Pebble Hunting: The Baseball Sandbox Experiment

When Ken Rosenthal wrote about his interview of Astros GM Jeff Luhnow last week, he included this line: “Some think the "team of the future" is making frequent use of shifts merely to gather information for the future. Not so, Luhnow said.”

Just like that, in two sentences, Rosenthal introduced us to a conspiracy that we were unaware of, then quoted a denial of it. I’ll take the Astros at their word, and trust that they’re—

Hey, why are you winking at me? No, I’m serious. I’ll take the Astros at their wor—

Stop winking at me. I mean it. I believe the Astro—

Ugh, you cynics are the worst. Really, if the Astros felt like testing some new theory of defense in game situations, it would be easy enough to do it on backfields, instructional leagues, rookie ball games, places where nobody’s watching and winning doesn’t matter. Look, money-saving service-time shenanigans are one thing; opting not to invest short-term expenses on a lost cause is one thing; but, thus far, so far as we know, no baseball team has used its pitiable place in the winning cycle as justification not to try to win once it actually takes the field. Doing so would be a whole new doctrine for tanking, one that I don't believe has existed at this level of the sport to date.

But now imagine with me: What if they were doing experiments just to gather research for the seasons that matter to them, or were at least open to the idea? What if you had a team that was, under practically no circumstances, going to compete for an achievement of significance this year? In a ruthless, business-oriented approach to running a team, every asset would be maximized for future profits, and one of the assets that a bad team has is that it gets to play 162 games however it wants to play them. You and I can’t run an experiment using real-life baseball action. A major-league team can, theoretically. If chaining a top prospect to the minors for a few months just to take money out of his pocket is considered moral in Major League Baseball, why not a few controlled studies?

So here are the rules: The experiment has to look enough like baseball that you don’t get expelled from the league or otherwise sanctioned. For that matter, it’s probably best that even your own players can’t suss out your motives, lest morale suffer when they realize that despite all their range they’re still just a rat in a cage. So it has to either be invisible or a credible effort at winning.

It can’t otherwise do damage to your team’s long-term goals. Studying what would happen if you let all your pitchers pitch through a UCL strain or labrum tear, for instance, might be fascinating, but we’re trying to win (future) baseball games here. If you learn something that nobody else watching you can learn, that’s even better, but we’ll consider any gained knowledge to be useful. What do you do? Here’s my team:

Catchers
Pitch selection randomized. Part of pitch selection involves trying to trick the batter. The batter, though, knowing you’re trying to trick him, adjusts his expectations to incorporate the possibility of trickery. Which might lead the pitcher to try to trick him by throwing the least tricky pitch. At which point the batter, annoyed that the pitcher is standing on the mound like a gol’ dumb Starting Lineup action figure, will call timeout, and that’s why no baseball game in history has ever been completed before mom called everybody home for dinner. The point is: The trickery aspect of pitch selection helps only if you’re trickier than the other team. If they’re trickier than you, then every choice you make costs you. So what happens if you randomize things? You take all the batter’s smarts and use them against him (or at least neutralize them). Maybe this is the right way to play! Further, nobody’s really come close to cracking the pitch sequencing mystery, partly because there are so many layers of context to each pitch selection that it’s hard to speculate on counterfactuals. Randomizing pitches could actually make it quite a bit simpler to isolate the effectiveness of each type of pitch in each situation and sequence.

Further, it’s probably best that you don’t just randomize pitch selection. Randomize which plate appearances will feature randomized pitch selection. Half the time, you can outsmart the hitter, who will never know he’s being outsmarted. The other half, you neutralize the smarts. (Also, this way you can actually compare the randomized pitch selection to the traditional method and determine whether it’s poppin’.)

Fielders
Two fast outfielders, five infielders, whole grip of groundball pitchers. Answers about four questions at once: 1. Could two outfielders, with exceptional speed, approximate anywhere close to the coverage of three typical outfielders—who, after all, overlap on a number of plays, and whose theoretical range currently covers ground (foul territory, where a ball dropping is less damaging; the stands, where a fielder can’t go; the lines, where contestable balls are probably hit less frequently) that isn’t all that valuable to cover. Certainly, two fast outfielders can’t literally cover as much ground as three average ones, but given the limited size of a baseball field, they might functionally cover an acceptably close amount, if… 2. The range added to the infield was sufficient. Might it be? Consider an example of what happens when a third infielder is added to one side of the diamond:

The shift has worked on Ortiz this year. The average major leaguer has a Batting Average on Balls In Play (BABIP) of .232 on ground balls. Ortiz has hit 81 groundballs, 70 of which were hit into the shift on the right side, and has only eight hits (.114 BABIP).

Alas, though, there’s a problem with such shifts:

Ortiz compensated by going the other way. He has hit 20 opposite field liners and grounders, and 16 have been hits (.800 BABIP), many of which would have been routine outs with a regular defensive alignment.

So add a fifth infielder and, instead of having a shift with a wide-open field for the batter to hit into, you simply choke off his favorite hole without sacrificing anywhere else. Anywhere, at least, except for the outfield, which then raises the third area of inquiry that this would illuminate: 3. How much can batters decide whether to hit a groundball or a fly ball? Part of what makes the shift interesting for an analyst is that it tilts the incentives enough that we get to see whether hitters are willing to do something different; further, we get to see whether they are capable of doing that something different. This year, we’re seeing whether all sorts of players are capable of laying down a rough bunt. But what about hitting a fly ball down an unguarded line by choice? And what about hitting a fly ball by choice 4. against a grip of groundball pitchers? Groundball pitchers generally already aim to get groundballs. But what if you tell them that a fly ball is particularly lethal, because there are no outfielders (or, at least, two-thirds as many)? Can groundball pitchers get even more groundballs in this scenario?

If a five-man infield can get a groundball pitcher’s groundball BABIP down to, say, .100, then does the groundball pitcher have the freedom to completely ignore strikeouts (because practically any ball on the ground will be an out) and see a corresponding decrease in walks and home runs? If he can, under these unusual circumstances, get his groundball rate up to, say, 80 percent, with a BABIP on grounders of .100, then perhaps he becomes totally unconcerned with walks, knowing that he’s got something like a 50 or 60 percent chance of a double play from the next batter. Might this pitcher, by being terrified of fly balls, totally immune to the threat of groundball hits, absolutely apathetic toward missed bats, and largely unworried by walks, be able to increase his groundball rate to even higher than 80 percent? Or would the batter’s role in this struggle then come into play, and might he be so single-mindedly focused on hitting fly balls that his agency neutralizes the pitcher’s will? We don’t know! We could know! All of these numbers are probably unrealistic! But maybe they're not! Long live teams tanking!

(Side note: If it turns out that five infielders is so good that it’s too good—that, in essence, it creates the same inefficient overlaps that we were trying to smooth out in the outfield—then it might conceivably free a five-infield team up to hire low-range sluggers to play the infield, even in the middle infield. Imagine not having to pay a shortstop premium anymore, either in dollars or in sacrificed offense.)

Pitchers
Starters all pitch in relief on throw days. Not an original idea, and we haven’t seen it implemented because there are good reasons it might not work. But it’s not an original idea because there are all sorts of reasons it might work. Imagine getting 30 elite relief innings from each of your starters at no cost. What’s that, three free WARP per year?

or

Pitchers not allowed to throw within, say, three mph of their max velocity. If they do, they get fined. What does this do to health? What does this do to movement, command, durability, stamina? Safe to say pitchers have an idea already, but what happens when they’re really committed to this way of pitching, spending months doing it? It’s probably the case that this era’s big-time velocity is good for pitchers and bad for hitters. It might not be. Would be fun to be the team that discovered this.

or

An all-bullpen staff, obviously. This is the first thing I’d test. This is the “I wish for unlimited wishes” of having a baseball organization to screw around with. It seems inevitable that at some point in 10 years, 30 years, 50 years, we’re going to see the all-bullpen staff, nobody throwing more than 50 pitches in a game, nobody facing a lineup three (or even two) times. The only obstacle to it eventually happening is that somebody will have to be first, and it could be scary, if not disastrous: Will you run out of pitchers? Why do you run out of pitchers—because of injuries, or because of extra-inning games, or blowouts, or because half of your pitchers will turn out to be terrible, or because you spend too many outings pinch-hitting, or what? How do you adjust for this? How hard is it, in reality, to find pitchers mid-season who can slot into this rotation? How many runs a game do you allow, and how cheap can you build that staff for? Can you reap extra benefit by switching pitcher styles, velo pitchers followed by junkballers? What does a knuckleballer in the third and fourth inning do? Do you still have a closer? Do the same guys start every time, or do first-, middle-, and late-innings distinctions disappear completely?

Is there a way to get more innings out of your best pitchers (as in the current model) without tiring them out? Is there a way to get your best pitchers in higher leverage (as in the current model) without ruining the rhythm of the experiment? Can a manager be trusted to pull a pitcher after three innings if he looks great, or does the temptation to stick with the effective pitcher always ruin the plan? Can you use your Triple-A team as little more than a storage warehouse for pitchers? Can you, as Zachary Levine suggests, rotate these arms from Triple-A to the majors and have them pitch eight out of 10 days in the majors, then basically just rest for 10 days in the minors, until they’re eligible to come back up? How strong is this staff in September? How healthy? How much do you sacrifice in platoon advantages? How do the players take to their new roles, knowing they won’t get wins or saves reliably? Bottom line: Can a staff of relievers mimic the league’s reliever ERA over 1,458 innings for the cost of 12 pre-arb relievers and scrub ex-starters?

(I’d also like to see pitchers skip the minors completely, or skip from Low-A to the majors. It feels somewhat likely (if not actually likely) that we’re wasting the best years of a number of pitchers’ careers by developing them in the minors, while their arm strength is at its strongest. Would be interesting to see how many Jose Fernandez’s are out there—not as good as Fernandez, obviously, but capable of pitching roughly as well in the majors as in the low minors.)

Hitters
Hit against the shift every time. I’m not as interested in bunting as much as hitting toward the hole. I just want to see to what degree batters can change the direction and trajectory of their batted balls. So far, based on Ben Lindbergh’s series, it appears that once a batter demonstrates he’ll bunt against the shift, the third baseman will stand reasonably close to a normal position, but the shortstop’s natural position remains unguarded. So hit for it every time. For all the analytic interest in the shift, and the possibility that this is a major change in the way the sport is played, such that we will never consider defensive positions in the same static way again, it seems possible too that the shift is incredibly beatable and can be eradicated like polio with just a bit of effort. (Might also not be.)

Manager
Just get some guy. Not just any guy, but some guy from the pool of guys who get hired to be GMs these days: Smart guy (or gal), baseball background, but not a former pro ballplayer. Let him do whatever crazy stathead stuff he wants—one of those optimized lineups that look so funny, a near-total ban on sacrifice bunts and pitchouts, closer by committee. Have him walk out and pull his famous, highly-compensated ace in the sixth inning of a no-hitter because we like our chances with the bullpen instead of risking a third trip through the order with the same pitcher. The point isn’t whether any of this crazy stuff would work; the point is just to see how the team reacts. There’s an idea out there that teams won’t respond to a manager who didn’t wear a uni from birth to death, and we might see that idea played out the first time a batter grounds out to the right side to move a runner over and gets scolded instead of buttslapped by his manager. But maybe not. Maybe the manager pulls the strings, and the players get used to it and get their leader-of-men fill from their coaches who would, naturally, still be Baseball Men capable of saying “horseshit” every eighth word. If it turns out that we’re overestimating the players’ reliance on ex-ballplayers to fill out their lineup cards, then the team that discovers this would have literally (literally!) billions more managers to choose from. Could be an advantage.

And if it fails, then you get to fire your manager before the next season, a guaranteed fire-up-the-team move before they start competing for real again.

You can certainly do better than this, given a team playing for nothing and 162 real-as-ruin baseball games to experiment with, so what would you do?