The decline in the median number of innings pitched by top starters is uneven, but unmistakable, falling by nearly a quarter.  The complete game decline is even more pronounced and more visible.  The two are, obviously, related.  All this we know.  But what of the explanation?

Proving why the observed trends are happening is considerably more difficult than demonstrating the pattern, but let's take a look.  The hypothesis is that pitchers are throwing more pitches per batter and more pitches under pressure now than was the case a generation ago.  Unfortunately we don't have a pitch counts database (at least not one that I'm aware of), so we can't definitively address the question.  The premise that pitchers were routinely throwing 150-175 pitches in 1969 is unproven.  Walks and strikeouts per inning pitches are displayed above because they are plate appearances that typically involve a greater than average number of pitches thrown per batter; as a general rule, the more strikeouts and walks a pitcher issues, the more pitches he throws.  Walks have in the 1990s, begun to creep upward notably, but aren't substantially higher than was the case in the early 1970s.  Strikeouts, however, are another matter.  While relatively high in the years immediately following the 1969 expansion, strikeouts predictably dropped in the years succeeding the introduction of the designated hitter in the American League.   With fewer pitchers batting there were fewer strikeouts.  But beginning in the early 1980s the strikeouts--unaffected by substantial rule changes like the DH, unaffected by sea changes such as 20% expansion in the number of major league teams--began to climb noticeably.  In the late 1970s and early 80s there were fewer than five strikeouts per nine innings.  As baseball moved into the second half of the 1990s, the average was in the mid-sixes and growing (the 1997 average was just below 6.7; the 1998 mark is 6.6).  That's a very large difference and, indirectly, demonstrates a rise in pitches thrown. 

The baserunners per inning column shows a somewhat less, but clear, increase over time as well, though most of it in the decade of the 1990s, from a mark around 12 per nine innings to more than 13 by 1996 (and continuing in 1997 and 1998).   More runners on base means more pitches from the stretch, more pitches in pressure (i.e. run-scoring) situations. 

Couple with that the pronounced increase in runs since the mid-70s and you complete the circumstantial picture.   There was a brief regression in scoring at the end of the 80s, but since then the run increase has found new life.  While the average team in the early 1970s was scoring fewer than four runs a game, the average club in the late 90s is scoring more than five.  That's a phenomenal increase (with no end presently in sight).  But as the MLB earned run average has moved from the mid-threes to the upper-fours there remains an inclination to use the same standards to evaluate pitchers from both eras.  We still look askance at a pitcher with an ERA over four, even though (especially in the American League) such a mark in 1998 signifies a starter as clearly above average.  Such a pitcher was a poor performer back in the early 1970s.  Furthermore, I daresay--though I can't prove it--that if the average big league starter in 1972 was allowing three earned runs by the end of the sixth inning (he wasn't; he was allowing less than two) he would have been much more susceptible to being lifted for a reliever.  Similarly, a 1998 pitcher who has allowed only a run or two through six is more likely to remain in the game.  Of course, it goes without saying that a pitcher who allows fewer runs throws, on average, fewer pitchers than one who allows more.  These things feed on one another.  In any event, this factor seems to me to imply a clear drop in innings thrown by starters.  It provides a less convincing explanation, in my view, of the complete game decline because complete games began an unmistakable drop before the sustained increase in offense was evident.

In any case...more pitchers per batter, more batters per game, more runners on base, more runs scored, more "pressure pitches" thrown as a percentage of the total number of pitches tossed...  We have indirect evidence that today's pitchers are throwing more pitches per batter, per inning, than in the past.   We have clear evidence that they are throwing more pitches in run-scoring situations.  A plethora of subtle changes in baseball over the past 30 years or so have conspired to make pitchers work harder.  Strike zones are smaller, ballparks are smaller (shorter fences, less foul ground), baseballs may be "juiced."   It's harder to get batters out than it was 30 years ago, despite the introduction of a number of new, difficult to hit pitches; it's harder to get batters out because it's harder to throw strikes, because foul balls are less likely to be caught, because balls are both traveling further and have less distance to travel to produce instant runs.   All of these factors have conspired to put more pressure on pitchers generally and that obviously includes starting pitchers. 

Now, this isn't the entire explanation.  The number of saves--and save opportunities--has exploded over time in major league baseball.  The apparent love affair with relief pitchers in baseball is not a new phenomenon.  The deliberate, tactical use of relief pitchers dates at least as far back as Joe Page's stint with the New York Yankees in the late 1940s.  Jim Konstanty of the 1950 Philadelphia Phillies appeared in 72 games--all in relief--and won the National League MVP Award.   The role of the late-inning reliever had become so institutionalized by the mid-60s that Major League Baseball--with a bit of prodding from sportswriter Jerome Holtzman--formally adopted the save rule.  The notion of the closer was already gaining acceptance by this time and saves continued to pile up.  Applying the save rule retroactively, from 1960 through 1975, 59 different clubs accumulated at least 40 saves in a season, led by the 1970 and 1972 editions of the Cincinnati Reds, each of which had 60.  The 1965 Chicago White Sox had 53.  But the evolution of relief pitching has shown increasing specialization (recall the statement quoted near the top of this diatribe).  It's not simply a matter of having a reliever enter in the eighth or ninth inning.  The notion of a setup man began to gain credence in the late 1970s and was a staple by the mid-80s.  There was no longer merely concern about the growing strain on starting pitchers.  Early closers, such as John Hiller of Detroit, Rollie Fingers of Oakland and Clay Carroll and Wayne Granger of Cincinnati began by averaging at least an inning and a half, and often more than two innings, an appearance.  Today's closers scarcely average an inning per appearance.  Lefty specialists--pitchers whose sole role is essentially to retire one or two left-handed batters--began to come into vogue in the mid-80s and now every club has at least one and many have two.  Right-handed specialists are now beginning to be implemented by some teams around baseball.  Teams continue to use more relief pitchers per game each year.  Consider the chart below:

The Reliever/T/G columns refer to relief pitchers used per team per game.   Note how the massive increase in relievers used really didn't begin until the early 1980s (that's the increased use of the setup man) and then exploded in the late 1980s as the lefty specialist role became institutionalized.  The increase continues, with no end in sight.  We're approaching the point where an average of four pitchers per team per game will be the order of the day.

The point is that this increased specialization has accompanied the drop in innings pitched by starters and, obviously, complete games.  It has also, quite obviously, paralleled the increase in runs per game.  (Perhaps, as outs become more difficult to get, managers believe that they're maximizing the opportunity to retire each batter by tailoring each plate appearance to their advantage.  Intentions notwithstanding, what they may be doing is ensuring that more bad pitchers are throwing a higher percentage of their club's innings.  The percentage of innings thrown by a team's top eight pitchers has been dwindling, not surprisingly over the past 30 years.)

Increased specialization of bullpen roles aside, the preponderance of our indirect evidence strongly implies that today's starting pitchers have to work substantially harder to get batters out than was the case a generation ago, which makes throwing a lot of innings and completing games a very difficult task to fulfill.  The pronounced drop in innings and complete games preceded the massive increase in the number of relievers used per game.  The use of "closers" was already well established by the early 1970s.  The percentage of games completed by starters has been falling, it should be pointed out, through the entire modern history of baseball.   In 1910, 62.3% of starts were completed; in 1920 that mark had fallen to 56.7%, and by 1930 it was down to 44.4%; by 1940 it was little changed (44.3%) but had slipped to 40.3% by 1950.  By 1960, after the notion of a regular role of a relief pitcher--ingrained by Page and Konstanty--had been institutionalized, the percentage collapsed to 26.9%.   Offense in baseball increased with the end of the deadball era in the 1920s and exploded in the 1930s, then remained high through the 1950s, which may well help to explain the increased use of relief pitchers (it's part of the argument that I'm putting forth for the increase over the past 30 years) over that span of time but declined markedly in the 1960s...and despite the substantial drop in runs scored from 1960 through 1968, complete games remained almost static, as the table below reveals. 

The implication of this--or at least the inference I've drawn--is that the influence of the ingrained notion of using the bullpen, which certainly was well-cemented by the early 1960s--had taken over.  If the increase of offense (and all of its accompanying consequences:  more pitches per batter, more batters per game, more pitches in pressure situations, etc.) was the only factor in the decline of the complete game, one would expect a large, sustained retrograde move--more complete games--when offense declined, as it plainly did in a major way in the 1960s.  As you can see, however, the "increase" in complete games over the decade was modest and essentially inconsequential (particularly considering the magnitude of the decline in runs scored) and certainly wasn't lasting.

Over much of the last thirty years the run-scoring trend of the 1960s has been heavily reversed.  Couple increased offense with the ingrained thinking bias, my argument goes, and you have a further decline in the number of innings that starting pitchers throw and a commensurate decline in complete games.  This is now a well-established trend, with no end in sight.  Barring both a reversal of the factors that have conspired to promote scoring and some other change--perhaps a move that directly or indirectly limits pitching staffs to nine or ten-- the trend is highly unlikely to be arrested, let alone reversed.