World Series Reveals Major Flaws in Third-Time-Through-Lineup Theory

An amazing thing happened in the last few World Series games. Boston manager Alex Cora and Dodgers manager Dave Roberts – handpicked by devoted sabermetrically inclined front offices – surprised many by flouting a firmly established but misguided analytic rubric.

It happened in Games 3, 4 and 5 when both managers tossed the “third-time through a lineup” rule out the window to let their starting pitchers go deeper into games than has been typical lately. In the Dodgers’ Game 3 win, Roberts stayed with Walker Buehler for 108 pitches and a full seven innings. Then in Game 4, Roberts let Rich Hill go beyond the 18-hitter mark to face 24 hitters over 6.1 innings, and then he was rightfully ripped for not letting Hill go further while working a 4-0 shutout.

As for Cora, in the series-clinching Game 5, he stuck with David Price into the 8th inning to face 25 hitters. The distances these pitchers went were fairly modest by historical standards, but rather unique in the current era when more managers and analytics departments have shifted more playoff innings to relievers. In fact, this year’s postseason saw playoff bullpens handle more than 50 percent of all innings for the first time ever. And this was even before Friday’s 18-inning marathon.

Consider these recent early-hook playoff examples:

• In Game 2 of the 2017 World Series, Roberts relieved Hill after exactly two times through the lineup. Hill trailed just 1-0 after four innings, while striking out seven and surrendering only three hits. Roberts went with the “fresh reliever” approach, using a succession of eight relievers to ultimately lose 7-6 in 11 innings and create a forever “what if” for Dodgers fans.
• In Game 7 of the 2016 World Series, the Cubs led 5-1 midway through the bottom half of the 5th inning with a very depleted bullpen. Although starter Kyle Hendricks had thrown only 63 pitches to just 19 hitters, Joe Maddon pulled that year’s NL ERA leader. The move nearly cost the Cubs the title, but the team’s hitters would rally to prevail 8-7 in 10 innings.
• Milwaukee’s 2018 rotation threw the fourth-fewest regular-season innings in the NL. So, very much by design, Craig Counsell gave his bullpen a whopping 61 percent of NLCS innings (not including Brandoff Woodruff’s 5.1 innings of “trick relief” in Game 5). This was probably the purest application of the “third-time-through theory” we will ever see, but it still seemed a minor miracle the Brewers fell one game short of the World Series.

I’ll grant Counsell’s bullpen strategy flowed at least partly from necessity given his poor rotation. (However, the undervaluing of innings-eating starters by analytics may also have led the Brewers’ front office to not properly invest in their rotation last winter.) In contrast, most other playoff teams willingly shifted large numbers of playoff innings from quality starters to less-talented relievers. In the process, as the data below indicates, they actually lowered their playoff-win probability, especially in longer playoff series.

Now, the third-time-through trend is not based on some crazy hypothesis. Regular season data does show hitters post an .800 OPS in their third at-bats against a starter, as compared to a .720 OPS against an “average” fresh bullpen arm. But some apples-to-oranges caution is recommended here. These averages were generated under regular-season conditions. Thus, applying their conclusions to very different playoff conditions is highly flawed.

For one, the principle edge relievers enjoy during the regular season is never facing the same hitter twice in a game and rarely ever more than two times in the same series. But given five- and seven-game playoff formats, the better relief pitchers lose this element of unfamiliarity by regularly throwing four or more times against the same team.

Take the 2017 World Series. Three Astros hitters faced All-Star Dodgers closer Kenley Jansen five times each. Not surprisingly, those hitters in their third, fourth and fifth at-bats against Jansen produced a combined slash of .285/.444/.714 for a 1.158 OPS. Further Jansen’s pen-mate Brandon Morrow appeared in all seven World Series games. This included pitching – with similarly poor results – to one Astros hitter five times and two others four times each.

This erosion in reliever effectiveness during the playoffs is only logical. As the playoff workload mounts on over-used relievers, their sharpness and effectiveness fade. Then factor in these two-pitch relievers facing the same hitters three and four times in a series. So it should be no surprise these relievers’ playoff effectiveness trails off worse than a quality starting pitcher trying to buck those “three-times-through” odds.

Over the past four seasons 14 relievers (see end of story below) appeared in at least five games in a playoff series and/or faced more than 27 batters in a series (the equivalent of three times through an order). This totaled 79.1 innings and 329 hitters faced. Here are their combined success rates by times facing the same hitter:Their first and second times facing a playoff hitter provided strong results, far better than the .720 OPS against by an average “fresh” reliever in the regular season. This is as it should be, as relievers in this group are far from an average regular-season reliever. A majority of them were elite, combining for a 2.33 ERA and 11.5 K/9 innings during the regular season

But elite as most were, their collective third time facing the same playoff hitter produced an OPS against of .804. This was no better than an average regular-season starting pitcher attempting a third time through a lineup (.800).

Keep in mind that playoff rotations aren’t average, either. Most playoff starters are top tier in baseball, thus their OPS against that third time through should be significantly lower than the “average” .800 mark. So statistically speaking, removing a quality starter early – like a Hill, Hendricks or Rick Porcello – and transferring those innings to a reliever with whom the opposing lineup has grown familiar actually improves the opponent’s chances for success, not lessens it.

In the last four years, the number of high-use relievers in a playoff series has increased each year. (Again, see list at end of story.) But if application of the “third-time-through” theory to the playoffs is so flawed, why are so many playoff managers continuing to use it?

Part of the reason may be selection bias. Most modern front offices embrace sabermetrics and have made it the emblem of their generational change. But as often happens when gen-gaps develop, a single trait can become over-embraced almost to the point of caricature, resulting in bad decision-making.

Further, when hiring new managers, front offices make it a prerequisite that winning candidates readily collaborate with and embrace the strategic suggestions generated by a team’s data department. Thus it is not in the nature of modern managers to push back against the assumed wisdom of their bosses’ numbers. Plus strange job security exists in failing collectively. Thus taking the risk of daring to break from the organizational crowd is less likely for newer, younger managers.

But both 2018 World Series managers actually produced the first cracks in the unquestioning over-devotion to “the numbers guys.” This is not to say most statistical analysis is wrong, only that healthy skepticism toward less questioned group think can be a wise move.

Alas, this realization dawned on Dave Roberts too little, too late to benefit him in either the 2017 or 2018 World Series. I doubt the Dodgers will part ways with him. He was, after all, just following their data, but after two consecutive runner-up finishes, it’s hard to see them mounting a third consecutive title charge. At least not without a major turnover of some of their wearied arms (Jansen, Clayton Kershaw, Hill, Madson) and wearying personalities (Yasiel Puig, Manny Machado).

As for Cora, all is forgotten in victory. Thus his mistakes got less scrutiny. However, he at least appeared to try to adjust on the fly. For example, he made a very poor decision in Game 3 to go by the (regular season) computer and lifted Porcello after exactly 18 hitters. The next night, Cora tried to correct for this, but chose to do so by putting too much faith in the far less effective and experienced Eduardo Rodriguez to go deep into Game 4. Good idea, wrong situation.

But perhaps the 2018 playoffs provided enough deviations from third-time-through strategy to convince more future playoff teams to limit application to more average starters, like the Brewers rotation and Rodriguez. Maybe when starting an elite pitcher who is dealing, we’ll see more managers just ride him and tell the paint-by-number crunchers to stuff it.

An ideal playoff strategy is to pitch your best starters as deep as practical, avoid throwing any one reliever more than four times in a seven-game series, and limit the number of times a reliever (other than a lock-down closer) faces any hitter more than three times in a series. In this way, playoff teams can shift a majority of innings back to their rotations and retain their bullpens’ regular-season advantage of unfamiliarity and freshness.

* Playoff relievers appearing at least five times in a series or facing at least 27 hitters:
2015: Kelvin Herrera (Royals), Addison Reed (Mets)
2016: Aroldis Chapman (Cubs), Mike Montgomery (Cubs), Andrew Miller (Indians)
2017: Jansen and Morrow (Dodgers), Chris Devinski (Astros), Carl Edwards Jr. (Cubs)
2018: Ryan Madson (Dodgers), Dylan Floro (Astros), Josh Hader, Jeremy Jeffress and Corey Knebel (all Brewers)