash can bangs upped the noise level significantly compared with when it was going to be a heater. When the White Sox were batting, there was a much smaller (10 times) difference in noise between their fastball and non-fastball pitches.

Depending on how you compute statistical significance, some methods will show this pattern to be significant and others won’t. But it’s kind of a moot point because we already know that the Astros were spanking a trash can lid to inform their hitters about the incoming pitch. What is more stunning here isn’t the statistical magnitude but rather just how obvious their metal whacking was—loud enough to leave a signature in the data.

With a clear audio fingerprint in hand, it’s possible to go back and review games from the Astros’ 2017 season and tell beyond the shadow of a doubt when and how they were sign-stealing. It’s easy to be fooled by drums in the stadium or lose focus thanks to announcers talking, but with the data to fall back on, it was plain when the Astros had broken their opponents’ codes.

In reviewing a handful of late-season Houston games, a couple of things became obvious. One was that the sign-stealing was almost immediate. The earliest I saw metal banging in the data was on the third pitch of a game. It was always within the first five or 10 pitches. Whoever did the actual signal-decoding for the Astros was very aggressive about making a call as soon as they thought they had cracked the opposing catcher’s cipher.

Second, that aggression seems to have been justified. In examining a few dozen throws, I didn’t find a single instance of an erroneous trash can bang predicting a breaking ball when the pitcher was actually tossing heat. It was always spot on. Whereas you can imagine an inaccurate predictor actually doing more harm than good, the system’s perfection would seem to be a significant advantage for a hitter.

Thirdly, the jury is still out on whether the Astros put their system to work in the 2017 playoffs. Houston claimed that its sign-stealing methodology wasn’t in use once October began. But an examination of several playoff games shows some telltale bumps in the audio, albeit without the clear connection to pitch calls that existed in the regular season. Below is an example from Oct. 14, 2017, in Game 2 of the American League Championship Series. (Fast forward to the 19:55 mark. Ironically, Joe Buck shouts out his audio crew while exclaiming, “You are going to hear good sound” right before the noise.)

It’s a bit harder to see this one in terms of raw volume because Buck is talking while the trash can clatters. To get a good view of it, you have to use what’s called a spectrogram, which looks like this:

Astros spectogram
Rob Arthur

In this kind of chart, time goes from left to right and frequencies from bottom to top, and the intensity at each frequency is reflected by color. Buck’s voice continues throughout the clip at the lower frequencies, but the bangs occupy the upper register, appearing as two distinct bands spaced at almost the same interval as the clip from earlier in this article.

But even though the trash can sound appears obvious in the clip above, the banging seems to come and go in that game and in other postseason contests. Unlike the regular season, where the code was as simple as “noise equals breaking ball,” the same sound doesn’t seem to always foretell an off-speed pitch.

Does that mean they really did cast aside their sign-stealing in a fit of good sportsmanship? There are reports the Astros switched to a different kind of signaling in the World Series, using either a visual cue (someone in the bullpen raising their arms) or a different sound (perhaps whistling). It seems likely that any one of these communication pathways would have had some problems. Whistling could have been lost in the roar of the crowd, but trash can smacking and hand signals from the outfield could have been dangerously obvious. One way to have overcome these issues would be if the Astros had adopted a hybrid system in October, using several different methods to convey signals. Perhaps they realized that the trash can was overly bold when Farquhar managed to notice their cheating within a few pitches.

Or maybe they really did abandon their cheating when it counted most, as they claim. It would seem out of character for an organization willing to break the rules in late September, long after they had effectively wrapped up a playoff berth, to then do the honorable thing and retire the system that helped get them there. And that’s not to mention the potentially severe handicap that could incur on hitters who had gotten used to knowing what pitch was coming. Considering the Houston organization’s other adventures in immorality, it’s hard to believe that they’d forgo a clear advantage when the stakes were highest.

Consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.

More from Baseball Prospectus:

The Visible Hand of the Baseball Economy

Moonshot: Rising Ticket Prices Explain MLB’s Attendance Problem