I want to write about player development. I want to do it mostly because there’s not a lot of writing about it among the statistically inclined sort, which I find to be very strange. Everyone’s always looking around for the #NewMoneyball and, frankly, it’s staring right at them. Young, cost-controlled players return value—on average—at a dollar per win rate that’s about half what a team would pay on the free agent market. And that’s just the average. If a team is any good at player development, it can assemble a roster of young, cost-controlled players and ride that wave for a long time. If a team could nail down player development, they’d have a bit of an edge, wouldn’t you say?

So, why do Sabermetricians rarely talk about player development?

Player development isn’t just the sort of thing where you can run a regression, write 2,000 words about it, and figure it all out. It’s not like there’s a shortage of baseball people thinking about player development. Each team has someone with a job title something like “Director of Player Development” and an entire minor-league and instructional system that is specifically focused on the topic. Player development is an entire field of study, kinda like “physics” or “literature.” But even if it’s too big to get completely right, it at least seems ripe as a field of study.

It’s strange that the state of the art research in this field is probably the study that looks at the WAR produced by “graduates” of each farm system, looks at the top three teams, and says “Well, I guess the Cardinals doing something right.” Can’t we, as researchers, go a bit deeper? Like any endeavor in Sabermetrics, it’s not that we will be able to uncover all of the answers just through poking around our spreadsheets, but maybe we could come up with something to add.

Before we get going, though, let’s talk about how this will be different than our usual areas of study.

1) The information chasm

When we write about happenings at the MLB level, we have plenty of data from which to work. Want to write about bunting strategy? There are hundreds of bunts (and non-bunts) dropped every year. Want to write about how valuable Luis Valbuena is? Everything he did (between the white lines, anyway) this year is recorded with some amount of detail. How many times did he swing and miss? How many times did he make a play at third that others might have missed? There’s an app for that.

The problem is that a lot of player development (or what we think of as player development) happens in the minors where the data aren’t quite as good. There’s PITCHf/x data in the Arizona Fall League and, if you really want it, play-by-play data are available for minor-league games. We have minor-league stats (although more on those in a minute). But really, that’s just what happens in the games. Player development is about more than just having a bunch of guys who hang out together in Oklahoma City and play a few games against some guys from Nashville. There’s a lot that goes on between the final pitch and “play bal" (and yes, I got that right.)

Teams spend a lot of time and resources working with the minor leaguers to try to teach them things. It’s not a big secret what they’re teaching. Here’s how you throw a curveball better. Here’s something you can do with your swing. And of course, if it were that easy, everyone would be an all-star. But the fact is that some guys get it and some don’t. We don’t know how the minor-league roving pitching instructor tried to teach the kid from Double-A that slider. We don’t even know that it was him. Maybe one day that kid was just goofing around in a side session and hit on something that worked.

And yes, teams have their own data on their own players, which of course they don’t share. In a perfect world, we’d all have access to all that, but we live in a world where you only get nine rolls of the Skee-Ball before you have to put in another token, so we know the world is not perfect.

So, how to get around it? Until we have better data at the minor-league level, we’ll have to just have to make do with what we have. In good news, players don’t just stop developing once they get to the majors. They grow and change over time, and sometimes teams will bring a guy up who’s “good enough” and have him work on something at the MLB level specifically because it’s best learned there. We won’t get a chance to know who’s teaching him what and how they taught it with any systemic precision, and we’re basically at the mercy of figuring that any improvements among a team’s pitchers must be because of the pitching coach. But at least we can study how some player develop and make inferences about how it might happen on the farm. Maybe someday we’ll be able to do studies on questions like “What is the best way to teach a curveball?”

2) We need to let go of WAR

For a moment consider the statement “He is a two-win player.” We toss that out rather casually around these parts. It’s not that WAR is a bad stat or that it doesn’t have its uses. And yes, the eventual goal of player development is to produce players who do the sorts of things that win games and thus drive up their WAR. Actually, the problem in that sentence “He is a two-win player” isn’t the reference to WAR. It’s the word “is.”

There’s something that happens when you start putting numbers on everything. It becomes easy to blur the line between “What Smith did last year was worth two wins of value” and “Smith is a two-win player.” The first describes what happened. The second says that there’s something inherent in Smith that makes him worth two wins. It’s a small thing, and perhaps I’m making too much of a common grammatical construction, but it’s a trap that’s easy to slip into. Player development is about growth and change in a player, and when we start using words like “is” we begin to assign characteristics to a player. Maybe it’s true, but in order to do player development right we have to think about what a player could be, not in the sense that every team blog is convinced that all of its prospects are future 6-win players, but in the sense that to do this right, a team has to believe that a player’s current WAR is just a temporary stop along the way to greatness. Or at least serviceable mediocrity.

3) Embracing the complexity

In Sabermetrics, we’re used to asking questions that have definable answers. Maybe they’re questions that we’d need better data to get the answer, but with enough work we could figure it out. How much is a home run worth? What are the optimal conditions in which to lay down a bunt? We’re used to finding one set of guiding principles that can guide decisions given any set of circumstances.

What to do when the question itself is a little bit more ambiguous. We have a lump of clay, or, perhaps more accurately, an 18-year-old kid fresh out of high school and away from home for the first time in his life. What shall we mold him into? What does he even need? If he’s a pitcher, does he need work on mechanics? Does he need to learn a new pitch? Does he need to learn the strategic game of pitching? Does he need to learn to control his emotions? Before we start answering the question, we have to figure out what the question is first. That’s not comfortable ground in Sabermetrics.

In baseball, there are plenty of ways to be successful. There are multiple skills that a player can have and multiple ways in which they can be employed. Some skills work better than others. Some are great in combination with others but pointless by themselves. (97 mph? Great! 97 mph with no movement… nevermind!) In another field which studies development, child development, we know that there are a whole bunch of paths for someone to end up as a “successful” adult. One person’s path wouldn’t work for someone else. At this point, we have such a poor understanding of what the underlying skills are for a baseball player, how to measure them, and how they all fit together.

This is where it’s helpful to have someone to call on who has seen a lot of different players over the years and has a wealth of knowledge on seeing different types of players. Yep, maybe there’s room for those baseball lifer guys after all.

4) Don’t scout the box score

Perhaps the biggest frustration in trying to study player development is that as Sabermetricians, we are trained to first go to the box score or the play by play log or something like that. What were the results on the field, because in the end isn’t that what it’s all about? Except that the box scores can lie. The dirty little secret of minor-league baseball is that the score doesn’t actually matter. That’s a bit of an overstatement and an oversimplification, but it’s more true than most people would like to believe.

Teams will routinely “take a pitch away” from a pitcher, especially if they feel that he’s overly reliant on it. Or if they feel he needs more practice with his other pitches for them to develop. They’ll ask a hitter to focus on hitting line drives rather than just swinging real hard in case he hits it. He’s not really a line-drive hitter and he’s not a great hitter in the moment because he’s being asked to do something that he’s not great at, but something that he needs to learn. And maybe he’s getting better, but if you scout the box score, you’ll see a guy who’s not pitching so well at Triple-A, because he’s pitching without his best pitch.

Yes, when a guy is hitting .400/.600/.800 at Triple-A and it’s not because he’s playing in Colorado Springs, it means something. But always treat minor-league stats with a grain of salt.

5) So how can I do good research on player development?

For one, it’s a good idea to just get comfortable with the idea that the type of third-decimal precision that usually appears on these pages just isn’t a good starting point for the topic. If you want to study 22- and 23-year-old major leaguers, that’s a very special subset of players, filled with all sorts of biases, and people will yell at you that you can’t generalize your findings. Sometimes you won’t have a perfect proxy for what you want to measure. Run your study anyway. Point out that there are biases like any good researcher should (and that there are no really good ways around them). There’s the fear that biased information is worse than no information, but sometimes that works the other way around.

For another, it’s perfectly acceptable to bring in expertise from other fields, even if it’s just theoretical musings with little data to back it up. Know a little bit about kinesiology? Great! I have no idea on any of that stuff and I’d love to read it. Have a master’s in public health? Cool. Want to talk about ways in which organizational philosophies can be disseminated to a big group of people? (It’s great to say “We want everyone to be able to lay down a bunt,” but how do you convince them to try and not hate you for it?) While baseball players are a special data set, they aren’t space aliens. If it works on other groups of humans, so it might just work on them too!

But the reality is that we might have to turn to data sources that we might not have even thought of before. For example, scan the scout and prospect writers every day for mentions of guys who are “experimenting with a new pitch” in Double-A. Use those mentions to build a database of guys who can populate a before-and-after sort of analysis. You won’t get everyone and it’s not the textbook way that a research methods professor (or in my case, former research methods professor) would draw it up, but that’s data. Use it.

The point is that where the field is now, anything done in even a semi-responsible way would be a welcome addition.

Thank you for reading

This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.

Subscribe now
You need to be logged in to comment. Login or Subscribe
I like this. Except I think you should occasionally scout the box score. You'll find players there. (Ahem Conforto.)
Instructional design is a missing link in baseball at this point, and I'd love to be able to research it more intently. Would be fascinating to see what big league teams are doing from an ID and human performance technology standpoint, as there's lots of room to implement learning design in orgs.