Consistency is important in many disciplines, but for this article, I’m writing exclusively about baseball pitchers.
ERA (Earned Run Average) is one of the most common metrics of a pitcher’s individual performance. It is defined as the number of earned runs per 9 innings. In others, a pitcher who gives up 1 run an inning will have an ERA of 9. By way of reference, an ERA below 4 is good, and an ERA below 3 is excellent. An ERA above 5 is bad. An ERA of 9 is beyond bad - it would be among the worst in baseball history.
A guy, let’s call him Neil
Our hypothetic dude ‘Neil’ is a major league relief pitcher. He consistently gives up exactly one earned run every complete inning he pitches. This gives him an ERA of 9.00. By one of the most common metrics we have, Neil is awful, possibly the worst pitcher in the history of baseball. However, such a pitcher would be one of the most respected and feared assets in baseball. Anytime his team reaches a state where their lead (in runs) exceeds the the number of innings remaining to be played, Neil can earn them an automatic win. As a side note, Neil would never actually earn a win for himself if used this way.
Mitigating Factors
There are a few minor problems with this hypothetical. First, it doesn’t account for unearned runs. Though, these are comparatively rare on even a mediocre defensive team, so I feel justified in ignoring them. Secondly, it doesn’t account for inherited runs - this is why I specified “complete innings”. Thirdly, he likely will have at least one decent looking statistic: saves. The majority of his appearances would almost certainly be save situations, and if used properly, he would never blow a save.
Making a Statistic to Capture Consistency
Admittedly, such a statistic might exist without me knowing. However, if one doesn’t exist, I propose it should. A naive approach would be to use the standard deviation of ERA/Game. The main issue is that it’s a hassle to calculate in that it requires individual game data.