clock menu more-arrow no yes

Filed under:

The best relievers of 2014 per RE24

New, 1 comment

There are many ways to get at a reliever's true talent level. But in addition to telling a story, RE24 can help us get at a reliever's true value.

Hannah Foslien

Relief pitchers are difficult to evaluate for several reasons. Sample size is obviously one of those things, and the small size of relievers' workload makes it difficult to root out true talent level. In addition, it's not always easy to tell how much a pitcher's true talent level actually matters as compared to his results; a LOOGY, for example, might do significantly worse if asked to face right-handed batters at a league-average frequency, but he's extremely unlikely to be put in that situation long term.

We can strip context away entirely with FIP, and we can put context back in with an ERA estimator like SIERA. But most pitching statistics ignore inherited runners, and overall, relievers inherit more runners than they leave on base. What they do with runners on base does matter, and it can't be attributed only to luck, or only to those pitching skills (like strikeouts and walks) that we normally care most about.

That's where WPA/LI and RE24 come in. WPA/LI is a counting stat that takes a player's Win Probability Added and divides it by the Leverage Index. Judging a reliever on WPA alone can be misleading, since an inferior reliever with inferior peripheral results could be placed in higher-leverage situations more often and still get a higher WPA than a teammate who may be the better pitcher. Using the Leverage Index counteracts that aspect of WPA.

But both WPA and LI are linked to a game's score. WPA/LI is one answer to the question of "what relievers have been most successful?" — and the answer the stat gives you helps tell the story of whether a pitcher was especially successful in important times in important ways, above and beyond simply being in the right place at the right time.

The story told by RE24 is more straightforward. Instead of being based on win expectancy, it's based on run expectancy, and is not dependent on the context of game score. Using a run expectancy table (here's one based on 1950-2010 data from Tom Tango), we can tell that on average, teams have scored 1.211 runs when they have runners on first and third with one out. If a reliever came in for that situation and recorded an inning-ending double play, he'd get an RE24 credit of 1.211. If the same reliever recorded a strikeout and then a force out, he'd still get a credit of 1.211. But if the reliever came in for that situation, walked a batter, and then was pulled, he'd get .42 taken away from his RE24 total for coming into the game in a 1.211 run expectancy situation and leaving it in a 1.631 run expectancy situation.

Unlike WAR, then, RE24 is based on what's average, rather than what's better than what Joe Shmoe from Triple-A might do. Theoretically, at least, RE24 is zero sum. Indeed, the total league average for RE24 so far this season is 31.00, which in the context of 25,586.1 innings pitched is essentially zero.

The RE24 leaderboard for relievers tends to be positive, though (188.33 so far this year), just like other pitching statistics (0.33 difference for ERA so far this season, 0.25 difference for FIP). And if you head on over to FanGraphs, keeping the relievers page on "qualified" will make the list look even more positive, as the crappiest relievers tend to get few innings.

Still, 53 of the 155 "qualified" relievers have negative RE24 totals. So it's worth keeping in mind: having a positive RE24 at all means you're probably pretty decent at your job. Here are the top performers so far this year:

Name Team G IP SIERA WPA/LI RE24 RE24/IP
Dellin Betances Yankees 40 55.1 1.45 2.02 22.10 0.399
Darren O'Day Orioles 40 40.2 2.69 0.89 15.98 0.393
Zach Britton Orioles 41 48.1 2.01 1.63 15.97 0.330
Jake McGee Rays 45 41.1 2.08 1.49 14.74 0.357
Pat Neshek Cardinals 43 38.1 2.64 1.82 14.49 0.378
Fernando Abad Athletics 45 38.0 3.05 0.98 13.93 0.367
Wade Davis Royals 37 39.2 1.83 1.54 13.78 0.347
Koji Uehara Red Sox 42 43.2 1.58 1.02 13.75 0.315
J.P. Howell Dodgers 44 34.0 3.18 0.78 13.60 0.400
Santiago Casilla Giants 31 35.1 3.06 1.24 13.03 0.369

One thing that might jump out at you: several of these guys have fewer innings pitched than games played. There are at least two reasons for that. One is luck; the more of your appearances that come mid-inning, the greater the fluctuation that's possible in RE24. But that's not terribly interesting.

The other explanation is that several kinds of relief "specialists" can be especially good at handling certain situations. Let's say you had an excellent reliever who was just filthy against same-sided batters — and because you had several other great relievers, you had the freedom to use that excellent reliever in platoon situations a higher-than-average extent of the time.

Like platoon matchup guys, though, there can be relievers who are particularly efficient at recording outs in particular base-out states. With one out and just a runner on third, for example, you want that strikeout guy more than you want a ground ball reliever -- even if they might have the same true talent level per a statistic like SIERA. The opposite might be true, though if there was a runner on first as well, since elite GB% relievers can "induce" weak grounders a bit more often then strikeout men can record strikeouts.

The fact that SIERA and other statistics don't account for these base-out states makes me love RE24 for relievers. There's something here. Brad Ziegler most likely has extra value to the Diamondbacks because his "double play efficiency" can be used in double play situations a disproportionately high percent of the time. And that might just be an example.

Dellin Betances is almost lapping the field in RE24 so far this season. He's killing it. But he's also pitched 55 and a third innings, which helps to pad his RE24 total since it's a counting statistic. Out of curiosity, I ran a "RE24/IP" to turn into a rate statistic. I got the same ten leaders, except that Jonathan Broxton (19.92 RE24, 0.392 RE24/IP) and Rafael Soriano (12.64 RE24, 0.342 RE24/IP) joined the party. And if you sort the above table by RE24/IP, you find that the specialists do even better.

Again, evaluating relief pitchers is not easy. Because different pitchers' profiles behave differently in different circumstances, I do like SIERA. Note in particular how well Zach Britton does in SIERA, which is due in large part to a ridiculous 78.0% GB% so far this season (only five times since 2002 has a qualified reliever finished over 70%). But RE24 is another side of the coin to look at, because while other statistics might better point to a reliever's true talent level, RE24 can help us determine his true value.

. . .

All statistics courtesy of FanGraphs.

Ryan P. Morrison is a writer and editor for Beyond the Box Score, and co-author of Inside the 'Zona, a site on the Arizona Diamondbacks with a sabermetrics slant. You can follow him on Twitter: @InsidetheZona.