Yesterday, Jonah Keri wrote an interesting and insightful article dissecting Game One of the World Series, including the various pre- and in-game managerial decisions. While I agree with most of what Keri says, there are a few things I think are worth taking another look at.
With respect to Tony Womack bunting with runners on first and second in the second inning with the Cardinals already down 4-0, Keri writes:
I didn’t like the bunt by Womack, but as I said, I think this was more than just one bad tactical move in play here.
Of course, most people with sabermetric leanings generally disdain any sacrifice bunt, let alone one in the second inning of a 4-0 game. In fact, sabermetric wunderteams Oakland and Boston had a total of 37 sacrifice bunts combined, all season long. Now, we all know by now that a perfect (batter out, runners advance) sacrifice bunt by an average hitter in an average context reduces a team’s run expectancy and reduces that team’s chances for a big inning–not quite what you aspire to if you are down 4-0 in the second inning against a team with arguably the best offense in baseball.
However, this is not an average hitter in an average situation and, as far as I know, Tony La Russa did not ask Terry Francona if he would be willing to take an out and place the runners on second and third. In other words, Womack is a below-average hitter with little power–followed by a similar hitter in Mike Matheny–and it’s possible that Womack, while attempting to bunt, might on average do better than a perfect sacrifice; he might produce a hit, reach on an error, or even in rare cases work a walk. Of course, he might also strike out, ground or pop up into a double play (not too likely), or otherwise make an out without advancing the runners.
So despite the fact that a successful sacrifice bunt with runners on first and second generally reduces a team’s run expectancy, we have several mitigating factors. One, Womack has a lifetime OPS of .681 and is facing an above-average pitcher in Tim Wakefield (whether bunting against a knuckleballer is more or less difficult I have no idea, so I’ll have to leave that out of the analysis), such that the difference between the run expectancy from bunting and from swinging away is reduced. Two, as BP’s James Click explained in an excellent series on the efficacy of the sacrifice bunt in various situations, the bunt is favored more when the following hitters are weak overall and in particular in the power department. In this case, the two hitters following Womack were Matheny and So Taguchi, inveterate lightweights.
Finally, and most importantly, Womack is an excellent bunter, often among the league leaders in bunt hits, and one of the fastest players in baseball. So what does that mean in terms of the sacrifice bunt attempt?
Let’s take a look at what happens when the number-two hitter (the most prolific bunting slot) in an average NL batting order attempts a sacrifice bunt with a runner on first and second and no outs. First of all, the average run expectancy in that situation, again from the “two hole,” when not bunting, is 1.709 runs (based on National League data from 2000-2003). After a successful sacrifice bunt, and the #3 hitter at bat, the average run expectancy is 1.682, thus the oft-quoted reduction in run expectancy when attempting to bunt. Of course, that reduction in run expectancy refers to a perfect bunt, and not to a bunt attempt. Maybe a bunt attempt, especially with a speedy, talented bunter, is better than a perfect bunt. Let’s see.
Here is what happens when an average #2 batter in an NL lineup attempts a sacrifice bunt with runners on first and second and no outs, early in a game (before the seventh inning), when the defense is ambivalent about expecting a bunt:
When the attempt ends in a bunt (or fouled third strike):
Batter out, both runners advance 60.9%
Force at third 10.7%
Reach on error 5.0%
Batter out, no advance 4.9%
Safe FC 1.7%
Force at second, one runner advances .7%
When the bunt ends in a “swinging away” (or strikeout, no foul bunt) or walk:
Batter out, no advance 18.3%
Force at second, one runner advances 7.1%
Batter out, both runners advance 5.6%
Reach on error 2.6%
Force at third 1.1%
Safe FC .4%
The total run expectancy for all of these results combined is 1.730, which is greater than the original run expectancy (1.709), and better than a “perfect” sac bunt (RE = 1.682). This suggests that early in a game, a bunt may in fact be warranted, especially if the batter is an average hitter or worse and the following hitters are singles-type hitters.
The above data is based on all bunters, fast or slow, talented or not. What if we only look at speedy players who are also talented bunters (based on their “speed scores” and bunt hits attempted per PA)?
When the attempt ends in a bunt (or fouled third strike):
Batter out, both runners advance 58.9%
Force at third 9.7%
Batter out, no advance 6.9%
Reach on error 4.4%
Safe FC 1.9%
Force at second, one runner advances .9%
When the bunt ends in a “swinging away” (or strikeout, no bunt foul) or walk:
Batter out, no advance 17.5%
Force at second, one runner advances 6.8%
Batter out, both runners advance 3.9%
Force at third 1.9%
Reach on error 1.9%
Safe FC 0%
The total run expectancy for the above is 1.760, .03 runs better than for all bunters, better than swinging away, and much better than a perfect sacrifice bunt. With Womack, an extremely good bunter and fast runner, we would expect a run expectancy even higher than that.
So once we include the potential results of a sacrifice attempt and take into consideration the fact that we have the “ideal” bunter in Tony Womack (poor hitter, excellent bunter, fast runner) batting out of the seven-hole, all of a sudden a sacrifice bunt in the second inning with his team down 4-0 doesn’t look so bad, and in fact might be the unequivocally correct play.
Once again, from Keri’s BP article:
Speaking of Womack, credit La Russa with abandoning the plan to keep his second baseman in the leadoff spot, instead opting for Edgar Renteria at the top of the order.
Again, sabermetric wisdom eschews the traditional scrappy, speedy, base-stealing, but low-OBP guy at the top of the order. However, there is more to constructing an optimal lineup that these sabermetric “rules of thumb.” In fact, if a player is fast and an excellent baserunner, he may be a good candidate to lead off, even if his expected OBP looks more like a batting average than an on-base average. Again, Womack is one of the fastest players in baseball and an excellent baserunner as well. Superlwts has him tied with Cristian Guzman and Rafael Furcal, also no slouches in the speed department, as the best baserunner in baseball from 2000 to 2003.
When I use a game simulator to simulate different lineup combinations for the Cardinals, Womack batting at the top of the order produces just as many runs as Womack batting seventh or eighth. The computer simulator uses up-to-date batting projections for all the players as well as their projected speed and baserunning abilities. So once again, traditional sabermetric wisdom is not always correct, especially when dealing with something as context- and player-specific as lineup construction. For what it’s worth, although La Russa has at times batted Womack down in the order during the regular season, his recent post-season decision to bat Renteria first and Womack seventh was predicated on Womack’s back problems as much as anything else.
Using a rare two-fer of lame managerial excuses, Tony La Russa opted to start light-hitting So Taguchi over a good hitter in John Mabry. The move, it was felt, would improve the Cardinals’ left-field defense at the face of the Green Monster, while also preserving Mabry for a pinch-hit appearance, given his skill at that role.
Using the same computer game simulator, Taguchi in left, and Sanders as the DH, fared as well as Sanders in left and either Mabry or Cedeno DH’ing. The reasons are actually quite simple. Taguchi is indeed a better fielder than Sanders at this point in Sanders’ career. In fact, my UZR projections have Taguchi projected at four runs per 150 games better than Sanders. In addition, Game One starter Woody Williams is a flyball and relatively low-strikeout pitcher, thus creating more chances for Taguchi or Sanders in left field. Taguchi is also a much faster baserunner than Mabry.
Although Mabry did in fact have an .867 OPS this year, he is a 34-year-old lifetime .741 OPS hitter. In other words, at any given time, we expect him to hit a lot closer to that .741 than his .867 in less than 300 plate appearances this year. As well, if you don’t start Taguchi, he has no value on the bench except as a late-inning defensive replacement for Sanders or as a pinch-runner. If you do start Taguchi, you have available a decent left-handed pinch-hitter in Mabry and switch-hitter in Cedeno.
Keri also suggests in the article that if you are going to DH someone other than Taguchi, Mabry is the clear choice over Cedeno. Is that true? As I already said, Mabry’s projected hitting is nowhere near his 2004 performance. When we do these kinds of analyses (e.g., who should be playing over whom), we use batting and defensive projections, which are based on a player’s entire career rather than his 2004 stats only. In fact, a player’s 2004 stats can be quite misleading, as compared to his expected performance, or projection, especially if he has batted only 200 or 300 times this year, as has Mabry.
Mabry, according to the batting portion of my Superlwts projections, is projected at -2 batting linear weights per 150 games. Mabry does in fact have an enormous platoon ratio edge against right-handed pitchers, but most of his historical plate appearances have been versus righties, such that his projected batting lwts versus a RHP is only a little higher than his overall projected batting lwts, about +2. Keep in mind that for various sound mathematical reasons that I won’t get into, it is not correct to simply look at a player’s performance against all RHPs (or LHPs) in order to project their performance against RHPs (or LHPs).
Cedeno, on the other hand, has a projected overall batting linear weights of -7 per 150 games, 5 runs worse than Mabry. Cedeno’s true platoon ratio (which, as with Mabry, we estimate from his actual or sample platoon ratio and regress it towards the average platoon ratio for switch-hitters) slightly favors him as a lefty hitter, such that his projected batting lwts versus a RHP is -6 rather than -7. So Mabry is still 9 runs or so better than Cedeno as a DH.
But offense and defense do not an entire player make. In fact, Mabry is a slow, poor baserunner, while Cedeno is much faster. Superlwts has Mabry as 5 runs (per 150 games) below that of an average baserunner and Cedeno, 3 runs above that of an average baserunner. So the difference between their baserunning almost completely cancels out the difference between their batting skills.
I guess the moral of the story is that when it comes to an analysis of manager strategy, be it the efficacy of the sacrifice bunt, or lineup construction, there is often more to it than meets even the sabermetric eye.
Mitchel Lichtman currently works as an analyst for a major league baseball team. He has been doing sabermetric research and analysis for more than 15 years and has a B.S. in Psychology from Cornell University and a J.D. from the University of Nevada Boyd School of Law. Mitchel can be reached via e-mail, here.