When Joe Maddon opted out of his contract with the Rays two weeks ago, there were immediately rumors that he would be joining up with all of the other 29 teams. (Yesterday, we found out that one of those rumors was true. He’s taking his talents and his back pocket card which drips with analytics to the North Side.) The rumors were understandable. After all, Joe Maddon is a certified genius. He’s gotta be better than that bum in our dugout. (Yes, Joe Maddon is a really smart guy, but so are the other 29 managers. All of them. Yes… even him.)
But then another conversation started that always pops up around this time of year, mostly because there are a lot of managers being fired (and hired) and there’s not much else going on. What is a good manager worth? What does a manager even do that produces value? We know that he makes the strategic decisions for his team, including making the lineup, putting the rotation together, and calling for the sac bunt. But we also know that most managers seem to manage out of the same “book” and while there are improvements over that “book” that could be made, the perfect Sabermetric manager probably clears a win or two more than a standard-issue manager. As Sabermetricians, we’ve spent a lot of time criticizing those strategic decisions, and while mathematically, we’re right, we’ve kinda missed the forest for the trees.
There’s an interesting paper that came out a few years ago in the field of clinical psychology (my home field) which looked at how well therapists did treating people who had panic disorder. The therapists were all conducting their therapy out of the same book. Literally. Panic disorder is one of those disorders that gives itself nicely to treating with a manual. There’s plenty of research saying that these types of treatments work great. Often, the manual specifies what happens in session 1, then session 2, and so on, specifies techniques, and gives standardized “homework” assignments for the client. There was a clinic that was evaluating one of these manuals, and so they recruited several therapists to follow this model in a specialty clinic that worked with people who have panic disorder. The researchers did careful measurement of panic symptoms over time and found that while most people got better, some therapists got better results than others. There was also no correlation between how well the therapists followed the manual and the treatment outcomes for their clients. If a good process is all you need, then it shouldn’t matter who’s administering it, but the researchers surmised (based on other research as well) that results were about much more than process. The relationship between the therapist and the client turns out to be rather important. Might the same thing be true for managers?
I’m fond of the thought that managers have three major jobs. They are in-game tacticians, PR spokesmen, and the guy in charge of wrangling 25 young millionaires who are on a six-month mission where they have to perform nightly. I’m sure the amount of amateur psychology that a manager has to do is staggering. Sabermetricians have focused mostly on the tactical aspect of management, because it’s easily visible and quantifiable. But let’s see if we can find evidence on some of the other work that a manager does. I don’t know that there’s a way to really get at how well a manager handles the media, but as far as his duties in managing the people who wear the uniforms, we might be able to learn a thing or two.
Warning! Gory Mathematical Details Ahead!
Let’s start with a reasonable assumption. Players don’t like losing and, even if in just some small way, it makes them sad. Not massively depressed, mind you. They’re big boys and they know that no one goes 162-0, but they also didn’t get to where they are by lacking in competiveness. We know that there’s even some evidence that being part of a team that loses a lot of games can stunt a player’s development, so losing must have some sort of effect.
I took all events from 2009-2013 (the 2014 Retrosheet file isn’t yet available) and coded everything, this time on a pitch-by-pitch level. I coded all pitches for whether the batter swung, and if he swung, for whether he made contact. I also looked to see, if he took, whether it was a called ball or strike. Once I had done that, I took the annual percentage for each of these for both the pitcher and batter (min 500 pitches faced/thrown). I converted these percentages into odds ratios, and created an odds ratio that expressed the chances of an individual pitch, given this batter and this pitcher, would involve the batter swinging. I have previously used this method in my work at the PA level.
Next I created a simple binary variable on whether the team at bat was losing during that at-bat. I wanted to see, once we’d controlled for the general tendencies of both the pitcher and hitter, whether players actually behaved differently when they were losing as opposed to winning (or being tied). I entered the control variable and the binary losing/not losing variable into a binary logistic regression. The results were interesting. When a hitter’s team is losing, he is less likely to swing, but when he does, he is more likely to make contact (and it’s more likely to go into fair territory). However, when he takes, it’s more likely to be a called strike. The effect sizes are a couple of percentage points, but that sort of change in approach can have big effects for some hitters.
We could make the argument that when a batting team is losing, it makes sense for them to swing less. Why take chances on iffy pitches when your team really needs baserunners. It pays to be more selective then. Plus, swinging less drives up the pitcher’s pitch count, and maybe it would be helpful to get him out of the game. Maybe we’re seeing a perfectly reasonable response to that situation.
But using the same approach, I looked at whether a batter’s behavior varied as a function of whether his team won or lost their last game. In this case, it makes absolutely no sense to change approaches based on what happened yesterday, and yet if a batter’s team lost yesterday, he is more likely to swing and less likely to make contact. Why the worse outcomes? One possibility is that even a slightly depressed mood—one that wouldn’t qualify as clinical depression, but is still mildly sad—can affect a person’s reaction time. I’ve previously argued that one of the most important things that a team can have is a way to deal with losses. Some sort of ritual, whether formal or silly, that says “That was yesterday” and allows players to move on. One might define that as one of the jobs of a manager. (The aforementioned Joe Maddon famously allows only 30 minutes after either a win or a loss for his players to revel or sulk. Tomorrow is a new game, guys.) You don’t want yesterday affecting today.
Well, now that we know that hitters swing and miss a bit more after a loss, we can ask whether that effect varies based on the identity of the man who put that hitter on the lineup card to begin with. To check this, I added new variables. I took the managers from that time period (2009-2013) and weeded out the interims and the bench coaches who managed a couple days while the regular manager was ill or on other business. I entered the manager into the equation (for the initiated, a categorical fixed effect), and the interaction between the binary “won/lost yesterday” variable and the manager. If the manager has some sort of impact on how well a team rebounds from its losses, then we would see that the manager by won/loss interaction would be significant. For swing rates, the overall multi-variate Wald on this variable was not significant, but for contact rate, it was. That means that for contact rate, once we’ve controlled for the batter and pitcher tendencies, we still see that hitters are less likely to make contact on the day after a loss, but that with the right manager, this effect can be blunted or even reversed.
The next thing to do is to look at which managers appear to be the best (and worst) at stunting this problem. We can tell this by looking at the regression coefficients on that interaction term. Because of the way that these analyses work, the coefficients themselves are actually read as “better or worse than Ned Yost.” Why Ned? His name was last alphabetically and the program needed a reference category. So, to list the coefficients would actually be a little misleading. But we can look at the coefficients relative to one another.
The best managers at preventing players from getting losing their contact mojo after a loss:
2) Bobby Cox
4) Tony LaRussa
5) Bud Black
Not a bad group of managers. Bobby V, for all his… quirks… apparently inspired his hitters to bounce back after losses in his one year at the helm in Boston. Cox and LaRussa are Hall of Famers.
The worst managers
1) Cecil Cooper
2) Bo Porter
4) Mike Redmond
5) John Russell
Collins and Redmond still have jobs. Then again, if this list were to extend to 6 spots, no. 6 would be Joe Torre. Like Cox and LaRussa, he also recently capped off a Hall of Fame career.
(For those wondering, Maddon came in middle of the pack.)
The range between the top of the chart and the bottom is something on the order of five percentage points. The way to understand that goes something like this. Assuming an identical batter facing an identical pitcher, we expect the outcome of a swing (contact vs. no contact) to be different based on whether a batter’s team won or lost in their last game by some non-zero amount. Let’s say a contact rate of 85 percent after a win and 84.5% after a loss, for a spread of half a percent (numbers made up on the spot.) We’d expect the “loss penalty” to be bigger for Cooper than for Bobby V (or maybe we might even expect Bobby V’s guys to increase their contact after a loss.)
Fighting the Grind
I’m happy to be the first to say that this is an experimental stab in the dark on this topic. I don’t know whether these effects are stable. I don’t know that even if they are that we could credit everything to the manager. He might have just been sitting in his office and the veterans on the team were the ones really powering the effect. I’m also hesitant to speak too prescriptively about whether making more or less contact is a good or a bad thing. Not making contact on a swing is obviously a strike, but that can be correlated with other good things, such as power. Maybe after a loss, Cecil Cooper’s crew realized that they needed to get back to what they were good at and swing for the fences a little more.
But let’s assume for a moment that the extra swing-and-miss is even somewhat bad. If there’s one thing that we’ve learned from the discussions of catcher framing, it’s that anything that has a slight impact on the ball/strike outcomes of pitches can add up, because a) balls and strikes are worth real runs and b) there are a lot of pitches over the course of a season. If there really are differences between managers in something as simple as being able to keep the players from getting too down after a loss, that would be valuable, especially because managers don’t consume a roster spot and can manage 162 games per year. And yeah, more valuable than the guy who wouldn’t bleed away a little bit of value by bunting once in a while.
I have a theory that we’ve missed the most powerful force in the baseball universe: the grind. Baseball, like any job, can get boring if you do it day after day. Most days, you only get one chance for that validation that you crave, because you only play one game per day. Lose that game and you have to sit with that for 24 hours, plus do a lot of travel and live in close quarters with a bunch of guys whom you may or may not like. After a while it can get depressing, and if the wins aren’t coming, you need something else to lift your spirits. When we speak of managers, we often talk about how he must manage personalities and make sure everyone is working together, or at least tolerating one another. Here, I’d argue, we get some glimpse of an idea of how valuable that could be.
Thank you for reading
This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus.Subscribe now