With the advent of “openers”, “bullpen days”, and avoidance of the dreaded “third time through the order”, recent seasons have seen a quickening of the already rapid decline in the average length of a pitcher’s start. Perhaps, though, there is reason to surmise that starts may begin lengthening soon. Find out why after the jump.

Here are the average lengths of a start for each season of the live ball era.

The trend in declining innings per start is well established and was fairly consistent throughout the live ball era, other than these few anomalies:

• Steep decline in IP per start in the transition from the dead ball era
• Rise in IP per start during World War II years, based primarily on economic considerations
• Rise in IP per start in the late 1960s (“second dead ball era”) and again in the mid-1970s (advent of the DH rule)
• Rise in IP per start in the early 2010s (end of “steroids era”)
• Rapid decline in IP per start beginning in 2015 (perfect storm of protecting or over-protecting high-value assets, heavy reliance on analytics, and maybe a juiced ball, whether intended or not)

That last bullet is an understatement. As shown in the chart, IP per start have fallen off a cliff since 2015, with a decline of at least 0.12 IP per start each season since. That trend will end in 2021, which is looking like it will finish in the 5.05 to 5.1 IP per start range, but that number will still be ~0.1 IP lower than in the last full length season in 2019. For a sense of how rapidly start length has fallen recently compared to the past, consider the table below.

The years listed under “Last Year Above” are subject to change (especially years since 2011), but suffice to say that it took 71 years (1924 to 1995) for the average start to drop from below 7 innings to below 6, but only 25 years (1995 to 2020) to drop from below 6 innings to below 5.

I attributed the decline in start length in recent seasons partly to analytics becoming firmly entrenched in MLB. But the response to ever more sophisticated metrics showing pitchers becoming less effective the more pitches they throw in a game (as if we needed analytics to tell us that) seems very much a knee-jerk reaction of “get that starter out of there” without a lot of regard as to whether the bullpen can do any better. In fact, the bullpen has been doing better than starters for a long time, even as relievers have been called upon to shoulder an ever increasing share of the pitching workload, as shown in the chart below.

The bars in the chart are showing the difference between the ERA of starters and relievers, plotted against the right axis. A positive difference (bars above the midline) indicates that starters had better ERAs, which was the norm in baseball until about 1950. But, since then, relievers have surpassed starters in ERA (bars below the midline), and not by just a little, with advantages of one quarter of a run or more per game in most seasons since the mid-1970s. That advantage, though, has shrunk to almost nothing over the past three seasons, a trend continuing in 2021 with 4.19 ERA for starters and 4.18 for relievers at the All-Star break. Comparing relievers to starters pitching the 3rd time through the order (see chart below) preserves a bit of an advantage for the relievers but one that has also rapidly declined over the past few seasons (in 2021 at the All-Star break, starters have a 4.30 ERA the third time through the order).

Part of the recent sudden decline in start length can be attributed to the type of pitchers making those starts. The following charts show statistics for three groups of pitchers:

• Starters: pitchers making starts in 60% or more of games
• Relievers: pitchers appearing in relief in 80% or more of games
• Swingmen: pitchers who do not meet qualification for starters or relievers

The following chart shows percentages of team games started by pitchers in these three categories.

After reaching a peak in 2011 of 93.7% of games started by pitchers classified as Starters (orange line), that proportion declined to only 83.6% of games in 2019. Similarly, pitchers classified as Swingmen (green line) started a record low of 4.7% of games in 2011, then saw that number shoot up to 11.9% of games in 2019. And, starts by Relievers rose from a low of 0.9% of games in 2013 to a high of 4.4% in 2019.

Some of those increased starts by relievers and swingmen in the last few seasons were actually “opener” starts. Adjusting the average IP per start to eliminate distortion from opener starts yields results shown below.

The adjustment above is probably overdone as there were undoubtedly some starts removed that were not intended to be “opener” starts. But, this provides a rough idea of the impact of opener starts on average start length.

While starts by pitchers not classified as starting pitchers has been on the rise in recent years, that hasn’t yet altered to a large extent the underlying trend to greater specialization in pitcher roles that started in the 1960s, as shown below.

The chart above is showing the percent of appearances that were starts by pitchers in the three groups. The trends of the past several years that I’ve been discussing are most evident among Starters (orange line) where the long upward trend that started about 1960 has finally tapered off just a bit, with Starters making starts in 92.9% of their appearances in 2020, down from a peak of 96.2% in 2010. Swingmen have stayed consistently in the 35-40% range for some 70 years now, while the big move down by Relievers has much more to do with the many more appearances they now make than with their number of starts.

Looking at innings pitched by pitcher role produces these results.

The bars show the percentage of innings pitched by pitchers in the three classifications. The biggest change, historically, is among Relievers who contributed barely 5% of innings at the dawn of the live ball era, but are now nudging the 40% threshold. Starters have seen their innings share decline from over 70% a century ago to just above 50% today, while Swingmen have ranged from highs above 30% of innings in the 1930s to lows under 5% a decade ago.

The lines in the chart show the average number of pitchers per team in each group with at least 20 appearances (10 appearances in 2020). Despite the much reduced workload by starters, teams have maintained a consistent average of about 4 Starters with 20+ appearances. In their heyday, teams employed about 2 swingmen with 20+ appearances but nowadays few teams will have more than one, and many have none. For Relievers, it’s been an almost tenfold increase in the past century, from fewer than one per team to an average today almost nudging ten.

So, what lies ahead? As we’ve seen, there is now essentially no difference in ERA between starters and relievers. Part of that is selection bias from removing starters from games before they can get too badly hurt. But, the flip side is the more innings relievers have to shoulder, the lower the bar falls on reliever performance. Still, the prospect of more diminishing returns from more use of relievers would suggest that we may be close to the bottom in terms of start length. But, how can starters go longer? Given the dollars being lavished on pitchers and the “max effort” approach employed by so many of them, it’s hard to imagine teams allowing starters to raise their pitch counts much higher than they already are. Thus, longer starts will likely happen only when pitchers become more efficient, getting more outs with fewer pitches, and that will happen only when the objective of “missing bats” starts to swing more towards “inducing soft contact”. For this observer, that can’t happen too soon.