Two weeks ago, I wrote an article about how the Astros shouldn't shackle Ken Giles to the closer role and instead should make him the league's first true "high-leverage reliever". Rather than force him to wait for a lead in the 8th inning or later, Houston should turn to Giles when "the opposing team's chance of scoring is a near certainty", although there are clearly caveats to this statement.
In the article, I detailed three instances in which Giles could have been brought into the game either to keep the score close or preserve the lead for the Phillies but instead was left to watch in the bullpen as his teammates let the game slip away. What I didn't do was suggest a concrete way to implement this plan, which in other words is known as whining. To make up for that, below are two methods that could be used together or separately to use Giles and other high-leverage relievers optimally.
If a team is serious about revamping their bullpen usage/hierarchy, leverage index is a fantastic place to start.
"During the course of a game, some situations are more tense and suspenseful than others. For instance, we know that a one-run lead in the bottom of the ninth inning is more suspenseful than a one-run lead in the top of the third inning. Batting with two runners on and two outs in the eighth inning is filled with more pressure than batting in the same situation in the second inning. Leverage Index (LI) is merely an attempt to quantify this pressure so we can determine if a player has been used primary in high-leverage or low-leverage situations."
On the surface, LI seems like the perfect statistic for pushing forward the evolution of the bullpen. By measuring the "pressure" of the situation, a manager could hypothetically use that to determine which reliever to bring in. While this is undoubtedly true, there's one issue that hurts LI's chances of being immediately adopted by baseball's gate keepers.
As FanGraphs' explanation of LI alludes to, not all run scoring chances are created equally, which inherently affects the statistic. For example, logically, a team down three runs in the 6th inning has a greater chance of making a comeback than a team down three runs in the 7th inning or later. However to a manager, and the entire team, three runs is three runs no matter what the inning is, and three runs is not an insurmountable deficit.
Unfortunately LI doesn't express that same optimism. As a result, if a team is already losing and multiple runners are on base, the LI would be low despite the high chance of scoring runs because of the score. To help illustrate this idea, the table below shows the 7th inning of a game between the Phillies and Nationals in which Washington was up 4-1.
|Jeanmar Gomez||Dan Uggla||0||4-1||Uggla flied out||0.19|
|Jeanmar Gomez||Ian Desmond||1||4-1||Desmond singled||0.14|
|Jeanmar Gomez||Jayson Werth||1||4-1||Werth flied out||0.25|
|Jake Diekman||Bryce Harper||2||4-1||Harper walked||0.19|
|Jake Diekman||Ryan Zimmerman||2||6-1||Zimmerman doubled||0.37|
|Jake Diekman||Wilson Ramos||2||7-1||Ramos singled||0.07|
|Jake Diekman||Danny Espinosa||2||7-1||Espinosa flied out||0.02|
Heading into the 7th inning, the Phillies were down three runs already, but by the end of it, they were down six (7-1). Philadelphia eventually lost 7-2, but had Giles or another high-leverage reliever been brought in during that 7th inning, the score might have been closer and more manageable to overcome. Herein lies the problem with LI. If a manager chose his relievers strictly by the LI score, a team down late in the game would never use a high-leverage reliever. Giles would still be sitting in the bullpen waiting for the game to get closer. Because of this, LI can't be the end-all-be-all stat to determine which reliever should be used. Fortunately, run expectancy can help.
RE (run expectancy) is a fairly simple concept to understand. Below is a matrix of the base-out states.
"A run expectancy matrix presents the expected number of runs scored between a given point and the end of an inning based on the overall run environment, the number of outs, and the placement of the baserunners. For example, in the RE matrix below (run environment set at 4.15 runs per game), the expected number of runs given a runner on first and no outs is 0.831 runs."
|Runners||0 Outs||1 Out||2 Outs|
|1 _ _||0.831||0.489||0.214|
|_ 2 _||1.068||0.644||0.305|
|1 2 _||1.373||0.908||0.343|
|_ _ 3||1.426||0.865||0.413|
|1 _ 3||1.798||1.140||0.471|
|_ 2 3||1.920||1.352||0.570|
|1 2 3||2.282||1.520||0.736|
By using run expectancy independent of LI, we can see what the average number of runs a team will score is, given a specific situation. While we inherently understand that runners on first and second with nobody out is a tough situation for a pitcher, through the run expectancy matrix we can see just how perilous it is. By putting a specific number on it, perhaps managers would be more inclined to bring in a high-leverage reliever, even if it's only the sixth inning.
On August 18th, the Phillies played the Blue Jays, and after five innings Philadelphia was leading 5-3. Through poor bullpen management, that two-run lead quickly became a three-run deficit.
|Elvis Araujo||Ben Revere||5-3||0||Revere walked||0.5||1.23|
|Elvis Araujo||Cliff Pennington||5-3||1||Penninton grounded out||0.89||2.13|
|Jeanmar Gomez||Chris Colabello||5-4||1||Colabello singled||0.69||1.71|
|Jeanmar Gomez||Troy Tulowitzki||5-4||1||Tulowitzki singled||0.53||1.94|
|Jeanmar Gomez||Josh Donaldson||5-7||1||Donaldson homered||0.92||3.22|
|Jeanmar Gomez||Jose Bautista||5-7||2||Bautista flied out||0.27||0.49|
|Jeanmar Gomez||Edwin Encarnacion||5-8||2||Encarnacion homered||0.11||0.33|
|Jeanmar Gomez||Russell Martin||5-8||3||Martin grounded out||0.11||0.22|
Starting with Revere, Toronto's run expectancy was over 0.50 for the first five batters and was near one on two separate occasions. In this case, the LI was incredibly high, as the Phillies began this inning with a lead. Given the context of the situation, this was the most important moment of the game. Instead of turning to Giles, their elite reliever, Jeanmar Gomez was brought in to "save" the game.
Giles was eventually called, but the score was 8-5 in favor of the Blue Jays. His contributions (a scoreless inning) were too late.
Thanks to the Royals' super bullpen that helped lead them to an AL pennant in 2014 and a World Series championship in 2015, teams across the league are attempting to copy their blueprint. The Yankees have what could be the best bullpen of all time with Dellin Betances, Andrew Miller, and Aroldis Chapman and will likely be able to use that trio in a way previously unseen.
With three fantastic relievers that could all be closers, Joe Girardi can afford to use one of his high-leverage relievers in innings before the 8th and 9th. Betances could be deployed in the sixth inning to make sure the Yankees hold their lead or keep the game close while maintaining a shutdown reliever for the later innings.
The bullpen is no longer viewed as just a place where failed starters go to keep their careers alive. The baseball world has shifted dramatically and now understands how vital a bullpen can be to the overall success of a team. While we're undoubtedly moving in the right direction, there's still a much needed evolution for how relievers are utilized. By using LI and RE for in-game decision making, teams can maximize their assets and reach the full potential of their bullpens.
. . .
Matt Goldman is a Featured Writer with Beyond the Box Score and a Contributing Editor at MLB Daily Dish. You can follow him on Twitter at @TheOriginalBull.