In 2017, the 84th best pitcher averaged approximately 10 runs prevented, establishing the 90th percentile performance point; in 2018, the 89th best pitcher averaged approximately 10 runs prevented, suggesting that the 90th percentile performance did not change. If you like false certainty, even here the decimal points look similar, as the 2018 cut-off was 10.221 average runs prevented versus 10.229 average runs prevented in 2017. Let’s call it 10 Average Runs Prevented for fun…

 Exploring Runs Prevented || Aces Do Not Exist || Rotation Spots || Third-Time Charmers

What is Runs Prevented? Runs Prevented is a relatively basic statistic that attempts to measure the quality of a pitcher’s performance within their run environment (which is roughly the combination of their ballpark, which can impact scoring, and their league, which can impact the mix of talent faced). Measuring pitching is difficult because unlike batting stats, you cannot simply “count up” with pitchers; a team that scores 750 runs is  likely better than a team that scores 700 runs, but a team that allows 650 runs is likely better than a team that allows 700 runs. So, assessing pitching in a “run environment” requires accounting for that “negative space” (in this example, between 650 RA and 700 RA). Throughout the 2018 season, I developed an Average Runs Prevented statistic that attempted to track variance in park factors (between Baseball Prospectus Pitching Park Factor and Baseball Reference Park Factors) as well as variance in league environments. Usually these differences did not amount to much, but tracking them is an important step to understanding variance and reflecting uncertainty in the Runs Prevented metric: I can say “Jeremy Jeffress prevented 24.63337 runs in 2018,” or I can say “on average Jeremy Jeffress prevented between 23 and 26 runs in 2018,” or I can say, “Jeffress prevented 25 runs.”

…The short lesson: if you’re an MLB pitcher that prevented 10 runs, you’re really, really good. This holds regardless of role. In 2018, the top ten percent of all MLB pitchers featured 53 starting pitchers and 36 relief pitchers, which is quite a swing in favor of starting pitching quality; in 2017, these 90th percentile pitchers featured 41 relievers and 43 starters. What’s important here is to takeaway that great relievers prevent enough runs to break into the top threshold of league performance; Runs Prevention need not categorically be dominated by starting pitching, which is one reason the Brewers did not need a starting pitcher at the trade deadline (thanks to their strong bullpen).

Aces are more difficult to define than “Average Runs Prevented.” I can at least give you a math equation for Average Runs Prevented at the end of the day. But an “ace” can be many things;

• Some fans want an ace to be a breathy “dude” or “guy”, a “stud” atop the rotation. File this one under the debate about playoff dudes,  “Do you want Jhoulys Chacin and Wade Miley to pitch in the playoffs?” (Yes, actually);

• Some people ascribe to a theory than an “ace” is simply the best pitcher on each MLB team (e.g., 30 teams means 30 aces, and no team can have more than one ace);

• Scouts like to define “aces” in very rigorous ways, focusing on the quality of a pitcher’s stuff (typically better than average fastball and off-speed or breaking ball is the minimum stuff requirement for acehood), as well as their command (must be great), and probably their frame, too (it’s good to project innings from an ace). But I take it that this is not how most fans mean “ace,” and it’s also no fun because aces rarely exist under this mold (#EveryoneIsAMidRotationStarter);

• Finally, there’s the simple stipulation that an “ace” is an elite starting pitcher, at the top of the league in terms of performance (presumably, hopefully, for multiple consecutive years. For example, this is why people say Clayton Kershaw is an ace, but Junior Guerra is not. I used to be sure of the importance of consistency, but….well, aces hardly exist under this requirement, either).

In this analysis, I am going to demonstrate two of the shortcomings with the idea of using performance to define the concept of an ace.

(1) Threshold of Greatness. Seeking the 90th percentile of MLB pitchers seems like a very high standard; it’s quite literally the top of the league, but includes a large enough group of pitchers to make meaningful comparisons. What I mean by this is, if we used a more strict threshold, there really would not be any aces; take the Top 10 pitchers by Runs Prevented in 2017 and 2018, for instance:

I rather like this, in the sense that it demonstrates that aces do not exist, but it doesn’t pass the eye test. If you’ve designed a threshold where only Corey Kluber, Chris Sale, and Max Scherzer are aces in 2017 and 2018, but not Kyle Freeland, Justin Verlander, Clayton Kershaw, Blake Snell, or Jacob deGrom, you’ve probably missed the threshold and poorly defined the concept.

An interesting problem arises with using the 90th percentile threshold, however: starting pitchers no longer dominate the proceedings, and 10 Average Runs Prevented is the measure for an ace. This surely won’t do for Brewers fans, for example, as 2016 Junior Guerra is absolutely, positively an ace under this regard; interestingly enough, so are 2018 Jeremy Jeffress, Josh Hader, and Wade Miley. Of course, this is conceptually interesting because Brewers fans absolutely did not want Miley to work in the rotation, and he was an “ace,” it turns out. Really, a truly fantastic pitcher, better than 90 percent of the MLB. Jeffress and Hader are difficult in terms of “acehood” because they do not start ballgames; but it’s not clear to me than an ace must start ballgames.

Consider the false controversy to open 2018, regarding whether or not Josh Hader should start or work in his relief role; as a starter, the concern is clearly that Hader does not have the command or pitching profile to work through a batting order multiple times, and based on news throughout the season, his delivery is probably too high effort to withstand a starting workload. But that was never the debate; the debate among angry fans was, “Josh Hader should start because the Brewers must see whether he can be an ace.” Ironically, based on the 90th percentile threshold of Runs Prevented, Hader is an ace, undoubtedly so (in fact, he’s better than 95 percent of the league in 2018; Jeffress was even better, beating 98 percent of the MLB). In this debate, it seems that the ideal of “finding an ace by making sure they can start games” misses the concept of what an ace ought to do (“consistently prevent runs at an elite level”), and so the Hader starting pitching controversy falls by the wayside.

Both Hader and Jeffress are aces, on this model. As they should be.

(2) Consistency. Josh Hader showed flashes of brilliance during the 2017 season, in which he established a clearly valuable MLB “floor” performance level, and raised debates about how high the “ceiling” performance could be. I gather this is in part why fans wanted so badly to see Hader start; when he began his career in the bullpen, it was certainly due to his tough season at Triple-A Colorado Springs, where he lost his breaking ball and had his stuff and command profile back-up a little bit. Yet Hader excelled in his MLB debut role, working through some command issues by offsetting bad outings with a dozen scoreless multi-inning appearances. Hader prevented between 13 and 14 runs in 47 and 2/3 innings.

Any team will receive considerable value from a pitcher that prevents more than 10 runs in a season; no team would balk at a chance at 90th percentile runs prevention production. Yet, there is more value to be had from pitchers that can repeat the feat in consecutive seasons. For example, 639 pitchers worked in both 2017 and 2018 seasons, with considerable variance in their production. The typical back-to-back pitcher in 2017 and 2018 found their runs prevented total shift by at least nine runs, with their innings pitched total fluctuating by 33. If you ascribe to the rule of thumb that approximately ten runs are equivalent to “one win” when balancing Runs Scored and Runs Allowed, the MLB pitchers that worked in 2017 and 2018 fluctuated enough to produce as many as 575 total wins (or losses), depending on how teams balanced their resources (as a side note, this is one reason organizations should not ever tank, and should always try to compete: there are always tons of runs that can be “captured” every year through trades, free agency, and player development fluctuations).

This is where I believe fans and analysts turn against “one year aces” like Junior Guerra. The idea is not that Guerra was not valuable to the Brewers in 2016, but that it is more valuable to have a pitcher that a team can “depend” on to produce 90th percentile performance year-in, and year-out. The trouble is, these pitchers do not readily exist. Looking at 2017 and 2018, here are the pitchers that were able to produce 10 Average Runs Prevented (or better) in both seasons:

This seems like a good list, for two reasons: (1) it’s intuitive enough to pass the “eye test” in terms of including pitchers like deGrom and Kershaw, but it also includes some “newcomer” aces like Aaron Nola; (2) it includes enough counterintuitive pitchers that we can look into new cases and further define quality pitching roles. Here, I’m thinking of Josh Hader (who is an ace in relief), JA Happ (who has morphed into an extremely consistent late career pitcher), and Dellin Betances (who is typically criticized as a middle reliever when it comes to contract negotiation matters, but should be thought of in the highest terms of consistency). Moreover, this table shows the benefit of being more inclusive in terms of defining aces, rather than less exclusive. Here’s what happens if you only consider pitchers with consecutive 20+ Runs Prevented seasons to be aces:

This list is visually appealing insofar as we all know that these pitchers are great, the top of the game in fact. This group would be the true 99th percentile of the game of baseball at the moment. But, it does not capture easily acquired pitchers (except for the Cleveland arms, perhaps, in Carlos Carrasco and Corey Kluber, this is an expensive group in terms of draft status, contract, or prospect resources required for acquisition). So, the cases of aces exclude most MLB teams from acquiring aces. Furthermore, there is no diversity of roles in this table, and to my eye that seems like the biggest shortcoming in this definition of the ace: for if we continue to define aces according to the highest possible standard, we will continue to replay and repeat the “Josh Hader should start” debate, and miss the reasons why Wade Miley or Jeremy Jeffress could be aces, too.

Building a pitching staff based around flexible roles, or based on elite relief roles and interchangeable starting rotation roles, does not preclude acehood. In fact, the 2018 Brewers demonstrate that effectively, both with multi-year consistent aces (Hader) and (potentially) one-off successes (Jeffress, Miley).

The playoffs present an uneven environment for predicting (or even projecting) potential pitching performances.

First, there’s the “any given day” concern, which holds that in any given day your starting pitcher simply might not have their stuff; this happens regularly throughout a 162-game season, but pitchers typically have anywhere from 10 to 20 to 30 additional starts in which to make adjustments. For example, Blake Snell, the best pitcher in the MLB according to Average Runs Prevented, allowed five runs twice, four runs once, and three runs once; Wade Miley, the Brewers’ best starter by Average Runs Prevented, allowed four runs twice and three runs twice. On those particular days, both of these solid-to-great pitchers were not of great use to their club (the bats would have to pick them up!), but over 12 additional starts for Miley and 26 additional starts for Snell, adjustments held up. The Brewers have even seen this with Clayton Kershaw, who got hit hard in Game One before adjusting for Game Five of the LCS.

Second, the playoffs are an absurdly uneven environment in terms of distributing runs. On average, each team is scoring approximately 3.9 runs per game in the 2018 playoffs thus far, which is notably lower than the regular season environment. So on the whole one might expect that better pitching prevails, even against the best offenses. But this really is not the case; in 26 games, the road team has scored more than four runs 11 times, while the home team has scored more than four runs 9 times. Thus 38 percent of the time a batting team has been better than average, and 12 percent of the time a batting team has been exactly average (four runs scored). Batting teams have been resilient in making adjustments, and the ratio of average or better games leads to questions about the extent to which “pitching wins championships” (if pitching is “better than average” half the time, it’s not clear that that’s the resounding vote in favor of loading up on elite pitching in order to win in the playoffs).

Third, regular season runs prevention figures should not be expected to hold up even in that same season’s playoffs, for the playoffs are not a “war against attrition” (as are 162 games). Instead, the playoffs force teams to make roster decisions to typically stick with their very best players for five-to-seven games, disallowing in-series roster moves (with the exception of injuries). This results in extremely uneven roster management, and example of which could be Craig Counsell leaning on three of his elite relievers rather than using his whole group of relief pitchers for much of the playoffs.

Finally, it’s worth diving into the particular match-ups that occur in each series. For example, both the Dodgers and Brewers bats have been held “cold” throughout the NLCS, averaging 3.20 Runs Scored per Game each (yes, the NLCS is “tied” at 16 runs apiece). Ironically, the Dodgers ace pitchers have not been as strong as their bullpen throughout the series, whereas it’s been the Brewers starting pitching that has notably out-performed the excellent bullpen. Both of these facts can be analyzed on a match-up by match-up basis given the nature of a short series; one particular relief move truly can neutralize a batter, or backfire. In terms of Houston, their “true ace” rotation was generally crushed by Red Sox bats, which is worth underscoring: the Red Sox had the best offense in the playoffs, the Astros the best pitching, and Boston still managed to score 29 runs over five games (a good total in any environment).

I’m not saying this to discourage discussion about aces (yes, everybody likes elite starting pitching), or to suggest that pitching is not important (it is!). I simply wish to emphasize that (a) regular season pitching quality does not categorically translate to playoff pitching quality, (b) you can stress the individual match-ups in a short series where each team is leaning on a very small percentage of their organization, and (c) you should not rely on bumper sticker narratives. Pitching wins championships, except for when hitting does.

Explanation: Exploring Runs Prevented

Following a new format for my articles (where possible), the tables will be presented first for those who do not wish to read about the guts.

Table One: 2018 Runs Prevented by MLB, AL / NL, cFIP, and DRA; Ranked by Current AL / NL Runs Prevented

If you cite this table, please link this article and cite Baseball Prospectus and Baseball Reference.

At the team level, Runs Prevented basically expresses the extent to which a club’s pitching staff is better or worse than the league average adjusted for their park factor. Thus, it scales team pitching production to the environment faced by pitchers (both in terms of overall league, and in terms of park). Yet, this is only one way to present the stat; strength of schedule could be included, and different park factors could also be used. I happen to prefer the multi-year factor published on Baseball Reference, but that’s not to say that other metrics are not valid for other presentations of this statistic.

Table Two: DRA Correction of 2018 Runs Prevented for 162 Game Performance

If you cite this table, please link this article and cite Baseball Prospectus and Baseball Reference.

To account for these types of variances, in recent years I have periodically scaled Runs Prevented to Deserved Run Average (DRA) and contextual Fielding Independent Pitching (cFIP) stats. The table above demonstrates 162-game Runs Allowed and “DRA Allowed,” along with an estimate using DRA to correct Runs Prevented. The benefit of using these stats to publish Runs Prevented metrics is that one can attempt to use two predictive metrics to assess teamwide performance. This essentially could help to scale expectations about whether a club’s Runs Prevented performance is sustainable; for example, as Table One showed, the Brewers’ exceptional pitching staff, a top five team by runs prevented, is basically on pace to be an average club over 162 games when assessed by both cFIP and DRA factors. So while there are reasons to think the club could sustain its performance in terms of winning (the bullpen is one of the very best in the league at 3.22 DRA), overall one might expect the club to descend back to an average Runs Prevented performance. (Should this team continue to contend for the playoffs, they could be one particularly strange playoff team, basically driven solely by elite fielding and elite bullpen performances). (Yes, the Brewers fielding performance is very, very good.)

Table Three: Variance of 162 Game Extrapolation for 2018 Runs Prevented

This is not necessarily a knock on the Brewers, however. As the table above shows, extrapolating runs prevented figures over a 162 game demonstrates remarkable variance. There is really a statistical sense in which the baseball could land any which way, meaning that an average pitching staff could be expected to land within a wide range of outcomes the vast majority of seasons. The variance figures above are not necessarily foreign, nor are they solely due to extrapolation (even if extrapolation contributes some of the variance here); in 2017 NL, the average NL club of 0 Runs Prevented demonstrated Standard Deviation of +/- 100 runs, while the average AL club of 1 Run Prevented demonstrated Standard Deviation of +/- 91 runs. (Hence, aces do not exist.) Internal to their own range of MLB, league, cFIP, and DRA Runs Prevented extrapolations over 162 games, each team does not fare much better, with a typical variance of +/- 56 runs. This range should be intuitive when one considers a club like Milwaukee, where the club could indeed finish 124 runs better than average at their current pace, or decline to 2 runs below average at a harsher DRA pace.

If the Brewers pitching follows their DRA correction course, the offensive troubles of the club will be amplified over the course of the remaining season. Thankfully, the thawed out bats in May are producing runs at a generally average rate, meaning that the club could have a completely different look by the end of the season. Milwaukee’s likely range of Run Differentials is quite well defined even after only 44 games, which places the club in an intriguing position: while the bats provide rather obvious spots for improvement, depending on how the front office views the underlying pitching performances, the pitching staff could be a site for roster improvement as well.

Photo Credit: Jennifer Stewart, USA Today Sports Images

References:

Baseball Prospectus. Team Pitching – Standard. CSV retrieved May 17, 2018.
Baseball Reference. Team Standard Pitching [MLB, NL, AL]. CSV Retrieved May 17, 2018.
Baseball Reference. Three Year Park Factors. Retrieved from baseball-reference.com, May 17, 2018.

Stats:

MLB Runs Prevented: Based on MLB RA9 of approximately 4.45, and Three Year Park Factors (Baseball Reference). [(IP/9)*(MLB_RA9 * Park Factor)] – [Actual Team RA]

League Runs Prevented: Based on NL or AL RA9 and Three Year Park Factors. [(IP/9)*(League_RA9 * Park Factor)] – [Actual Team RA]

cFIP Runs Prevented: Based on NL or AL RA9 and Three Year Park Factors, plus each club’s cFIP according to Baseball Prospectus. [(IP/9)*(League_RA9 * Park Factor)] – [(IP/9) * (League RA9 * cFIP/100)]

DRA Runs Prevented: Based on NL or AL RA9 and Three Year Park Factors, plus each club’s DRA according to Baseball Prospectus. [(IP/9)*(League_RA9 * Park Factor)] – [(IP/9) * (DRA)]

162 Game Pace Stats: For DRA and standard RA/G, take either stat and do the following: [RA/G  * 162] or [DRA/G * 162]

DRA Correction: Building on current Runs Prevented, this stat estimates what a club’s final Runs Prevented would be if the remaining games follow DRA. [Current Runs Prevented] + [ (162 Game DRA Pace) * (Percentage of Remaining Games)]