What to look for in a good reliever



Cards' fans had to look at a scene like this -- from Holland and many others --- far too often in 2018.

(This is the 2nd part of a multi-part series on the bullpen and the steps that the Cardinals need to take to fix it.)


A couple of days ago, I took a look at the free agent options that were available last offseason and how they performed this season. The Cardinals, of course, signed Bud Norris, Luke Gregerson, and Greg Holland but had lots of other choices. Those other guys were a bit of a mixed bag; some pitched well, some were awful (sort of like the 3 the Cardinals signed).


This post isn’t really going to be a Cards-centric post at all. What I want to do here is look at all of baseball’s relievers to figure out what characteristics the Cards’ front office needs to target when looking to add some bullpen pieces this offseason. It’s easy to look at stats like ERA, FIP, or WAR but all those are really a result of some underlying peripherals. It’s the peripherals I want to look at today. What do good relievers do well?


It seems to me that good pitchers might do at least 1 of 4 things well – keep batted balls on the ground so that fielders can make outs and the ball can stay in the yard; force soft contact from hitters in order to reduce the number of extra-base hits; keep batters from making contact with the ball by striking them out; and minimize the number of batters who reach base by limiting walks. It’s tougher to score runs if batters don’t hit home runs and if batters are pounding the ball into the ground. A high ground ball rate might be doubly important now that there is such an emphasis from hitters on elevating the ball. It’s also difficult to hit home runs and extra-base hits if they’re not making hard contact on the baseball. On the other hand, if hitters never make contact at all and have to U-turn and head back to the dugout they’re not going to be very successful at the plate. Finally, if batters are routinely behind in counts and aren’t able to reach base via the free pass, it’s going to be more difficult to score runs.


The 3 peripheral statistics I wanted to examine more closely, therefore, are relievers’ ground ball percentage, hard hit rate, and strikeout minus walk percentage. I wanted to examine these peripherals to determine to what degree there was a correlation between each of them and the pitchers’ fWAR and WPA. Though WAR is certainly not an exact science – and it’s even less exact for relievers – it is reasonable to believe that, generally speaking, relievers who have a higher WAR pitched better than those with a lower WAR. A pitcher’s fWAR is based on his FIP and includes the leverage index of his performances. Pitchers, therefore, who strike out more, walk fewer, and give up fewer homers in tougher situations will have a higher fWAR than those who don’t do those things.


WPA (win probability added) is a statistic that isn’t discussed as often as WAR is but it can often be a pretty good way of evaluating a reliever’s performance. WPA measures and sums the amount by which a team’s probability of winning a game increases or decreases while the pitcher is in the game. When relievers perform well their team’s WPA is likely to increase and when they pitch poorly their WPA is likely to decrease. For example, if a closer enters the game when his team has a 91% chance of winning the game and he successfully closes the game then his WPA for the game is going to be +0.09 (9%). Add all that up throughout the season and that will tell you the player’s WPA for the season. If that closer blows that game, his WPA for the game will end up being -0.91. Relievers, thus, with higher (and positive) WPA’s had a lot more good outings than bad ones over the course of the season.


Neither WAR nor WPA are perfect measures of how relievers performed but if you look at the Fangraphs leaderboard for relievers for both statistics, it’s a who’s-who of the best relievers in the game. Clearly, the Cardinals’ pen is going to be better if it has more guys with high WAR’s and high WPA’s. The point of this exercise then is to figure out what kind of relievers are more likely to end up with high WAR’s and high WPA’s.


First, a caveat…anyone with any familiarity at all with statistics knows that correlation does not necessarily imply causation. What we’re going to be looking at here is the correlation between the underlying peripherals and WAR and WPA. So we will be able to see if a relationship exists but without running a regression we’re not going to be able to definitively assert that the peripheral stat causes the player’s WAR or WPA.


For my methodology, I looked at the performance of all of baseball’s relievers who threw at least 40 innings this season. This was a total of 183 relievers. I didn’t total all the innings but it’s reasonable to assume that we’re looking at something in the neighborhood of 10,000 innings or so. I then created scatter plots to see the relationships between these variables. A strong correlation between the should give one the ability to place a best-fit line through the points. A weak correlation should appear to be just a big jumble of dots with no discernible pattern.


So let’s first examine the relationship between ground ball percentage and WAR and WPA. The logic again is that if relievers get a lot of ground balls they keep the ball in the yard and give their defense a chance to make plays. The Cards are well-known for valuing ground ball pitchers and we all know by now how valuable a double play can be. The idea here would be that a high GB% correlates with a high WAR and a high WPA so if there’s a correlation between the 2 we should see the makings of a clear upsloping line.



It’s obvious from looking at both of the scatter plots comparing GB% to WAR and WPA that there’s just a big jumble of dots. There is clearly no discernible pattern here. This tells us that some pitchers who get a lot of ground balls pitched well in 2018 and some did not. Some of those who pitched especially well had a much lower GB% and so their success must have been attributable to something else. A high ground ball percentage, then, may help a pitcher perform well but it’s certainly not the secret to success.


Now let’s look at hard hit rate. It stands to reason that pitchers who induce soft contact are less likely to give up extra-base hits and, therefore, runs. Pitchers who pitch well should, if hard hit rate tells us what we think it might, have a low hard hit rate. Now we’re looking for an inverse relationship between the 2 variables which should show us a pretty clear downward-sloping best-fit line.



A strong correlation between the 2 would be implied by a best-fit line that slopes from the upper-left corner to the bottom-right. Is that what we see in the 2 graphs? It certainly doesn’t appear so. Again, it appears as though we’re looking at a big jumble of dots. It just doesn’t seem as though we could draw a very good best-fit line through that scatter plot. This implies that there’s not much of a correlation between hard hit rate and a pitcher’s WAR or WPA. Why is that? I’m not sure but I would hazard a guess that there must be something that is a lot more closely aligned with a pitcher’s success than the hard hit %.


Finally, let’s look at K-BB%. Pitchers who score high here will strike out a lot of batters and walk very few. Clearly, if a pitcher strikes out a lot of hitters, he’s going to be tough to score against. A low walk percentage means that batters are going to have to hit their way on base and it also implies that the pitcher probably is in fewer hitters’ counts. If there is a strong correlation between K-BB% and WAR or WPA we should see a strong, upward-sloping line and, preferably, with both statistics.




Now those 2 scatter plots seem to be able to accommodate a pretty solid best-fit line. It’s pretty clear looking at those 2 graphs that there is a relatively strong correlation between K-BB% and WAR and WPA. The fact that that the peripheral stat is correlated with both WAR and WPA implies that there is some validity to each correlation as well. If K-BB% showed a correlation with just one, that would be pretty strange. We should expect it to either be correlated with both or neither. Thus, relievers who have a high K-BB% tend to also have high WAR’s and WPA’s. There’s some logic there.


The increase in strikeouts is one of the most noticeable changes in the game over the last few seasons, especially from relievers. The best teams seem to have several guys who are able to mow hitters down, one after the other, in the late innings. They don’t allow unnecessary base runners by walking them and are able to get their teams out of jams by striking out opposing hitters. This shouldn’t really surprise anyone but it is an important confirmation of what our eyes seem to tell us.


In the next part of our series we’re going to look at how the Cards’ relievers fared by K-BB% this past season. Since the pen as a whole performed poorly we would expect to find that most of them also fared poorly by K-BB% and then we will use this as a guide to inform us as to which relievers the team might be able to count on in 2019 and which ones should go the way of Greg Holland.


Thanks for reading.


The great – and very appropriate – Greg Holland pic comes courtesy of the great and very appropriate @cardinalsgifs.


All the stats here come courtesy of Fangraphs