bump

He's a lesser version of Lee Smith and Smith hasn't gotten close to the HoM.

Franco has the better ERA+ (138 to 132) in a similar number of innings (1245 to 1289). However, he doesn't have nearly as many big years as Smith. Smith has a top season of 4.5 WAR ('83, using baseball reference's numbers) and 3 more of 3.0. Franco has a top season of 3.2 WAR in '88 and 2 more of 2.9. The career totals favor Smith as well, 30.3 to 25.8.

It is practically impossible to create an objective career measurement for relievers to make the Hall without putting Franco in. Total games (3rd), saves (4th), ERA+ (18th), excellent postseason record.

StHendu - I have Franco between Tug McGraw and John Hiller as the 9th most valuable reliever of all time.

I've detailed the system before, but I try to account for everything from team defense to inherited/bequeathed runners, to in-season league quality, leverage, etc.. Even hitting.

I don't account for chaining - but that shouldn't matter when comparing relievers to relievers.

I get this as my result for eligible relievers. I'll insert the true 50/50 SP/RP types without ranking them.



Considering Rivera added .097, .094 and .077 PA from 2004-2006, I'd assume he's #1 now and close to passing Eck even with his starting.

I never did figure Trevor Hoffman. As a proxy for team defense I used NRA and DERA from the BPro player cards (the difference between the two was a great indicator for team defense), but they changed (removed?) it awhile ago, and I haven't found a good proxy since, which is a shame.

Eckersley's breakdown:

SP, .645, 111, 2587.7, 43.1
RP, .522, 142, 1290.0, 35.0

He's basically Burt Hooton and a slightly less effective Tom Henke, rolled into one.

Also look at that list - it starts to tighten right after Smith. The distance between Sutter (who follows Smith) at #5 and Tekulve at #11 is less than the distance between Smith and Sutter.

For me Lee Smith is the line between in or out.

If you go below him, the door has to open wide. Above him it's pretty obvious they belong. He's sort of right in the grey area. I'm getting won over to his case more and more.

Joe, do you have a season-by-season breakdown for Lee Smith? I'd like to use your pitcher WARP to rate relievers, and I need information about Smith's individual seasons to do that (top season, top 3, top 5, top 3 consecutive, seasons above 5 WAR, seasons above 2 WAR).

No problem Alex, here you go . . . I adjust strike seasons to full-time. Also, I don't adjust for chaining, so I probably overstate leverage (meaning the number of adjusted innings). My replacement level is 5.48 R/9 and average is 4.50 R/9. In this system 7 or 8 WAR can win a Cy Young Award in a typical year where no one goes insane.



Smith comes out 4th when you compare him on a JAWS scale using my WAR for everyone eligible. He comes out 4th using Pennants Added. He's behind Gossage, Wilhelm, Fingers in JAWS and Wilhelm, Gossage, Fingers using PA.

Using a system similar to what Bill James uses in the NHBA, which is extremely peak heavy (WAY too much so, IMO), Smith comes out 8th. He's behind Gossage, Sutter, Hiller, Fingers, Wilhelm, Marshall and Radatz on that system.

Any system that rates Dick Radatz above Lee Smith is too peaky for my taste. :-)

Hiller's peak is truly phenomenal. His 1973 is one of the most valuable seasons of all time, starter or reliever. He saved 15.6 inherited runs from scoring, and his relievers cost him another 7/10 of a run. He only gave up 21 runs all year. He essentially pitched 125.3 innings at 1.72 leverage and allowed 4.7 runs. His DRA+ was 1180. That's not a misprint. He did this in an environment that had a 107 run factor. It has to be a record. I get him at 12.3 WAR, which is basically Walter Johnson 1914 caliber - in 125 innings.

Or if you account for leverage, he threw the equivalent of 207 innings with a 0.38 RA in a 4.50 RA environment. Even if you adjust for chaining - which wasn't nearly the issue with the smaller bullpens of the 1970s, it's really impressive.

Just for fun, here are all the 5+ WAR seasons by relievers, per Chone WAR

I'm afraid I can't follow your math on Hiller's '73, Joe. Walk me through?

No Mike Marshall 1974???

If Hiller allowed 21 runs, but prevented 15.6 inherited runners above what an average pitcher would have allowed, you can subtract those runs from the runs he allowed when figured his runs prevented.

So 21 - 15.6 - .7 = 4.7 effective runs allowed.

I have Gossage's 1975 really high too, but I like his 1977 even better. His LI in 1977 was 1.97, vs. just 1.45 in 1975.

I have Marshall's 1974 as high, but not crazily so. His LI that year was just 1.38 and he wasn't all that good, below average with inherited runners (-.7) and his own relievers saved him another 3.8 runs.

His 1973 was a more valuable season. I have them 7.9 WAR vs. 5.7. Mainly on the strength of him saving 9 inherited runs above average, and his relievers costing him 1.8 more. It's a +10.8 for inherited/bequeathed in 1973 vs. a -4.5 in 1974. That's a big difference. And in 1973 his run environment was 104 vs. 94 in 1974.

Wow, Mark Eichhorn slipped through the cracks, I never worked him up. Didn't get Doug Corbett either, although I remember his 1980 very well, it's the first year I paid attention to baseball.

I think Chone's WAR understates the big reliever seasons a bit, just looking at those numbers. I get Hiller at 12.3, Gossage 1977 at 10.6, Fingers 1981 at 9.3 as the 3 best reliever seasons (amongst those whose careers finished before 2006), although I did miss Eichhorn and Corbett.

I have Rivera's 1996 at 6.1, because his LI was a lot lower (1.30) than most ace seasons (he was a setup man that year), and we are closer there, so I wonder if leverage has something to do with it.

Joe Dimino, this sounds like a rather circuitous way to go about it. If you're trying to capture game-state impact by looking at inherited runners and leverage index, why not just use WPA?? Hiller's WPA in 1973 was 8.4. In the modern game, the chaining effects mean that reliever WPA is usually pretty close to true reliever WAR, but I'll give you that that was likely not the case in 1973, given that Hiller pitched enough innings to cover both a modern closer and a setup man, and that there weren't relief specialists floating around like there are today. So if we put reliever replacement level at, say, 1.5 wins below average per 200 IP (compared to a starter's value of 2.4 in my system), then Hiller comes out at 9.3 WAR. That is indeed an astonishing figure, far better than I would have thought possible for a reliever, and significantly higher than the 6.9 he's credited with by CHONE. It does suggest he was a legit rival to Blyleven as the most valuable pitcher in the league, and that both were significantly better than the top position player (Grich if you believe some fielding numbers and Carew if you don't). But it ain't 12.3, or anything approaching it. 12.3 is peak Bonds, dude.

why not just use WPA?

Even more fun, here are all the 6+ WPA seasons by relievers

WPA does not account for ballpark (Hiller's was a great hitter's park in 1973) or the defense behind the pitcher (Hiller's was quite below average).

What is wrong with giving the pitcher credit for the inherited runners he saved? Am I fundamentally botching something? Or does it just seem strange because Hiller was so phenomenal that year it tests the limits of the system?

And I would say that a pitcher effectively giving up 4.7 runs in 125 high leverage innings, in a hitter's park, with bad defense behind him, is "Bondsesque".

WPA doesn't account for ballpark?? It most certainly should, by adjusting the run environment used to determine win probabilities. Defense is a fair point--I'd probably just add that in assuming the pitcher got the same fielding support as his teammates and using some standard runs-wins conversion.

I think that when your numbers diverge from WPA by that big of a margin, it is prima facie evidence that something is gravely wrong with your system. My guess is that you are double-counting, because you are first giving full credit for leverage and then incorporating inherited runners separately, when inherited runners of course drive up a pitcher's LI. It seems crystal clear to me that the proper methodology is to start with WPA and then adjust for defense, league-contextual chaining, and (if it really isn't baked into the WPA calculation, which it should be) ballpark.

"Win Expectancy and Run Expectancy Stats" at Baseball-Reference

According to that explanation, BB-Ref's edition of WPA incorporates a single adjustment for run environment, which in turn depends on league-average scoring and the so-called park factors. In turn that may be a difference from Fangraphs' edition.

In part:
Now that we can get [Win Expectancy] and [Run Expectancy] for a particular run scoring environment what environment should we use for a given park[?]
There are a couple of options:
# Look at RS/27outs scored in that ballpark for the year. For example, Padres home games averaged 3.65 R/27outs in 2009.
# Look at the league run scoring for league home games. The 2009 NL home game averaged 4.48 runs/27outs.
# Take the league run scoring and then apply our park factor to it. The Padres PPF was 87, so we take 4.48 * .87 and get 3.90 R/27outs as our run environment.

Baseball-Reference.com uses the last technique. I can see arguments for any of the above, but I feel that since we are comparing a player's performance to an "average team" the average team scores 4.48runs/27outs and in Petco they would score 3.90 runs/27outs.

I know that most of these numbers are available at FanGraphs.com and that our numbers differ from theirs. I believe the difference lies in the differing run environments we use for each park.

Now that we have a run environment for every park, we can put a WE and LE (and WPA and RE24) on all 9 million plays in our database and add them up.

They're not eligible yet but I'd love to see how Tom Gordon and Billy Wagner fit into the chart in #4.

Yup, OK, so their WPA does account for park factor as I thought. Next objection?

WPA wasn't readily available when I built this many moons ago. B-R WPA, which I didn't realize accounted for run environment wasn't available either.

Didn't pick up that using LI and inherited runners would double count but that makes sense.