How deep should we expect Arizona Diamondbacks’ starters to go?

You might be inclined - I know I certainly was - to think that the average length of a starter’s outing has been trending down for a while. After all, in 1998 there were 161 complete games tossed in the National League. Last season, the same number was less than a quarter of that, at only 39 complete games. However, that is not reflected in overall averages, both in innings pitched and number of pitches per start. Those have remained relatively constant, at least until the past couple of years. In 1998, the average starter went 6.1 innings and threw 95 pitches. In 2014, the average starter put up almost identical numbers, going 6.0 innings and throwing 95 pitches.

The reason why that’s the case, despite the sharp drop in complete games, is because we’ve also seen a drop in really short starts at the other end of the spectrum, and an increase in the middle. The chart and associated graph below show a breakdown of starting length in the National League, sampled every three years since the Diamondbacks entered the league in 1998 - which conveniently means the most recent season was last year.

You can see that the middle blocks have generally been increasing in size, squeezing out the extremes on both ends. Last season, 58.0% of all starts went between 5.0 and 6.2 innings, compared to only 43.6% in 1998. However, in the last couple of years, things have changed. Last season, the average start lasted only 5.6 innings. This was driven both by a decrease in “long outings” - only 5.0% reached eight frames, compared to 8.1% just three years prior - and an increase in “disaster outings”, the ones where the starter couldn’t even get through five innings. At 20.6%, this was also sharply up on the 15.4% number from 2013.

Another factor comes into play here. Despite the innings drop, the number of pitches thrown has barely changed, at 92. This suggests the average starter now takes more pitches to get through an inning (last year, 16.4 pitches per frame, compared to 15.8 in 2014). A likely cause is the uptick in K-rate - at-bats that end in a strikeout will always take at least three pitches. For starters in 2016, the K-rate was 7.88 per nine innings, more than half a strikeout up on the figure just two years previously, when it was 7.36 per 9IP. [Relievers had a much smaller increase: 8.62 to 8.70] As someone noted, Ian Kennedy had a higher K-rate last year than Bob Gibson EVER had.

This does suggest we may need to re-calibrate our opinion as to what constitutes “adequate” length from our rotation, if getting the second out in the sixth inning makes for a longer start than most. This won’t be easy: I still haven’t come to terms with a world in which a .255 hitter is actually major-league average. But last year, there were only two National League teams, the Giants and Cubs, who averaged six innings from their starters. Arizona’s 5.5 was good enough for eighth in the league, and not a single qualifying NL starting pitcher averaged seven innings per game. This may become the new normal going forward.

For the trend doesn’t show any signs of stopping, with the average start to date in the NL being a mere 5.5 innings [though this may be in part due to pitchers still being stretched out over the first week]. This is actually slightly below what the D-backs have managed thus far. 5.6 innings, good for sixth in the league. [I note the Giants and Cubs occupy the top two slots again - not that it’s exactly helping San Francisco so far]

The quality start

There has been debate over this metric, invented by Philadelphia sportswriter John Lowe in 1985. Its definition - a pitcher going six-plus innings, while allowing three or fewer earned runs - has remained unchanged, despite the complete game becoming an endangered species over that time [in 1985, the NL has 267. The Cardinals and Dodgers each had 37, virtually as many as the entire league last year, as mentioned above, 39], and ERA varying. But the percentage of starts which qualify as “quality” didn’t vary that much over almost the first three decades it was in use. It was 56% in 1985, 49% in 2000, when ERA was over a run higher, then rebounded to 57% in 2014.

However, it has declined sharply of late, to 46% last season, likely in line with the innings drop. If the decline in starter innings continues, it may be worth looking at “revaluing” the quality start, especially if we want to be able to use it across different eras. Dropping it to five innings and two earned runs would keep it in the same realm - it was achieved by 48% of starts last year - while also addressing one of the most common arguments against it. Namely, that the current standard equates to a 4.50 ERA, which is not very “quality” after all. A shorter, tougher standard of quality, with a 3.60 ERA might better reflect pitching in the modern era.