If asked how baseball has changed most over the past three decades, the answer could be the precipitous decline of the complete game. Lots of things change in scale, in frequency, in importance. But perhaps no baseball feat has traversed from routine to nearly extinct to quite this degree. In your lifetime, you might very well see the last complete game.
Just 30 years ago, there were 622 complete games, roughly four every day. There were 42 all of last year, despite there being four more teams in the league (and thus 648 more pitcher starts). This decline, which is certainly not a total revelation, steepened over the three years prior, from 104 (already an alltime low) in 2015 to 83 in 2016 to 59 in 2017. Roughly speaking, that’s 20 to 30 percent of complete games disappearing every year, long after the achievement had seemingly already become an anachronism.
(The blue line shows complete games by year since 1988. The red line shows how many complete games per 4,860 games played — a full season these days — to control for strikeshortened seasons and expansion.)
So, if one were to extrapolate the past three years the simplest way possible, we’d say there will be around 32 complete games this year, 24 or so in 2020, in the teens by 2021 and in single digits by 2023. By 2028, there will be two complete games a year, fewer (paradox alert!) than there are completegame nohitters. Within 10 years of that, there will be a complete game about every decade, and by 2047 there will be a complete game per century. So, to answer one question, you — you, specifically — will see your final complete game sometime in the mid2040s.
We mostly jest, of course. Simple extrapolations are too simple. Not all complete games are threatened equally, and not all complete games are threatened by the same forces. So to really get into this, we need to know which complete games are gone, and where they went.
It helps, perhaps, to think about this decline as it relates to two players. In and around 1988, Roger Clemens was probably the game’s best pitcher, winner of the previous two Cy Young Awards and a contender for it in ’88. He threw 14 complete games that year. In and around 2018, Max Scherzer was probably the game’s best pitcher, winner of the previous two Cy Young Awards and a contender for it in ’18. He threw two complete games, same as he threw in 2017. Each pitcher was his era’s horse — Scherzer led the National League in complete games last year and in 2017 — but the expectations had all changed. So where did all but 42 of those 622 complete games go? Where did all but two of those 14 go?
1. The Ones Lost to Safety
In 2018, only one pitcher was allowed to throw as many as 130 pitches in a start — Sean Newcomb, who was chasing a nohitter. (After allowing a hit with two out in the ninth, he was immediately pulled.) Only two more pitchers were allowed to throw more than 125 pitches: Tyson Ross (nohitter pursuit) and Trevor Bauer (selfprofessed pitchcount skeptic/durability unicorn). The most pitches anybody else threw was 122 (Justin Verlander, along with Bauer again). We can surmise, then, that in 2018, most managers had an unofficial hard cap of 122 pitches, barring nohitter pursuits, which are no longer at all automatic, either.
That’s one way of deducing how many pitches a pitcher gets in his quest for completion. Another, more universal one would be to look at the 95thpercentile pitch count for starters in a season, which would give us a pretty good idea of where almost all managers would set the ceiling on almost all pitchers in almost all circumstances. The 95thpercentile pitch count last year was 109. For most pitchers, 109 is the soft cap, with maybe one more batter allowed under some circumstances.
Both “limits” — 122 pitches and 109 pitches — are not only way down from 1988 limits, but are continuing to drop every year. As to the hard cap, in 1988, at least 18 pitchers threw at least 150 pitches in a start, and even five years ago, 17 pitchers topped 125. As to the soft cap, here are the 95thpercentile pitch counts in each of the past five years, as well as in a few years further back:

1988: 131

1990: 127

2000: 124

2010: 117

2014: 114

2015: 112

2016: 112

2017: 111

2018: 109
That 122pitch hard cap eliminates at least 206 of the 612 complete games that were thrown in 1988. (There were 19 starts with incomplete pitch information.) It eliminates nine of Clemens’ complete games. These disappeared complete games have nothing to do with ingame strategies, thirdtimethroughtheorder penalties, bullpenning or openers. They’re just now broadly seen as unsafe, and good riddance.
2. The Ones Lost to Modern Offense
An additional 194 complete games in 1988 took at least 110 pitches, but 122 or fewer. These would be our 95thpercentile pitch counts, strenuous but still within the limits of modern pitcher usage — especially for a pitcher like Clemens (then) or Scherzer (now).
But it’s not that easy. In 1988, the average plate appearance lasted 3.58 pitches. In 2018, it lasted 3.90 pitches. There were also 33 percent more outs made on the bases — caught stealing, pickoffs or runners thrown out trying to advance — and slightly more double plays in 1988, so even with the same OBP in both seasons the modern pitchers have to face more batters to get the same number of outs. Bottom line: It takes an average of 13 more pitches to get 27 outs now than it did 30 years ago, which means the seemingly attainable 110pitch start in 1988 would demand an outofrange 123 pitches today. That means the 194 complete games between 110 and 122 pitches in 1988 join the other 206 in the category of FlatOut Unsafe range, including two of Clemens’. Unsafe, good riddance.
3. The Ones Lost to Strategy
There were 162 complete games in 1988 that took at least 97 pitches but not more than 109, which — adjusted for the 13 extra pitches required to get outs today — we can now adjust up to 110/122, the 95th percentile of pitch counts last year. These 162 represent our pool of complete games that are narrowly within modern standards, but only if everything works just right.
But there were only 15 complete games in this pitch range last year, which is one way of saying everything rarely works just right. A few examples of potential complete games interrupted:

On Sept. 4, the Yankees’ J.A. Happ threw six innings in 70 pitches, having allowed only two baserunners and only one run. But after a leadoff single in the seventh, he was quickly replaced by David Robertson, who was replaced by Zack Britton, who was replaced by Dellin Betances — all of them former AllStars, all of them closers, all of them with ERAs in the 2s for New York. There were 110 relievers last year — nearly four per bullpen — who had ERAs at least 20 percent better than league average, more than double what there were in 1988. Almost every team’s bullpen is deep.

On Sept. 26, Jhoulys Chacin was pulled after throwing just 60 pitches in five innings. He had a 21 lead, but the top of the order was coming up, and teams are increasingly skittish about letting their pitchers face batters, and particularly heartoftheorder batters, a third time. (Offensive numbers spike the third time hitters see a pitcher, from either pitcher fatigue or hitter familiarity.) The Brewers pulled their starting pitcher after exactly 18, 19 or 20 batters 38 times, the most in baseball and the secondmost in history. (The 2012 Rockies, who were running a fourmanrotation experiment, had 43.) Chacin had nine such games, the secondmost in history (behind one of those Rockies.) There were almost 600 such starts across baseball last year, 100 more than in any other season and more than twice as many as there were just eight years earlier.

On April 15, Verlander completed eight innings in 100 pitches. He had 11 strikeouts and hadn’t allowed a hit since the third inning. But the playoff format isn’t the same as it was in 1988 — more teams make it, and the eventual champion might need to win as many as four playoff rounds instead of two. That meant the frontrunning Astros could be (a) much more certain than a 1988 team that they’d need Verlander to pitch in October and (b) more worried about the extra workload in October, which could add 30 or 40 innings to his season. (Verlander had, in fact, thrown an extra 37 innings in the previous postseason.) It made more sense to save his arm than wring it dry.

And on June 5, Scherzer had thrown only 99 pitches through eight innings, with 13 strikeouts, no walks and a 42 lead. But running a regimented bullpen means keeping pitchers on something of a schedule, neither over nor underworked. Closer Sean Doolittle hadn’t appeared in the previous couple of games, and if he hadn’t pitched that day — and with a Nationals travel day coming up — he risked going almost a week between outings. In 1988, there’s a pretty good chance Clemens would have just stayed in, but in this case, out ran Doolittle, and Scherzer’s pitching line ended at 8.
Now, as we said, there were 15 complete games in this highpitchcount range last year, so 15 times when everything worked out right. But for the most part, the league’s 42 complete games in 2018 came on especially low pitch counts, including five eightinning complete games (where the visiting starter lost and didn’t have to pitch the bottom of the ninth), one rainshortened shutout (which is how Trevor Williams led the NL in completegame shutouts despite never throwing a pitch after the seventh inning) and Rick Porcello‘s 82pitch novella.
There were 41 complete games from 1988 that took 96 or fewer pitches, which we’re considering the equivalent of 109 or fewer pitches in 2018. Now, knock that 41 number down a little bit to allow for bullpenning, the 5 percent of games when a team sent a reliever out to start the game, making a complete game totally unthinkable. And knock it down a little more for the extra caution teams use with young pitchers, since completegame rates for under25 starters (three last year, two the year before) really, truly are nearing zero. You’re left with about as many lowpitch complete games in 1988 as there were in 2018. And the decline is, more or less, all accounted for.
So, Then, Looking Forward
We found three primary factors for this nearextinction: a low cap for pitch counts, longer atbats, and strategic developments, which broadly prefer relievers to starters. The low cap continues to get lower, by roughly a pitch per year, and in fact dropped by two pitches last year. The longer atbats continue to get longer, by about one hundredth of a pitch each year, and in fact over the past three years have gone up by about double that rate.
More important, the preference for relievers was perhaps the dominant strategic story of last season, from the Rays’ regular use of the opener to the A’s bullpenning in the wildcard game to the Brewers Josh Hadering their way into the playoffs.
Nine teams didn’t have a complete game last year, from small markets (Tampa Bay) to large (Los Angeles Dodgers), such that it’s almost impossible to expect anybody in those rotations to complete a game. The two Cy Young winners combined for one, such that it’s hard to imagine any pitcher at this point racking up more than a couple.
So, since this is a predictions piece, we’ll say there will be … 32, like the simple extrapolation said. A little more than half will be thrown in the American League, perhaps one will be thrown by a pitcher under 25, none by those teams that have gone most heavily into bullpenning or 18batter starts, none by the teams with the best and deepest bullpens. Two by Scherzer.
And the trajectory beyond that? Well, if Major League Baseball were to expand by six franchises, then bullpens would be diluted, and even tiring starting pitchers might once again become preferable to the fourthbest reliever. Until then, though, most teams will be trying to find ways to use their relievers more, not less. We were just joking with our simple extrapolation. It also might be just about right.