The Twins have said goodbye, for the moment at least, to leftfielder Eddie Rosario, who led the team in rbi’s the last two or three years. It used to be thought that the rbi was the statistic that best indicated a team’s most productive hitter, but not only did the Twins show little uncertainty in releasing Eddie, but no other Major League team offered to pick him up at his projected salary of $10 million. I can’t help but think this means that the rbi has been severely downgraded in this new age of analytics, and I can guess why.
Let’s look at the variable factors that may contribute to rbi totals. How many games did he play? Where in the lineup did he hit? The more at-bats, the more potential rbi opportunities. We can account for this by dividing rbi’s by number of plate appearances. Not “at-bats,” because a sacrifice fly, squeeze bunt (rare), or bases-loaded walk or hit-by-pitch can produce an rbi without an official at-bat. But not all ABs are created equal. Obviously, someone hitting cleanup after three .300 hitters will have more chance at rbi’s than the number 8 hitter on a weak-hitting team. Maybe a rough corrective would involve factoring in a team’s total runs, so that an rbi for a weak offense will count more than one for a powerhouse, such as the Twins were in 2019 when they set the all-time home run record. I’d like to burrow down more deeply, though, and I wonder whether modern analytic data-collection can go this far.
Every rbi could be categorized for the situation it occurred in: the number of outs and the number and locations of base runners. If a hitter bats in the number of runs in a given situation that the average of all hitters did in that situation in, say, the previous year, his rbi average for that situation, in my proposal, is 1.00. Twice as many, 2.00. Half as many, .500. Computing the average of all the situations he faced would produce his overall rbi average. This would give us apples v. apples comparisons, not apples and oranges. To spell out one example: runners on second and first, no outs would be one “situation.” Bases loaded, one out, another. I would break the situations down by outs for the simple reason that the third base coach is more likely to send a runner home if there are two outs than no outs, so it’s more likely to result in an rbi. And of course if there are two runners on, rather than one, there is the chance for more rbi’s. I don’t know how many permutations there are, but it’s closer to 9th-grade math than infinity.
Is there any need to weigh the scores from the different situations differently, given that the rbi probability is much greater if the bases are loaded than if they are empty? Not at this point in my analysis. For our purposes, a run is a run, and we’re comparing one player against the league average. We’re looking for someone’s value as a run producer, whether in the 1st or 9th inning. But once we’ve given a player his rating for each possible situation, we can massage these data for other purposes, and the one that comes first to mind is “clutch hitting.” MLB tried to quantify this with something called the “game-winning rbi” in the 1980s. It died of its own uselessness because it couldn’t distinguish, for instance, between a bloop single in the 1st inning of a 9-0 blowout and a 9th-inning grand slam in a 4-3 squeaker. So, rather obviously, there are two new factors to consider if we are to put a value on the rbi: inning and game score.
Here the permutations and combinations are pretty close to infinite, so I’ll propose a different test: just as we now have a “quality start” for starting pitchers (a stat with its own problems), we should have “quality rbi’s.” A quality rbi is any rbi from the 7th inning on, or in the last inning of a rain-shortened game, in which the hitter’s team is tied or one run ahead, or results in his team’s being one run behind, tied or ahead. For this statistic, it doesn’t matter where in the order a batter hits. It does somewhat matter how good his team is, for the better team will have more runners on base and, thus, more chances for an rbi. Given the arbitrary nature of this statistic, however–unlike the more pure rbi average, above–I’ll let this unaccounted variable slide. Compared to the nothing that is here now, it is a meaningful marker. And this variable is less of a factor here than in such other recognized statistics as wins for a pitcher.
One last word on rbi’s. At some point in the not-too-distant past, the accepted abbreviation of “runs batted in” became “rbi,” instead of “rbi’s,” presumably on the theory that the plural “s” comes after the “r” for “runs,” not the “i.” I have always considered “rbi” an entity of its own, capable of taking an “‘s” after the entire abbreviation. I still do.