What is the best stat for judging relievers?

[sweepscc]

I wanted to find if Melvin has ever put together a very good top 10 bullpen in his career as GM. I think Melvin had one crucial chore this off season and that was to put together a bullpen that was playoff caliber. I understand his philosophy of retreads and maybe one high paid reliever. I also understand he's been crunched because we haven't had an abundance of high caliber arms come through the farm system. I just wanted to shed some light on his past performance in regards to constructing bullpens. Maybe he's been solid I don't know.

[dlk9s]

Off the top of my head, I would think WHIP is a good stat by which to judge relievers. Basically, the fewer runners you allow per inning, the less likelihood there is of allowing runs. This would probably not apply as well to when relievers come in with runners on base compared to starting an inning, as they could allow runs (and thus "not do their job") without allowing a hit/walk.

[InsomniacInkRoss]

I would agree that WHIP is the closest to best

 

Until they develop a better stat which I am sure they are working on

 

The TBAtROBaNBF

 

Total Bases Allowed to Runners On Base and New Batters Faced

[Brawndo the Thirst Mutilat]

What do you want to know?

If you want to know how good a reliever HAS BEEN, particularly in a single season, WPA is a good one. The problem with ERA is that it assigns relievers no blame or credit for one of the most important aspects of the job - stranding inherited baserunners. It sounds like you want to evaluate how well Melvin's past bullpens have performed. I'd be tempted to use WPA if that's what I was doing.

Stay away from WPA if you want to project future performance. It is not a predictive stat. If you want to know how well a reliever is going to pitch in the future, I'd look at FIP here or xFIP, which normalizes the rate of homers allowed, here (plain old FIP is available as well).

Finally, if you want to know how effectively a manager employed his relievers, use leverage index, particularly gmLI, which gives you the average importance of the situation when a pitcher entered a game. If the best relievers do not have the highest leverage index numbers, the manager was bringing in the wrong guys in the most important situations.

[DrWood]

a reliever has to by a gritty gamer. everything else pales in comparison.

[Fear The Chorizo]

I think limiting walks is the key to a successful reliever and a successful bullpen - force a team to put together more than 2 hits in an inning to generate a crooked number on the scoreboard. I think the Brewers' current bullpen walks to many hitters, and puts themselves in terrible situations where 2-3 runs can be scored in an inning with 1 hit.

[TooLiveBrew]

WHIP is dicey for RPs, since they operate over so relatively few IP. I'd agree with BTTM that FIP & xFIP are decent measures.

[Thurston Fluff]

given how many of them are great for a year or two then fade away I'd day the most recent trend is as good as any. Releives may be the one area where stats are not good predictoers of the future since they never really get enough time in any given season to have anything but a small sample size. Maybe the best stat is longevity. If they managed to stick in the bigs for 5+ years they must be doing something right a majority of the time.

[DrWood]

given how many of them are great for a year or two then fade away I'd day the most recent trend is as good as any. Releives may be the one area where stats are not good predictoers of the future since they never really get enough time in any given season to have anything but a small sample size. Maybe the best stat is longevity. If they managed to stick in the bigs for 5+ years they must be doing something right a majority of the time.

 

Hmmm. Carnac says the question is: "What is the best way to assemble an expensive bullpen?" http://forum.brewerfan.net/images/smilies/wink.gif

[rluzinski]

WHIP over one season is a really bad stat for a reliever. There's just too much randomness with balls in play over 60 IP, or whatever. WHIP in general isn't all that, as it treats a single and a HR as the same thing.

 

If we are talking about estimating the skill level of a reliever, we shouldn't just restrict ourselves to the stats from this year. I'd start with their strikeout and walk rates over the last 3 season or so. Then, I'd look at their FB/GB ratio and make some inferences about their HR rate from that. I'd probably just ignore their BABIP, since I don't think I could infer much from it anyway (sample size!).

 

If you want to just know how their performance has been this year, look at runs allowed and inherited runners allowed (accounting for the base and out situation).

[DrWood]

If you want to just know how their performance has been this year, look at runs allowed and inherited runners allowed (accounting for the base and out situation).

 

runs allowed, or earned runs? errors could have a big impact on the small sample size. The rate of ER, is ERA, so is that what it comes down to? ERA and % of inherited runs allowed?

[Brawndo the Thirst Mutilat]

On the question of RA vs. ERA, I find myself increasingly wishing that the latter would just drop out of existence entirely, whether you're talking about starters or the pen. Errors are assigned, it seems, in an increasingly bewildering fashion with each passing season. The thinking behind the creation of the stat was sound - you want to somehow be able to divide the blame for runs allowed between bad pitching and bad defense, but I don't think using errors as your signpost for bad defense really gets you anywhere.

If you want to cleanly separate pitching and defense, we've got a much better metric these days - FIP. The only problem is that there is a small element of pitcher skill being omitted - some pitchers really do have BABIP skill. But over the short haul (say, less than 5 full 175 IP seasons), the signal to noise ratio on that skill is so low that, statistically speaking, you're better off just regarding BABIP allowed as 100% luck, at least for the purposes of projecting future performance. So, FIP is a better predictive stat than ERA.

If you want a descriptive measurement that tells you how valuable a pitcher was over some stretch (say, for Cy Young voting), I prefer RA to ERA because while both have a ton of "but, how much of that is fielding?" noise, at least with RA you know that up front (no claims are being made to separate pitching and fielding). In my opinion, the extra information contained in ERA (how many of the runs "wouldn't have scored" if it weren't for the strangely-assigned "errors") is very close to completely worthless (predictively AND descriptively), so I'd rather just use the metric that is a purer measurement of something (RA).

Since this isn't a very big deal (ERA tells you very close to the exact same thing as RA), and since the mainstream loves its traditional stats, you won't see too many stat geeks digging in on this particular point. Gotta pick your battles, after all.

With relief pitchers, I think that the best descriptive stat needs to take into consideration how many runners you put on and allow to score (or leave to the next pitcher), how many more/less of the runners you inherit score than would be expected, and how important the situations you pitch in are. WPA ain't perfect, but all of that is in there.

[joepepsi]

Doesn't it come back to K:BB ratio or K/ip? Isn't the best thing a reliever can do is take the ballgame into his hands by striking everyone out?

[rluzinski]

Why use ERA when you just basically want to know the results of their pitching (as opposed to estimating their actual skill). ERA is just dumb, anyway.