Baseball, like every other sport, is a results-oriented business. At the end of the day—or the season—the only things that truly matter are wins and losses. Sometimes this is unfortunate, as it often muddies the way coaching decisions are evaluated.
Should we pinch hit with a righty? Should we go for it on 4th-and-1? Should we foul before they can attempt a three-pointer? Either side of those coins will routinely be praised or ridiculed, depending on whether or not the game is won.
One popular draw of the sabermetric movement is that the reliance on empirical data forces observers to judge the process instead of the outcome. While the majority of people focus on what happened, the statistically-informed focus on what should have happened.
Enter Charlie Manuel. Manuel was criticized because on Wednesday night he used Jim Thome as a pinch hitter, and then left Thome in when a left-handed pitcher was brought on to face him. Had Thome gotten a hit (or even a sac fly), the old-school crowd would have praised Manuel because of the outcome. He “knows his ballclub.” He “pushes the right buttons.” He “instills confidence in his players.”
Thome struck out, and Manuel was ripped.
Interestingly, his rationale for leaving Thome in was more disturbing than the act.
Matt Gelb writes that Manuel explained, “Thome is 2 for 11 off the guy [Javier Lopez] with three strikeouts. That means he put the ball in play eight times. If he hits a ball, as big and strong as he is, we have a chance to score a run.”
This is troubling for several reasons. First, note that Manuel particularly liked the matchup even though Thome was batting .182 off of him. Second, Placido Polanco was available to hit for Thome. Polanco is not only right-handed, but was the third hardest batter to strike out in the NL last season. Finally, most damning of all, let’s dive deeper into Thome’s 11 at bats that were the basis for the decision.
As you can see, there were three in 2003, three in 2004 and only five within the last seven years. The decision was made primarily based on at bats that occurred when Thome was an everyday first baseman, with an OPS over .950 and a WAR above 4.0.
Manuel has always relied heavily on batter vs. pitcher data, so it’s hard to be surprised by his thinking. The disturbing trend is that he continues to do so with an increasingly older team, relying on increasingly older sets of data. Manuel has to understand that his players, particularly his aging players, are not who they were in their primes.
Reliance on outdated statistics displays a strange form of bias, akin to being unable to separate in his mind the players on his roster from the players they used to be. Making decisions today based exclusively on data from more than half a decade ago is like trying to win on Jeopardy! by studying yesterday’s clues.
Consider the first game of the Giants’ series, against Tim Lincecum on Monday night.
Gelb wrote in his game preview (and I have no reason to doubt him), “Why is [Juan] Pierre starting over John Mayberry Jr., who is the superior defender, in a big ballpark? It probably has to do with these career numbers vs. Lincecum.”
Gelb didn’t break them down by year, but I will.
Sure, Pierre has had success against Lincecum. But the data shows just 12 at bats, much of it four or five years old, all during a stretch when Pierre was a .294 hitter, which he has not been in the years since.
While these at bats are at least more recent than Thome’s against Javier Lopez, it’s wrong to allow this miniscule sample size to carry more weight than the hundreds of at bats since.
There’s a reason small sample sizes aren’t dependable. For an example, let’s go back to Wednesday night’s game and think about why Polanco was available to pinch hit in the first place. Polanco was on the bench not only because of his early season struggles, but because Manuel wanted Ty Wigginton to start at third base against Matt Cain.
“How did I decide on it,” Manuel asked himself. “Nix is 8 for 20 with a homer off this guy. Wigginton is 3 for 9 with a homer. Those guys have seen them, have some hits on him, so why shouldn’t I put them in the lineup?”
Yes, Wigginton was 3-for-9, for a shiny .333 average and a homer. But stats can fluctuate so much over nine at bats. One of those three hits was a ground ball through third base in a 10-1 ballgame. If that ball had been gobbled up for an out, making Wigginton a career .222 hitter against Cain, would he have been on the bench? The point is that one measly ground ball on one random day, and one homerun (at Coors Field by the way), shouldn’t matter as much as his 1,000 at bats over the last two years, in which he batted under .250.
The blame doesn’t all fall on Manuel when the Phillies’ offense comes up empty. Ruben Amaro built this roster, and Manuel has other coaches to help him make in-game decisions. Plus, I’m just like everyone else: Had Thome popped a homerun, I’m probably not writing about this.
The roster and the injury bug have dealt Manuel a bad hand. To his credit, he has not publicly used that as an excuse. If he did, it probably wouldn’t be well-received, but I think it would be defensible. If he disregarded important statistics like career platoon splits or contact rates to play a star, or go with the hot hand, that too would often be defensible.
But for him to repeatedly explain after games that his decisions are based on statistics—and then use insignificant or outdated ones—shows a lack of understanding about how the decision-making process works. That, to me, is indefensible.