A closer is a fairly special person in the baseball ecosystem. If being a starter is like being a marathon runner, closing is more like running the 100-yard dash. Closers are expected to pitch for one inning at a time, not six or seven.
Unlike starters, who need a wide repertoire of pitches so that opposing players can't catch on to them, closers just need one or two; say a fastball and hard slider. That's what Hanrahan has. Melancon features a fastball and curveball.
The formal requirement for being a good closer is to earn a lot of "saves" to your credit. Hanrahan has that in spades (100 as of May 2, 2013), which is why he was highly valued in the market for closers. That's something that can be earned only by having closed before.
Even so, Hanrahan, whose ERA skyrocketed to 9.82 in 2013, was on his way to being demoted before his injury in May.
In Moneyball, Michael Lewis made the case that almost any above-average "flamethrower" could become a good closer.
Accordingly, Oakland's Billy Beane would trade established closers (with many saves on their record), for the players he needed, and they would quickly find a suitable replacement. Neal Huntington has taken this art form to a new level by trading yesterday's closer for tomorrow's.
So why did Huntington trade a closer with a 2012 ERA of 2.72 (Hanrahan) for a reliever with a 6.20 ERA in the same year (Melancon)? Because something told him that they were actually more or less equivalent (as of the end of 2012).
That something was FIP (field independent pitching statistics). It is a formula based on the number of home runs, walks and strikeouts pitched by the pitcher.
Last year, Hanrahan's was 4.45. Melancon's was only 4.58, so statistically they were almost a tie. The underlying assumption is that pitchers' BABIPs (batting average on balls in play, which is the main remaining driver of pitchers' ERAs), are basically under the control of the fielders, not pitchers.
But Melancon's inferior ERA may have given Huntington the leverage to demand three prospects for the rest of the trade, while giving up only replacement outfielder Brock Holt in return.
Baseball players, particularly pitchers, tend to peak between ages 28 and 32. If there are two players with comparable ability, one aged 28 and the other aged 32, the former is likely to get better while the latter is likely to get worse. Coming into 2013, Melancon was 28, while Hanrahan was 31.
Even with a FIP no better than Hanrahan's, the younger Melacon was deemed to be more upgradable. For instance, his FIP and ERA went down sharply in 2013, after he dropped the sinker and slider from his repertoire and concentrated on his two best pitches. (He had been slated earlier to be a starter who would need all four pitches.)
When the Pirates acquired Melancon, they made him a set-up man, a role considered by many to be an "assistant closer." The greatest closer of them all, Mariano Rivera, actually won the MVP award in the All-Star Game in the set-up role.
Some would note that Jason "Grilled Cheese" Grilli is a good closer at 36. That's the exception, not the rule. And older closers are more injury-prone.
Grilli ended up on the disabled list, which is how Melancon landed his (likely temporary) gig as the closer. Melancon actually has the better ERA (below 1.00), while Grilli's is a quite-good 2.34.
Grilli will probably be restored to the closer position when he returns to the Pirates. But we're talking about someone who will be 37 years old in 2014 and 38 in 2015. Melancon will be 29 and 30 in those years, indicating that he will be the closer of the future.
In the past, I've made the case that players' records (in college, the minors or early in the majors) are a more reliable guide to their future production than the "tools" that they may sport. Closers are the exception to the rule. Here, tools are probably a more reliable guide than the record.
The Pirates are more of a tools team than record team. That may be why they are particularly good at identifying closers (and other relievers).
*Data and Statistics are courtesy of FanGraphs unless otherwise stated.*