clock menu more-arrow no yes mobile

Filed under:

Jake Diekman, Miguel Sano, and the danger of small sample sizes

Jeff Banister cited 2015 platoon splits as his reason for letting Jake Diekman pitch to Miguel Sano on Tuesday. Does that rationale make sense?

Jesse Johnson-USA TODAY Sports

As we all no doubt remember, on Tuesday, the Rangers took a 2-0 lead into the eighth inning.  Jake Diekman was brought into the game, and clearly was having an off night.  With a runner on second and two outs (after Brian Dozier's potential tying home run went foul by a foot, and then saw his deep blast to center field be run down by Delino DeShields), Joe Mauer and Miguel Sano were due up.

Diekman stayed in the game to face the lefty Mauer, which made some sense, given that Diekman is also a lefty.  Mauer doubled, bringing up the righthanded Sano.  And while Rangers fans (and some media) were clamoring for Jeff Banister to go get Diekman, he was left in the game, giving up the game-tying double.

The Rangers went on to lose in the 9th.

Evan Grant's game story yesterday included Banister's explanation for the decision to leave Diekman in the game:

Banister stuck with Diekman against right-handed rookie Miguel Sano, one of the best young prospects in baseball. Sano has not hit lefties well in his rookie season, but doubled off the right field wall to tie the game.

Banister’s reasoning: Diekman actually has better numbers against lefties (.233 entering the game) this year than against right-handers (.262) and Sano, for his 128 major league at-bats, had difficulty with lefties (.176), while crushing right-handers (.306).

Scott Lucas offered a sequence of tweets last night that sets forth the fallacy of this rationale, and rather than summarize them, I'll just include them here:

To summarize...there's nothing in their history to lead you to believe that the Diekman/Sano reverse splits in 2015 are anything other than small sample size noise.

And anyone with a basic understanding of statistics should know that.  Here's what Fangraphs says about the stabilization of sample sizes for platoon splits:

● Obviously, we all know that batters typically do worse when facing a same-handed pitcher (e.g. a lefty batter facing a lefty pitcher). However, we can’t make any overarching statements about how much batters struggle against same-handed pitcher; the size of a lefty/righty platoon split varies from batter to batter.

● The same can be said for pitchers: while pitchers are normally better against same-handed batters, the size of a pitcher’s platoon split varies from pitcher to pitcher.

● Batters don’t have their platoon splits stabilize until at least 1,000 plate appearances against each hand (around 2,000 for right-handed batters), while pitchers have their platoon splits stabilize much faster (500-700 plate appearances against both hands).

So there are two problems with Banister's explanation:  1)  Nothing in the pre-2015 data suggests that either player has reverse splits, and 2) the 2015 sample sizes for the two players are way too small to draw conclusions from.

Add to that that Diekman is a fastball/slider pitcher, the type of pitcher that has normal (if not larger than normal) platoon splits, while pitchers who have smaller or reverse splits generally rely on a cutter or change up (pitches that neutralize the opposing-handed hitter), and it makes the decision even more baffling.

And of course, there's the reality about Diekman's command and stuff on Tuesday...per Evan's story:

Diekman entered with a 2-0 lead in the seventh and walked Escobar when he couldn’t get his 98-99 mph fastballs down in the zone. He tried to take a bit off to locate better, dropping to 95-96, but the problem was that the fastball was, in Diekman’s words, “extremely flat.”

So Banister, according to Evan's game story, opted to stay with a struggling pitcher who didn't have anything close to his best stuff in order to take advantage of a reverse platoon split that appeared to be a statistical anomaly based on a very small sample size.

And I could also go on about how nonsensical it is to be willing to use Shawn Tolleson for a second inning on Saturday with an 8 run lead, but not be willing to use him for four outs on Tuesday to protect a 1 run lead, but I'll leave that alone.

I've defended Jeff Banister this season, and I generally think he's done a good job.  And bullpen management is probably the most difficult part of a manager's in-game decision making, particularly when you don't have a good bullpen (and the Ranger pen isn't good).

But the rationale that Banister has offered for his last couple of baffling decisions is very discouraging.  Maybe its just ex post facto rationalization on his part for decisions he'd do over again if he had the opportunity.  I don't know.

But is worrisome.