“In scenarios where the next batter was intentionally walked, I went to the first pitch thrown to the subsequent batter after the free pass.”
I do not quite understand the logic behind the practice of doing this for your analysis. The pitcher is in a 3-2 count, the coach visits the mound, the pitcher throws a ball to walk the batter which this pitch isn’t included in your analysis, and then the first pitch to the very next batter is included in your analysis. Isn’t there a huge situational difference between that ball four to the previous batter and the first pitch to the next batter? Or am I not understanding correctly?
I would love to see this analysis with that ball four included rather than the first pitch of the next batter.
Never mind, I glossed over the word intentional for some reason, but I still think the situation is different with the first pitch to the next batter is included. In my opinion, it should not count in the analysis.
You misunderstood me, or maybe I did not explain that well. In that situation, I added the ball thrown on 3-2. On the occasions where there were intentional walks, the coach came out to the mound and ordered the pitcher to walk the next batter. Presumably, the coach came to the pitcher to talk about walking the batter and how to attack the next one, so I included the first pitch to the batter after the intentional walk.
I understand now. You are basing your analysis on the implication made by Matthews that the coach will always instruct the pitcher to throw a fastball for the next meaningful pitch (i.e. the intentional walk not being meaningful) regardless of batter or situation. Thanks for clarifying.
I have a suggestion to control for the adverse selection problem you mentioned with respect to comparing to league average pitch frequencies (i.t. that “after visit” sample is more heavily weighted towards pitchers that require more visits than the league average sample). Instead of computing an unweighted league average, compute a weighted average where individual pitchers in the sample are weighted by the total number of mound visits over the period examined.
Gary Matthews was wrong. It would be big news if Matthews were ever right about anything.
I guessed that the pitch percentages would be about the same as usual, but I don’t have Matthews’ immense information base.
Nevertheless, this was an interesting and informative study.
The only actual way I was able to do this was by hand, so going back and finding out which pitches were thrown before the mound visit would be very difficult to do now. If there were an easier way to control for mound visits, this would definitely be a good way to attack the issue.
I really enjoyed reading this. If you happen to watch/listen to as many Phils broadcasts as I do, you’ll hear a whole lot of nonsense coming from Good Ol’ Sarge that would’ve been FireJoeMorgan-worthy just a few years back. That he was somehow correct about Rothschild is, I can only assume, the result of dumb luck. Thanks for posting – looking forward to the next one.
As a hitter “sarge” is about right, expect a FB 45% of the time after a mound visit. This may improve his odds of putting the ball in play successfully.
Maybe we need to be more critical of the coaches advice or when they chose to visit the mound
I’m not sure that “A pitching coach most often comes to the mound when a pitcher is struggling” is a true statement. Stalling for a reliever to get in a few more warm-up pitches and delivering pitch around (unintentional intentional walks) or defensive positioning instructions could easily account for half more of the pitching coach visits, especially later in the game.
Before even delving into the analysis of how pitch selection changes after a pitching coach visit, I’d start with a simple table showing when the pitching coach visits were made broken down by game status (run differential/inning/outs/count/bases occupied). You could then eliminate the bulk of the cases of stalling visits by selecting only those visits made in the first 4 innings of a game or so. I also suspect that very few stalling visits are made with the bases unoccupied, a score difference greater than 5, etc. That table mentioned above would make for an interesting post on its own – when do pitching coaches make mound visits?
Nice article, though I wish you presented the statistics in a normalized fashion. As in, for each pitch reference a database with that individual pitcher’s averages, and construct the overall number of each pitch that should have theoretically been thrown. One way to get around the question of biasing the sample with types of players who are more susceptible to mound visits.
There are a variety of situations in which the coach or manager will come to the mound, complicating drawing any statistical inferences from the next pitch thrown. Often, the visit will be made to a struggling pitcher, to calm him down, and not to call the next pitch. On the other hand, the pitcher may not be particularly struggling, and the coach will come out to consult about the strategic decision as to what pitch to throw next. You’d think the fastball would result more from the “calm the pitcher down” visits, and less from the “subtle strategic decision” visits. I’ve even seen for example, Jim Leyland come out as a feint, making the batter think about subtle strategy when the next pitch will be the obvious fastball that the pitcher planned on throwing anyway in this situation.
Great article! I’ve heard other tv commentators say the same thing. More so back in the 80s, compared to now. I wonder if the fb has just been replaced by the sinker to induce a double play like you said.
Great article. I would love to see some analysis on how effective those visits to the mound are. (My own data crunching abilities are non-existent). So often, it seems that the mound visit only gets the pitcher more rattled than he already is. Also, some coaches, I imagine, have a more beneficial effect than others …
Every time I hear the phrase “mound visit” I can’t help but think of cunnilingus.
Comment by Justin Bailey — March 9, 2012 @ 7:54 pm
This is different at lower levels of play. As a high school pitcher myself, I have never thrown anything but a fb after a vist. Not only that, but when I am hitting I know that after a visit I’ll get a fastball. Not once have I gotten anything else. The reason MLB is different is because those pitchers are so good most of them can throw changes and sliders for strikes just as easily as a fast ball. If you don’t have confidence in your off speed stuff, you’ll be in a or aa instead of the majors.
A fair percentage of the time, the mound visit is meant to give the next guy to come in the ballgame a few more minutes to warm up. They’re probably talking about tee times the next day and stuff like that every once in awhile too.
I believe Sarge said some pitching coaches demanded fastballs, and others had other pet pitches they’d call for, so who knows how right he was?
Comment by EastFallowfield — March 10, 2012 @ 11:28 am
FYI: in the bottom of the 4th of today’s Cubs @ Dodgers game, and Len Kasper cited this article as a part of an extended discussion on pitch tendencies and data analysis.
Comment by Matthew Bultitude — March 11, 2012 @ 4:45 pm
I think there could be more to this. Such as whether the mound visit followed a hit or walk. What was the last pitch thrown and is it different than the first pitch thrown after the visit.
Whether the first pitch to the batter is different than the 1st pitches the same batter has seen during the game.
Sometimes coaches go to the mound just to calm a guy down or talk about the situation. But sometimes they go with a specific bit of info like “This guy hasn’t swung at a breaking ball yet, so let’s start him off with one and get ahead” or “we need this guy to pull the ball on the ground so let’s pitch him soft away and get him to roll over on one” or “we’re playing him to pull, so pitch to your defense and stay inside” and things like that.
We might see different data based on quality of hitter, location of runners , number of outs, etc. by looking at all situations we might simply have too much data to discern any real trends due to the diversity of scenarios encapsulated in such a large average.