By S. Christopher Michaels
As the title suggests, effective pitching in Major League Baseball is knowing when to accept the challenge with an opposing batter. To illustrate this, I’d like to share two examples before we examine data from the 2021 season for its predictive insights into games played today. The first is innocuous, buried in a game where both teams were already out of playoff contention. The second is both more recent and divisive in its risk-reward calculation.
On September 28 last season, the Kansas City Royals were forced to bring a reliever out of the bullpen in the first inning. Crafty veteran Ervin Santana took the mound in an unexpected situation and calmed the choppy waters. His scoreline won’t impress you should you find yourself reviewing historical box scores. What stood out is how Santana used a wealth of knowledge and experience to produce the desired result the Royals needed on a day when the starter didn’t even last an inning.
In the third inning of Santana’s late-season outing, he struck out Amed Rosario with a brilliant hesitation pitch ( skip to the 28-second mark ). Rosario was a young, league-average hitter in 2021, not yet 26 years old at the time of their duel. At 38 years old, Santana was in the twilight of his career. In fact, Ervin made only one more appearance in his sixteen years under the spotlight. On that day—yet another telling of a timeless classic—the old bull bested the challenging young pup by outsmarting him.
Our second example occurred just
last night
. In the fourth inning of an early-season contest between division foes, Los Angeles skipper
Joe Maddon
intentionally walked in a run with the bases loaded in a game his team was already trailing. Maddon had already brought in an inexperienced reliever,
Austin Warren
, to mitigate the damage wrought by starter
Reid Detmers
. In what was lambasted across social media as an inexplicable mental error by the
Angels
manager, the opportunity to pitch to a seemingly-underwhelming
Corey Seager
with the bases
loaded was yanked from young Austin’s hand. (
Pitch tracking data comes from Statcast at
baseballsavant.mlb.com
)
To better understand the context of this exchange, you have to look at Warren’s resume at the Major League level. He has a grand total of 24 1/3 innings pitched on the biggest stage, including last night. His repertoire includes a fastball, sinker, and slider—all of which he throws well. However, the game situation of a bases-loaded appearance facing Seager—who raked against breaking balls the past few seasons—was not where Maddon wanted to place his fresh-faced reliever. Bucking tradition, the wily manager preferred Warren face the Rangers catcher, Mitch Garver . Mitch hasn’t hit the breaking ball well lately. In fact, you have to go back to 2018 to find a season where he slugged over .200 against the funny stuff.
In a moment steeped in second-guessing, Warren watched three runners cross the plate in that fateful fourth inning. Regardless of how it played out, the decision to forego the potential challenge between Austin Warren and Corey Seager, battling strength against strength, remains defensible. Maddon clearly didn’t want to rattle his wet-behind-the-ears reliever. He opted for a data-driven matchup and lost the gamble. Yet, this example is as important as our previous instance in understanding when and how pitchers should accept the challenge. With that, our
analysis of 2021 pitching effectiveness begins.
The 2021 Major League Baseball season promised to be exciting. It marked the first full season following the Covid-shortened season of 2020. After a year of empty stadiums to mitigate viral transmission, it also saw fans return to the ballpark. Thinking back to this time one year ago, I was most intrigued to see how pitchers would perform over an entire slate of games.
It should come as little surprise that pitchers dominated the first half of the season. Their arms hadn’t been abused to total capacity in more than a year. Sports publications lamented the decline of scoring with wild suggestions to spark dying offenses. By season’s end, the data reverted to the mean. The pitchers had a strong season but didn’t shatter any game-outcome records. Runs-per-game, strikeouts-per-walk, home runs allowed, and a handful of other metrics were all consistent with previous trends. Another year in the books with an off-season full of wondering led me to the questions that produced this study.
What’s the most effective pitch in baseball right now? And who is throwing it best?
I wanted to provide a thorough approach to answering these questions. I elected to include all Major League pitchers who faced at least 162 batters. It’s a much lower threshold for qualification than what is used in the Bigs for awards consideration. At the same time, the current iteration of baseball includes so many different situational opportunities that any review of pitchers needs to be equally inclusive of different types of pitchers. All told, I compiled data entry for 411 such pitchers.
Before I continue, I would like to thank the folks at Baseball-Reference and Baseball Savant (Statcast) for their records collection. I wouldn’t be able to conduct research like this without them. All the same, none of the data I used for this comprehensive analysis was available in a single source. I pieced together data from multiple pages at BR. I entered pitch-type volume data from Statcast by hand because it doesn’t yet exist in a downloadable format. All told, it took me a few weeks to gather, enter, collate, and review more than 55,000 data cells.
Feast your eyes on the data that matters… ( well, it matters to me )
2021 saw more than 181,000 total plate appearances. Combined, the 411 pitchers in this analysis accounted for more than 83% of those appearances. Almost 590,000 pitches were reviewed. Needless to say, the sample size for this study is massive.
Before Opening Day, I shared some of this data on the Absolute (Sports Betting) Degeneracy Podcast . Visiting with the host—and my friend—Arch Stanton, I offered the split-finger fastball as baseball’s most dominant pitch. Ever-insightful, Arch asked me if that was due to its limited use by Major Leaguers. Undoubtedly, any pitch thrown less than 2% of the time has a strong likelihood of puzzling opposing batters. This phenomenon is amplified when that rarely-thrown pitch is delivered with unusual effectiveness. Having said that, there is no direct relationship between the frequency of a particular pitch type and its ability to produce desired outs. Look no further than the pedestrian numbers associated with cutters and sinkers. Those pitches rank fifth and sixth among the seven pitch types recorded by Statcast data in terms of OPS.
Globally, pitches that move make the batter’s box experience less fruitful for hitters. Unless you are a novice to baseball, this shouldn’t raise an eyebrow. Still, it is nice to put actual numbers to something we’ve long suspected. Split-fingers, curves, and sliders were each thrown for strikeouts in more than 30% of the plate appearances where that particular pitch was the final pitch thrown. The dominant trend continues with low home run rates and low OPS lines. Naturally, splits of more or less potent pitch types beg the question of why individual pitchers aren’t shuffling their repertoire to focus on those more dominant pitches. Unfortunately, it’s not that simple.
As much as the total population data appears settled regarding which pitch types each pitcher should select, the disaggregated individual data shows puzzling results. The starting pitcher who threw the most effective pitch the best— Shohei Ohtani and the split-finger fastball —ranks 95 th among our sample in pitching OPS+. I know what you’re thinking; what in the wide, wide world of sports is pitching OPS+? Well, it stands for adjusted pitching on-base-plus-slugging percentage. I was somewhat shocked to discover this metric didn’t already exist, given the prevalence of the same metric for batters. Designing the measurement wasn’t difficult either. I inverted the formula from batter-adjusted OPS to divide pitchers by the league average because pitchers would want to have a lower-than-average OPS. In contrast, batters want it to be higher-than-average. The formula for adjusted-pitching on-base-plus-slugging ( pOPS+ ) * is:
Again, lower is better for pitchers. Getting back to Ohtani , his pOPS+ of 127.9 ranks 204 th among all pitchers from 2021. Looking only at our sample of 411 pitchers who faced at least 162 batters, Shohei’s OPS+ was recalculated to reflect his performance relative to the sample population. In this case, his line drops to 123.8. For as dominant as the MVP was with the splitter, he was unimpressive with his four-seam fastball, ranking 280 th in OPS+. He also threw the 34 th best cutter and the 130 th best slider.
It defies logic to consider Ohtani throws his worst pitch 44% of the time while throwing his best pitch 18%. By percentage of pitches thrown, Shohei prefers the fastball followed by the slider, splitter, cutter, and curve. Sadly, Shotime is not alone in dominating with secondary pitches while hurling a lackluster primary pitch. This trend is repeated time and again among our qualified population. Pitch-type batted-ball data ( PITCHf/x ) has been around since the 2006 postseason. I will slowly work my way backward to analyze each season it has been used.
On another note, the most intriguing pitch type seems to be the sinking fastball, typically referred to as a sinker. Yes, this pitch type ranks sixth in OPS and last in strikeout percentage. The silver lining is sinkers are tied for second in lowest home run percentage behind only the mesmerizing splitter. Admittedly, pitchers continue to employ the sinker because of its low home run rate. While that is all well and good, the data suggests batters are not fooled by sinking fastballs, recording an alarming on-base percentage of .364. As a result, it’s hard to justify the reduced home run rate when batters reach base more than 36% of the time. Still, two pitchers threw exceptional sinkers.
The two pitchers who threw ridiculous sinking fastballs in 2021 were Aroldis Chapman and Nestor Cortes Jr.
Chapman’s sinker was extraordinary in 2021. Over 28 plate appearances where the final pitch was a sinking fastball, no batter reached base against Aroldis. His OBP and SLG were both a .000, which means his OPS+ was infinity. The sample size was undoubtedly small. All the same, let’s recognize greatness where it exists. For what it’s worth, Nestor Cortes was no slouch here. Over 21 plate appearances, only one player reached base. Nestor’s OBP ** was .048, while his SLG matched Chapman’s .000. Continuing this examination, the Kansas City Royals ( this author’s beloved team ) own Carlos Hernandez ranked third among 205 sinker-ballers with a recalculated ( sinker-only ) OPS+ of 385.6. In fact, there were fourteen pitchers in 2021 with a sinker-only OPS+ of 200 or higher. Only one of those pitchers faced at least 200 batters. That was none other than St. Louis’s Adam Wainwright , who recorded a sinker-only OPS+ of 225.4.
The data suggests the sinker and cutter were more effective pitches for relievers than starters. It is part of their intrigue because it informs us when to expect these pitches in a given game. By contrast, the change-up was clearly thrown more by starters. Curves and sliders seem to be well-represented among starters and relievers. Finally, nearly 85% of the 411 pitchers sampled threw fastballs *** . Still, the question of knowing when pitchers should accept the challenge presented by a given batter remains. To that end, I designed yet another statistic… ( cue the laugh track and impending facepalm )
To find a meaningful yet palatable metric that mirrors our current baseball lexicon, I crafted a statistic that looks and feels like the earned run average. It is based on the Outcomes Runs algorithms I designed to measure runs scored with greater precision. As an aside, the correlation with runs scored for my standard Outcomes Runs ( sOR ) algorithm is .9988, with regression analysis results that are orders of magnitude more accurate than even those publicly touted as the gold standard, such as Bill James Runs Created technical ( RCt ). I’m also pleased to share the Outcomes Runs betting algorithm ( ORbet ) has a near-perfect correlation with runs scored ( .9998 ). I said all of that to say I won’t design a metric that isn’t directly related to runs scored.
Whew. I always feel wrong to dump data when I’m talking to everyday folks. I don’t know your threshold for intensive statistical analysis. I want to be able to discuss baseball with all sorts of folks, so rest assured that’s it for the minutiae.
The ERA-like metric for pitchers is Outcomes Runs-27 ( OR27, for short ). We take a pitcher’s Outcomes Runs ( the total number of runs he is responsible for based on corresponding event outcomes ), divide that by the total number of plate appearances, and multiply the remainder by 27 ( because there are 27 outs in a game ). The result mirrors an ERA and uses the same concept that lower is better. The formula for OR27 is:
Among our qualified pitchers, Jacob deGrom had the best OR27. He’s a stud among alpha-studs. You’ve heard his name mentioned for stellar performances and record-shattering achievements if you follow baseball even casually. Last season was no different before deGrom was injured in July. His 2021 OR27 to that point was 0.04. The league average OR27 for the season was 3.24. By comparison, the league-wide ERA for 2021 was 4.26. I’ll be transparent that I haven’t done this level of analysis on other seasons yet. I cannot say where the league average OR27 from last season ranks across seasons. I can say that regardless of seasonal rank, an individual pitcher recording a 0.04 OR27 feels absurdly superhuman and is likely on par with the most celebrated seasons of all time. I’ll have to get back to you good folks when time permits with all of the baseball projects I’d like to undertake.
Twenty-nine different pitchers had an OR27 under 2.00. Most of them are household names. I’m not reinventing the wheel by telling you which pitchers are the most effective at rising to the challenge. Instead, I’m giving you the data to substantiate your barroom claims with hard evidence firmly rooted in the only data directly correlated with runs scored. You can thank me later. I guess the ultimate answer of when to challenge hitters falls somewhere between having dominant stuff and being able to fool those batsmen with your wits. If neither describes the man standing on the mound for your team, let’s hope your manager is brave enough to consider a Joe Maddon move.
I hope you’ve enjoyed this column. I want to challenge your thinking about baseball statistics. Someday, my own research on the game will become outdated. Please feel free to spar with me about the ideas I’ve presented here—I enjoy the discussion because it challenges my thinking. I can be reached here on Baseball Almanac, via email at christopher.s.michaels@gmail.com , and I’m on the social media ( Facebook , Twitter ). As always, this has been the World According to Chris . Thanks for tuning in and be sure to check the links below to the data tables and Outcomes Runs calculators.
Postscript Notes:
*The purpose of using adjusted statistics for batters or pitchers is to compare them across seasons and eras. Because a metric like OPS+ yields an absolute value by virtue of measurement against a given year’s league average, it can be easily compared to any batter or pitcher from any season for dominance relative to a player’s competition for that season. 100 is the league average for OPS+ (or any adjusted statistic), with values over 100 representing above-average performance and below 100 representing below-average performance.
Ex. Jacob deGrom had a pOPS+ of 267.4 in 2021. It implies his season’s performance was 167% above-average.
**The data set compiled for this analysis is as complete as possible. The pitch-type data available on baseballsavant.mlb.com does not record sacrifices for individual pitch types. The formula for OBP requires the inclusion of SAC flies. To determine the likeliest OBP numbers for pitchers, all plate appearances not counted as official at-bats for a particular pitcher were recorded as a batter reaching base. It does mean that a pitcher’s OBP for a specific pitch type MAY be off by no more than a few hundredths or thousandths, but not enough to substantially change the data as presented or analyzed.
***Pitch-type data only includes data publicly available on baseballsavant.mlb.com. Even on that website, there are internal errors where the subtotals of one set of data do not match the corresponding totals found in another data set on the same website. Players in italics have inconsistencies in their recorded data. Every attempt was made to minimize the total number of discrepancies in these data sets.
2022 ORbet Outcomes Runs Calculators Final – Public
Outcomes Runs Calculators – README – Public
