In part one of this series, we found that teams today retain leads better than any decade since the deadball era. However, the reason for this is that teams establish bigger leads today since their success in holding smaller leads (one-, two-, or three-run leads) has changed very little over the last fifty years or so. However, one has respect the way that middle relievers have stepped in as starters are getting pulled earlier all the time and have not missed a beat.

In the second part of the series, we found that late-game situations with any sorts of leads, small or large, teams on average have improved and continue to improve their ability to hold that lead, and apparently, that's thanks to the closer.

In the latest installment, I want to look at how the success rates across the majors varied over time.

My thinking is that if how much these rates vary (via their standard deviations) goes down, that means that the approach du jour has become more universally employed and the level of talent devoted to that approach is more uniform. If it goes up, then different approaches are being employed (e.g., some teams are using starters deeper into games or using closers in a non-save situations) or the teams have varying devotion to the approach and are therefore, devoting pitchers of varying talent to the role it defines.

Fortunately, I will not be publishing tables of standard deviations for the last hundred or so years, no matter how much I am inclined to do so, so this will be a shorter entry than the other two. I would publish the graph of how they varied but don't have a ready way to do it in this blog.

However, if there's one thing that makes for better reading than loads of table, it's describing said tables. So here goes…

I ran the standard deviations per year from 1901 to 2005 for team success rates in four situations: leads retained after six innings, leads retained in the ninth inning, leads of three runs or less retained in the ninth inning, and one-run leads retained in the ninth. Of these, the success rates for leads (of any kind) in the ninth had the least variance and the rates for one-run leads in he ninth had (by far) the most. The variation in success rates for leads after six and for leads of three runs or less in the ninth were very close, though the former was slightly higher on average.

OK, teams with big leads in the ninth usually win, no big surprise. However, I also ran a trend for each set of data (using Excel's exponential trendline). I wanted to see if teams were getting closer in success rates over time or farther apart. My expectation was that all of the data would be getting closer together as modern relief pitching matured and teams developed more uniform use of bullpen pitchers.

The standard deviation in success rates for any type of lead in the ninth did go down slightly overall (from about .035 to .033). For three-run leads or less in the ninth, the standard deviations also went down (from .055 to .052) as did those for leads kept after six (.068 to .056). But I was surprised that given that longer schedules and more teams should help standard deviations become more uniform, I am surprised that they variation between them did not go down more.

The one big surprise was that the variance for one-run leads has gone up over the decades (from .099 to .111). This means that teams are witnessing more and more wildly varying success with holding one-run leads in the ninth as the closer role becomes more and more well defined. This phenomenon appears to be independent of how well teams are scoring runs in the given year. Low scoring seasons in the Seventies saw the same sort of increase as the high-scoring Nineties in general.

They did drop right after Bruce Sutter defined the modern closer role in 1979 and had an even bigger and longer drop after Dennis Eckersley redefined the role in 1988. However, three years since 1993 have seen the highest variance since 1963 (though none came close to the all-time high set in 1940, .147).

One encouraging trend is that the variation has gone down two of the last three years, but it is too early to tell if it's a new trend or just a blip on the radar screen.

Given that I am anticipating a new trend in reliever use to take shape anytime soon (since baseball had been averaging one every ten to twelve years previously), I'm expecting some monumental change to occur at any moment. Now that teams have the middle relief roles well established, maybe the one-run leads in the ninth will become more of an issue. Who knows, maybe it's time for the return of the starter-sum-closer to stop the scoring in the ninth. Or maybe the trend change a few years ago is real and the new post-Eckersley definition of a closer (more in the Gagne role) is already here.

Next, I would like to update my old reliever study to see if there is indeed a new approach that teams are using with closers.