Beyond a handful of reliable hurlers that manage to be consistent in their production, projecting what a reliever is going to do in any given year can be a risky proposition. For every Mariano Rivera or Jonathan Papelbon, you have dozens of inconsistent Joe Borowski or Lance Carter types to sift through every year. Luckily, there are some sabermetric tools that you can use in order to identify which pitchers are capable of consistently delivering the numbers they put up, and which ones are better left on your league’s waiver wire, notwithstanding any lofty save totals they may have.

Here’s an example of the kind of differences that can be found using nontraditional statistics: Brad Lidge threw 69 1/3 innings last year, with 41 saves, a 1.95 ERA, and a K/BB of 2.6. Joakim Soria threw 67 1/3 innings, racked up 42 saves, and posted an ERA of 1.60 with a K/BB of 3.5. Soria had fewer strikeouts but better control (and therefore a lower WHIP), hence the significant difference in K/BB, but otherwise the two are very similar. It’s hard to believe, but Lidge was actually two wins better than Soria last season based on BP’s Win Expectancy stat. WXRL is Win Expectation above Replacement (adjusted for the batters faced), and can be used to determine the value of a relief pitcher. Lidge led the majors in WXRL with 7.6 in 2008, while Soria came in fourth with 5.4 despite the similarity in their surface statistics; Lidge’s opponents combined for an OPS of 770 (fifth in the league, minimum 50 innings pitched) while Soria’s came in 20 points lower (114th).

Now you wouldn’t say no to either of these closers on your fantasy team, but it’s good to know that beyond the traditional stats used in fantasy, there are more advanced numbers that take into account factors-like the quality of batters faced-that can help you project future performance. Each year, you end up with relievers who perform over their heads due to the way they were used, who they faced, or just plain luck resulting from the small sample sizes that go hand in hand with working in relief.

In addition to WXRL, you can also use Leverage, which also uses win expectancy, though it does so in order to measure the important of the game situation. Situations that are less crucial, such as pitching in relief when your team is up by more than a few runs or when the game is out of reach, are given a leverage score under 1.00, while key situations (bottom of the eighth, bases loaded, the team pitching is up by two runs) are given higher leverage scores. Leverage scores can help you to see who is pumping up their numbers by pitching in less critical circumstances. Joe Nelson is a good example of this; he posted some impressive numbers pitching out of the Marlins‘ bullpen in 2008, with an ERA of 2.00 and 10 strikeouts per nine over 54 innings pitched. If you take a look at his Leverage and WXRL, you see a different story, since he was not utilized with any consistency in significant situations; his Leverage was just 1.09 (ninth on the Marlins), and he wasn’t much above replacement level as a reliever, compiling an 0.87 WXRL despite pitching in 59 games. Combine that with the fact that his adjusted ERA figures are higher than his actual numbers by a considerable amount, and you can see that he isn’t someone likely to repeat his performance without some help.

Leverage can also be used to explain blown save totals for some pitchers. Francisco Rodriguez shattered the record for saves in a single season last year, but he also picked up seven blown saves along the way. He had 70 save opportunities in 76 games pitched, but the Angels still utilized him in important situations-his Leverage score for the year was 2.21, and his opponents’ OPS was 764, tied for the 27th highest mark in the majors. While he most likely won’t approach those save totals again, his overall numbers were probably hampered by his opposition, and those blown saves had more to do with the way he was used than they did with Rodriguez’ ability.

On the flip side there’s Jose Valverde, who had to face even tougher competition (opponent OPS of 773, the third-highest), but who was used in situations that were much easier for him, with a Leverage score of 1.39. Like K-Rod, Valverde also blew seven saves, and despite not pitching in anywhere near as many high-leverage situations as Rodriguez, he allowed 3.1 more inherited runners to score than a league-average reliever would have (while Rodriguez allowed 2.6 more to score). For some perspective, that puts Valverde in the same neighborhood as set-up men Duaner Sanchez and Ryan Franklin, neither of whom are closers at present (nor should they be). While Valverde’s season was certainly not bad, if he were to be used in higher leverage situations going forward-situations that will most likely result in his continuing to face stiff competition-his numbers would assuredly suffer for it. It’s a good thing to keep in mind when you need a tiebreaker on draft day between two closers, or if you’re shuffling your roster around during the season.

This is also useful for leagues where holds are a statistic, as these numbers are available for all relievers. For instance, Jeremy Affeldt increased his strikeouts and dropped his walk rates significantly, turning in one of the better seasons of his career last year while with the Reds. Take a look at his Leverage and WXRL though, and you see that part of what helped him was his use in less-critical innings (Leverage of 0.63, WXRL of 0.22). If the Giants use him in higher leverage situations this year against better hitters, we may see his numbers head downward again, and you won’t pick up the holds you were expecting from him in your fantasy league.

Projecting relievers is difficult, but it isn’t entirely impossible. With the right tools and a thorough approach-stats like Leverage, WXRL, and the like are available for your perusal all season long and get updated daily-you can make some headway and outfit your team with a reliable core of closers and set-up men.

A version of this story originally appeared on ESPN Insider Insider.