Bullpen revolution: Teams continue to emphasize late-game relief

Travis Sawchik of the Pittsburgh Tribune-Review has a good look at the evolving state of major league bullpens, especially for budget-conscious teams like the Pittsburgh Pirates, who are often priced out of the starting pitching market. He writes:

The Pirates enter the season with 23.1 percent of their payroll allocated to the bullpen, an all-time high under (GM Neal) Huntington, up from 15.4 percent last season and 13.8 percent in 2014.

The game always is becoming more specialized. But the growing role of the bullpen threatens to fundamentally change the way we think of starting pitchers and pitching-staff configuration. Analytics suggest bullpens should be relied upon to an even greater degree. Power relievers continue to increase in velocity, reducing hitters' reaction time. The amateur market is changing the types of arms available. Teams like the Kansas City Royals, with elite bullpens, have essentially shortened games to six-inning affairs. And the Pirates -- and the New York Yankees and Houston Astros -- are trying to follow in their path.

The Pirates intended to shore up their rotation this offseason after losing J.A. Happ to free agency and A.J. Burnett to retirement, but once the market exploded for starters they had to trade for Jonathon Niese and then spent what little money they did have on a couple arms to supplement a bullpen that already may have been the best in baseball in 2015.

As Travis points out, only 28 pitchers reached 200 innings last season, the lowest total in modern MLB history. That means the top starting pitchers are more coveted than ever and the average cost per inning for starters and relievers has diverged:

In 2009, based upon opening day payrolls, every inning pitched by a starting pitcher cost teams an average of $26,603.

The average cost of an inning thrown by a reliever? $26,855.

But last season teams spent $41,000 per inning on starting pitchers and $33,343 per inning on relievers.

What's interesting is -- in contrast with popular belief, I suspect -- the total usage for relief pitchers hasn't changed much in the past 15 years. Travis reports that relievers accounted for 34.9 percent of all innings in 2015 compared to 33.7 percent in 1999. That translates to 5.8 innings per start for starters in 2015 compared to 5.9 in 1999.

There have been two significant changes, however, in how pitchers are used. In 2015, relievers averaged 3.0 outs per game; in 1999, they averaged 3.5 outs per game. There were 15,108 relief appearances in 2015 in 4,858 team games, or 3.1 per game. In 1999, there were 12,420 appearances in 4,856 games, or 2.6 per game. Yes, your eyes aren't deceiving you; we're seeing more pitching changes than ever. These numbers are the result of expanded bullpens -- more relievers leads to shorter relief outings and thus more appearances.

Also note that 1999 came in the heart of the steroids era, when teams averaged 5.08 runs per game compared to 4.25 in 2015. That led to more short stints by starters than we saw in 2015. But that's evened out by the workhorses pitching more innings. In 1999, the 20 starters with the most innings combined for 4,574 innings versus 4,332 in 2015, a difference of 12 innings per starter. A minor thing, but top starters are certainly used a little more conservatively these days. Those differences become even more acute with the second-level guys: 56 pitchers threw 180 innings in 2015 compared to 70 in 1999. What we had in 1999 were a lot of bad starters: 29 pitchers threw at least 100 innings with an ERA of 5.50 or worse; last season, we had just seven.

What's the impact on this bullpen usage been on win-loss record? At least compared to 1999, it's led to only small improvements in late-game winning percentages, even though we seem to have more dominant relievers than ever. Check out the winning percentages in 2015 versus 1999:

When leading after six innings: .882 (2015) vs. .854 (1999)

When leading after seven innings: .925 (2015) vs. .893 (1999)

When leading after eight innings: .965 (2015) vs. .947 (1999)

Teams lost 116 games in 1999 that they led heading into the ninth inning. Last year, they lost 76, so 40 fewer defeats, or just over one fewer loss per team. In 1999, teams averaged a 61-10 record when leading after six innings versus 61-8 in 2015. When leading after seven innings, they averaged 65-8 in 1999 and 65-5 in 2015.

So, yes, bullpens are slightly more efficient. Relievers collectively had an ERA 6 percent below the average MLB ERA in 2015 compared to 5 percent below the average ERA in 1999.

In baseball, however, every team is striving to find that extra 1 or 2 percent that can lead to an extra win or two. That's why teams are going to continue to load up on as many strong-armed relievers as they can find and why managers will continue their non-stop trips to the mound to bring in a new arm. And why those teams with the best and deepest bullpens are good bets to contend for the playoffs.