# A Wild Weekend of Bunts

A few weeks ago, I introduced seven questions I expected to explore over the course of this season with regard to the use and efficacy of bunts. I’ve been tracking bunts by National League teams in an attempt to gain a better understanding of their effect on win probability and whether they’re being used wisely. As is the case with anything one might study for three weeks in April, the results bounced around a little bit early on, before three crazy days of bunts this weekend.

More on the weekend, and the season, in bunts after the jump.

On Friday, teams attempted 10 bunts. Four of ten resulted in the batter reaching base- two via singles, one on an error, and the fourth on a fielder’s poor choice. Another was the rare successful sacrifice that actually yielded positive WPA. Gerardo Parra dropped one down in a tie game in the 12th inning with no outs and runners on first and second. A base is never worth an out, in terms of WPA, but two bases can be, and in this case, the gamble paid off, with both runners scoring on Aaron Hill’s subsequent single in Arizona’s 4-2 win over the Dodgers.

All told, the ten bunts yielded 29 points of WPA, an average of almost 3% improvement in the bunting teams’ chances to win their respective games. Seven of nine runners moved up by those bunts went on to score, and eight of ten bunters’ teams won that day (teams went 5-2, but three winners laid down two bunts apiece). Excluding the three bunts by pitchers, win probability improved by an average of more than five percent and all but one advanced runner scored.

Saturday was a relatively uneventful day in the world of bunts, but Sunday was a disaster. After Ben Revere singled on a bunt off Boone Logan, the last 12 bunts of the day produced negative WPA, totaling -46 percent. Marco Estrada bunted into an inning-ending double play against the Pirates. David Hale hit into a forceout against the Mets. Josh Collmenter struck out trying to bunt against Josh Beckett and the Dodgers. Gerrit Cole popped one up against the Brewers. Even the position players struggled with bunts, as Carl Crawford and Curtis Granderson grounded out via bunt (a .333 daily average on bunt hit attempts is no disaster, but bunters have batted .493 in non-sacrifice situations in 2014). Of the five sacrifices executed on Sunday, only one runner came around to score.

Bunting is a fickle pursuit within a fickle game.

Let’s take a look at the seven questions I posed earlier this month and see what kind of results the year’s first 265 bunts have yielded. Over/unders were based on rates established over the first three days of baseball games.

**What percent of bunts will be attempted by pitchers? (O/U 62.5%)**

Way under. 113 of 265 so far, or just 43%. Pitcher bunts have averaged a WPA of negative 2.5%, while position player bunts have averaged a positive 0.6% swing.

**What percent of bunts will come outside of traditional sacrifice situations (fewer than 2 outs, runners on first, second, or both)? (O/U 12.5%)**

Way over. 67 of 265, or 25%, were clearly intended to put the batter on base, and more than a few of the 198 that have come in sacrifice situations certainly weren’t intended to give up an out. More on that later.

**What percent of sacrifice attempts will successfully advance the runner(s)? (O/U 43%)**

Way over. This was an easy one. 67% of sacrifice attempts have moved the runner(s). That adds 107 sacrifices to the 25 batters who reached base and moved the runner(s).

**What percent of all bunters will reach base? (O/U 25%)**

Under (21.9%). 25 of the 198 sac bunters reached, in addition to 33 of the 67 who bunted for hits. That .219 on base percentage on all bunts is far short of the league’s .311 figure, but position players have reached base at a .362 clip when bunting, vs. .319 in all situations. It’s the pitchers, who often use the bunt as a means of helping create runs without having to run the bases, who are dragging this number down, reaching only 3 times on 113 bunts.

**What percent of runners advanced via sacrifice will score? (O/U 25%)**

Over, but not by as much as you might think. 57 of 128, or 45%, of runners advanced by sacrifice have scored. If we add in runners advanced in sacrifice situations in which the batter reached base, we’re up to almost half (77 of 155). If only we knew how many of those runners would have scored if the batter had swung away. At least we have WPA…

**What percent of total bunts will lead to an increase in win expectancy (O/U 25%) and what will be the cumulative WPA of all bunts?**

Over, barely. 69 of 265 bunts, or 26%, have resulted in positive WPA. Another 18 have rounded to 0, while the other 178 have reduced win expectancy. Total WPA has been -192, or -0.72% per bunt. As noted above, this is heavily influenced by pitchers. Position player bunts have averaged 0.61% WPA, which doesn’t sound like much, but the average plate appearance by a non-pitcher has yielded a 0.04% WPA.

That the average position-player bunt has aided win probability more than the average position player non-bunt doesn’t necessarily mean teams should be bunting more, as more bunting might lead to defenses better situated to turn bunts into outs. But it does show that NL teams have had some success with the bunt in 2014, and has runs become a scarcer commodity, it might not be a bad idea for some light hitters to drop a few more.

**Subscribe to: ** RSS feed

Bryan, just an awesome read. I appreciate the tone you put in as you go through this. It’s far from a dry read. Keep at it!

I’m trying to wrap my mind around this.

Do we know what the WPA is for a typical pitcher who hits away in what would normally be a sacrifice situation? I understand that your results showed that pitchers bunting in a sacrifice situation resulted in a negative WPA but is it possible that it’s a LESS negative WPA and therefore a net positive?

I’ve never had a problem with a batter attempting to bunt for a hit- although I am surprised at the success rate- but I might have to rethink my views on non-pitchers sacrificing if your findings are indicative of typical results over the long-term. That’s at least if I’m understanding them correctly.

Pitcher bunts have averaged a WPA of negative 2.5%I would agree with this statement only if pitchers hit as well as an average batter. Since they don’t and their batting effectiveness is much more than 2.5% below the average hitter, then I would say that pitcher bunts are actually quite positive, relative to what is likely to happen otherwise (striking out mostly, or hitting into DPs).

WPA of -2.5% doesn’t mean 97.5% as well as an average batter. It means losing 2.5% game winning chance in that plate appearance relative to league average.

So you’d have to know how much a typical pitcher gives up in a PA versus a league average hitter. Looking at some active pitchers, it looks like Bref gives them roughly 1 Rpos per 7 PAs, which should be a rough indication of how much the average pitcher gives up versus a league average hitter. about 1/7 of a run per PA. Currently runs convert to wins at about 10 to 1. So that’s about 1/70th of a win or 1.42%. So that suggests pitchers have done considerably worse bunting than their average so far this season. That said, I’m using a rough proxy based on other years. To really get a better idea, you’d have to look at the rBat for all pitchers non-bunt PAs and see how much they are giving up this year.

Thanks for the estimate, Michael. That was, indeed, my point – that the -2.5% was how much the bunt play cost compared to what an average hitter might have done in that spot, NOT compared to what that pitcher might have done hitting away.

It always depends on the situation, of course, but, as a general rule, if I have a pitcher who can’t hit a lick but can lay down a bunt, I’ll take that base for an out, rather than just the out.

Interesting take, Michael.

If I’m interpreting fangraphs right, pitcher WPA has actually been worse when bunting. Using fangraphs’ leaderboards, selecting “League Stats”, filtering for NL and pitchers, total WPA to-date is -9.98.

Over 835 PA, that’s an average of -1.2%, significantly better (or just less awful) than the -2.5% they’ve incurred when bunting. This is probably biased by leverage, since most bunts come in higher-leverage situations. For instance, in a scoreless game in the third inning, if a pitcher comes up with no one on, win probability is probably around 50% and will probably remain there if he strikes out. In contrast, if he comes up with a runner on first, WP is probably up a few points above 50%, but giving up an out, whether via K or sac bunt, drops it back closer to 50%.

If any play index wizard can identify situations in which pitchers have come to the plate in traditional sac bunt situations and swung away, I’d be interested to see what an average WPA is. I’d guess, as Doug deos above, that it’s worse than 2.5%.

You need to use RE24 rather than WPA for this type of analysis.

Here are the YTD results for 2014.

- 187 times with pitcher batting, 0 or 1 out, runner on 1st, runner on 2nd, or runner on 1st and 2nd.

- 53 times didn’t bunt with WPA of 0 (exactly) and RE24 of -0.27

- 12 times with 0 out, 41 with 1 out

- 25 times with 2 strikes, 9 with one, 19 with none

- 14 strikeouts, 19 batted ball outs, 4 FC, 3 ROE, 1 HBP, 1 BB, 11 singles

So, that’s a .216 BA indicating that managers are choosing this option with better hitting pitchers.

Here’s the list.

http://bbref.com/pi/shareit/ectKg

Thanks, Doug. I wonder if the better-than-expected pitchers’ batting average is a result of managers only allowing better hitters to swing away or that all the pitchers in your sample were throwing from the stretch, potentially distracted by baserunners.

Mosc, I’d like to understand more why RE24 is a better metric for this analysis. That RE24 ignores inning and score actually seems detrimental, since the bunt is a one-run strategy, which makes more sense in the late innings of close games, when WPA is likely to swing more.

I mentioned above that leverage could cloud a retroactive comparison of all bunts by pitchers to all non-bunts by pitchers, but I don’t see how RE24 would compare them more accurately or relevantly. Since these events tend to take place in different base-out situations, it seems that RE24 would be subject to the same bias. What am I missing?