We Should Account For Inherited Runners Better
On April 21, Grant Anderson inherited a hot mess. With the Brewers ahead 3-0 in the fourth inning, starter Kyle Harrison lost his feel. He walked Riley Greene and Spencer Torkelson in two uncompetitive plate appearances, then gave up a rifled line drive on one of his slowest fastballs of the day, a center-cut cookie to Hao-Yu Lee. Pat Murphy called Anderson in from the bullpen to face the bases loaded with no one out.
Anderson delivered nearly flawlessly. He got Javier Báez to ground into a first-pitch double play, then struck out pinch-hitter Kerry Carpenter to escape the inning with only a single run allowed. That run, of course, went on Harrison’s ledger. Anderson got credit for a scoreless inning, no more or less.
On May 16, Chase Silseth tried to pull off the same trick. José Soriano fought through five strong innings against the Dodgers, but he didn’t have it in the sixth. After an inning-opening groundout, he walked four of the next five batters and hit the fifth, driving in two runs and leaving the bases loaded. Silseth came in to put out the fire – but he might as well have poured kerosene on it. He hit the first batter he faced, then gave up a two-run single immediately after, pushing the score to 6-0. He finally got the last two batters of the inning – which meant that in the game’s official log, he pitched two-thirds of an inning and didn’t allow a run.
These two pitching performances went quite differently. Anderson had a tougher task and performed better. But the two of them each got credit for a clean sheet. This is far from the only problem with the way we calculate ERA, but it’s one that stands out to anyone following. Anderson and Silseth didn’t deserve the same counting statistics there. Likewise, Soriano got tagged for three runs, while Harrison got tagged with only one. But that didn’t reflect what happened to them – both of them lost it and had to be removed from the game because of all the runners they’d allowed.
I didn’t pick Anderson and Silseth at random. Anderson has been good this year even by regular run-scoring stats, with a 2.76 ERA and 3.88 FIP. But he’s inherited 16 runners this season and allowed only two of them to score. On the other hand, he’s departed the game with six runners on base – four of those guys have scored. ERA is doing him dirty; he’s actually been incredible when it comes to preventing guys from scoring. Silseth, meanwhile, has inherited 34 runners – 15 have scored. He, too, has been relieved with six total runners on base. None of those six has scored.
Inherited-runner nonsense feeds into ERA. But you could imagine a statistic that doesn’t have that same limitation. Or, at least, I could – so I made one. Every time a pitcher leaves the game, he leaves with some run-scoring expectation. That’s what a run-expectancy table is. For example, after reaching a situation with the bases loaded and one out, teams have scored 1.62 runs per inning this year (excluding innings where the home team won the game before recording three outs). A pitcher who departs in that situation has allowed those runners to reach base, and has left his team with some expectation of those runs scoring, regardless of what happens afterwards.
You can’t exactly just credit those 1.62 runs to the pitcher, though. That’s because teams who have no one on base with one out have still scored 0.27 runs per inning this year. So the net expectation of those runners on base is 1.35 runs – the difference between how much teams score on average starting from scratch and how much they score on average starting with a tailwind.
I debited every pitcher who left the game in the middle of an inning by that difference – the difference between his exact situation’s run expectation and the run expectation with the same number of outs and no one on base. I then credited those runs to the reliever replacing him. When the actual runs did or didn’t score, I put those on the reliever’s ledger. I did this all in RA9 space instead of in ERA, because I didn’t want to deal with scoring decisions.
As an example, let’s take Soriano and Silseth. Soriano allowed three runs before being removed with one out in the sixth. I debited him an additional 1.35 runs for the state he left the game in. That works out to 4.35 runs allowed. Silseth came in with 1.35 runs of credit, but then he allowed three runs to score before closing out the inning. My method says that he “allowed” 1.65 runs in that two-thirds of an inning, the difference between the three runs that actually scored and the 1.35 runs we’d already assigned to Soriano.
In Anderson and Harrison’s case, I used the same method and got some interesting results. Harrison left the game with the bases loaded and no one out, or 2.08 runs worse than leaving with the bases empty. I debited him with those runs. Anderson started 2.08 runs to the good and allowed a run to score. That means that in his inning of work, he “allowed” -1.08 runs. In other words, I gave him a negative “RE-RA9” — my name for this run-expectancy adjusted stat. Oh, one last minor adjustment: In extra innings, the pitcher who starts the inning gets some negative runs added to his ledger for the ghost runner. A scoreless extra inning is actually a little bit better than not allowing a run in a regular inning, so I accounted for that.
Anderson benefits enormously from this re-framing of runs allowed. His 3.07 RA9 is good to begin with. But that’s dragged down by the runners who reached base against him scoring after he’s removed and by the odd ghost runner, and it doesn’t credit him for the excellent work he’s done in preventing inherited runners from scoring. Account for all of that, and he’s pitched to a 1.72 RE-RA9. That’s one of the largest gaps in baseball.
Silseth has the opposite issue. His actual RA9 is a sterling 1.77. But that ignores the fact that none of the runners he’s allowed to reach base have scored, as well as the fact that he’s been atrocious at preventing inherited runners from scoring. Add all of that together and his RE-RA9 is 3.61. That’s not a bad mark – but it’s a lot closer to how he’s actually pitched than a one-handle number.
Articles like this always need tables, so here are the 10 relievers who benefit most from this way of assigning runs:
| Player | RA9 | RE-RA9 | Diff |
|---|---|---|---|
| Pierce Johnson | 3.27 | 1.52 | -1.75 |
| Gordon Graceffo | 2.41 | 0.69 | -1.71 |
| Brant Hurter | 3.20 | 1.57 | -1.63 |
| Alex Vesia | 2.70 | 1.20 | -1.50 |
| Kevin Kelly | 2.27 | 0.83 | -1.44 |
| Clayton Beeter | 3.38 | 2.00 | -1.38 |
| Grant Anderson | 3.07 | 1.72 | -1.34 |
| Orlando Ribalta | 1.80 | 0.47 | -1.33 |
| Isaac Mattson | 3.86 | 2.56 | -1.30 |
| Fernando Cruz | 2.01 | 0.75 | -1.26 |
Of course, the other side is just as interesting:
| Player | RA9 | RE-RA9 | Diff |
|---|---|---|---|
| Chase Silseth | 1.78 | 3.61 | 1.84 |
| PJ Poulin | 3.08 | 4.77 | 1.69 |
| Greg Weissert | 4.00 | 5.67 | 1.67 |
| Tim Herrin | 2.92 | 4.56 | 1.64 |
| Evan Sisk | 1.80 | 3.23 | 1.43 |
| Jake Woodford | 6.94 | 8.32 | 1.38 |
| Ethan Roberts | 2.14 | 3.38 | 1.24 |
| Calvin Faucher | 4.25 | 5.48 | 1.23 |
| José Suarez | 5.71 | 6.82 | 1.11 |
| Gus Varland | 4.73 | 5.82 | 1.09 |
Woodford’s line stands out. He’s been handed eight inherited runners. A whopping six of them have scored. And they weren’t even inherited in particularly difficult spots – he’s just had particularly poor luck at preventing them from scoring.
A few odds and ends: Among relievers who have thrown 20 innings, Robert Suarez has the lowest RE-RA9 in baseball, a sterling 0.35. Justin Lawrence has the worst RE-RA9 at 8.70, but that’s because he has an 8.54 RA9. This statistic mostly looks like actual ERA – most of a pitcher’s runs are rightly credited to them already. But if you’re wondering why a particular pitcher’s ERA doesn’t match up with what you think of him, this is a likely culprit.
Starters don’t have the same kinds of swings that relievers do in this accounting, because far more of their statistical record covers innings they both started and finished. But that doesn’t mean there’s no effect. Anderson’s teammate Brandon Sproat has had the worst luck with the runners he’s bequeathed to relievers who enter the game after him – he’s left 11 runners on base, with an aggregate run expectation of 3.5 runs, but seven of those runners have scored. His RE-RA9 of 5.18 isn’t exactly great, but it’s much better than his 5.70 ERA:
| Player | RA9 | RE-RA9 | Diff |
|---|---|---|---|
| Brandon Sproat | 5.70 | 5.18 | -0.52 |
| Rhett Lowder | 4.79 | 4.28 | -0.51 |
| Freddy Peralta | 4.12 | 3.62 | -0.50 |
| Spencer Strider | 5.54 | 5.08 | -0.46 |
| Trevor McDonald | 4.85 | 4.46 | -0.39 |
| Noah Schultz | 6.05 | 5.67 | -0.38 |
| Erick Fedde | 5.05 | 4.70 | -0.35 |
| Connor Prielipp | 6.16 | 5.83 | -0.33 |
| Colin Rea | 5.47 | 5.14 | -0.33 |
| Landen Roupp | 4.72 | 4.41 | -0.32 |
On the other side of the ledger, Cam Schlittler and Dylan Cease stand out as good pitchers who are also getting great work from their relievers. Schlittler has left 12 runners on, with an expected 5.2 runs worth of scoring contained therein. Only one of those runners has scored. Only one of Cease’s seven inherited runners have scored. Sean Burke has left 11 runners on base and not a single one of them has scored. Not bad work if you can get it:
| Player | RA9 | RE-RA9 | Diff |
|---|---|---|---|
| Nick Lodolo | 5.21 | 5.73 | 0.52 |
| Sean Burke | 4.28 | 4.71 | 0.43 |
| Cam Schlittler | 2.12 | 2.55 | 0.42 |
| Christian Scott | 3.32 | 3.73 | 0.41 |
| Shane McClanahan | 3.52 | 3.85 | 0.33 |
| Noah Cameron | 4.63 | 4.94 | 0.31 |
| Dylan Cease | 3.31 | 3.61 | 0.30 |
| Patrick Corbin | 4.57 | 4.87 | 0.30 |
| Brady Singer | 6.05 | 6.35 | 0.30 |
| Brandon Young | 4.31 | 4.60 | 0.28 |
I don’t know how meaningful this statistic is in terms of predicting future performance, because runs allowed just isn’t all that predictive. But this version of a runs-allowed metric does a little bit better at allocating those runs, and thus I think I like it a little more. I’m not saying that you should stop using ERA, or that this is the new best statistic in baseball, or anything like that. But if you’re wondering who really deserves the runs when a reliever comes in and melts down with men on base, I think this formulation is the right way to think about it.
Ben is a writer at FanGraphs. He can be found on Bluesky @benclemens.
It’s great information to use, but almost like a ground ball pitcher with a bad defense behind him, It’s less luck-based than people think (if they’re looking at BABIP). Therefore a bit more predictive than we’d guess.
There will be exceptions, but when taking starting pitchers, looking at their bullpen matters.
I like the analogy of having a bad defense, except even more environmentally determined. There are occasionally pitchers who do outperform or underperform their estimators even when they switch teams or the team gets rid of a whole bunch of bad / good defenders at one time.
But I am having a really hard time thinking of a case where this sort of “talent” for not letting runners in that you left on base after you left the game is attributable to the removed pitcher.
It might be predictive but if so it is because the starter is in the same environment being followed by bad relievers from year to year. The “bad luck” here should be due to your team not being good at developing or finding bullpen arms.
I think Backdoor Slider’s point is more that a starting pitcher with a bad bullpen behind him will continue to have a bad bullpen behind him for the rest of the season, unless he’s traded or the bullpen is really overhauled at the trade deadline. Thus, that starter is likely to continue having a lot of his inherited runners score over the course of the rest of the season rather than for his ERA to regress towards his RE-RA9 like you would normally expect.
But of course, bullpens are highly variable year-to-year, even if a pitcher stays on the same team, so the metric works better on a multi-year basis.
Yes—I would agree with that part. I thought I said that, but maybe I was unclear.
I wonder if this could be used as a first approximation for quantifying bullpen management
I like this and feel it deserves a catchier name than “RE-RA9.” Let’s see what we can do.
Runs Allowed Considering Impact of Substitution Technique (wait no let’s not use that one)
hmmm try again
Inherited Runner Adjusted ERA (IRAERA)
not quite
Baton Passing Adjusted Pitching (BPAP)??
Leverage Adjusted Run Prevention (LARP)??
LARP might be a tough google search.
Well, a really high LARP might mean the pitcher actually is LARP-ing
Runs, Adjusted For Average Expected LOBs (RAFAEL). We just need to find a season where Rafael Soriano was huge by this stat.
I’ve always thought that an inherited runner on first who scores should go against the RP and any inherited runners on second or third who score should go against the SP. Not perfect, but more sensible than what we’re currently doing.
Either that or allow fractional runs. A baserunner on second would count as half a run for the RP and half a run for the SP. Doesn’t have to be exactly half, but that’d be nice and simple
Could be half. Sure. If you’re noting base occupation anyway, could allocate based on which base. A man left on 3rd is 3/4 the responsibility of the starter, 2nd base is 1/2, and a man left on 1B who comes around to score is 1/4 the fault of the starter.
There is a hitch, of course, when a baserunner survives through a second pitching change. The SP is pulled with a man on 1st. The RP gets two outs and allows a hit, moving the runner over to 3rd. A new RP comes in and allows that man on 3rd to score. Who gets how many runs on their record?
Noting the RE difference handles all of that nicely, though it is, of course, codifying a league average as an individual stat.
Seems straightforward enough, SP gets .25 runs, middle reliever gets .5, 2nd reliever gets .25
Yeah, exactly. Each base you allow a runner to take equals .25 runs against
I’d count it against the pitcher who let him on base and the pitcher who let him score, while the rare case of any intermediate pitcher is off the hook.
So a RP who comes in with the bases loaded and 0 outs could face two batters, throw six pitches, get a groundball double-play where the batter isn’t even credited with an RBI, and then a strikeout, and they’re getting .25 runs charged against them? 0.00 WHIP, ~0.50 SIERA, but a 2.25 ERA?
This is exactly what my thought process was. To flesh it out, my fractions were 1/4 for each base. An inherited runner on 3rd is charged 1/4 to RP and 3/4 to SP. 2nd base is 50/50. 1st base is 1/4 SP and 3/4 RP.
This article lays out a much more sophisticated calculation but same thought process
Or just count all inherited earned runs as half-runs given to each pitcher. We already have fractional innings pitched, so why not half-runs?
Not really. Fractional innings don’t math as fractions or have anything to do with fractions, it’s just shorthand for batters retired where .1 = 1 out or .333 of an inning, .2 = 2 outs or .667 of an inning, and .3 = 3 outs or 1.00 inning.
A pitcher can’t actually get a partial out, or get credit for a partial out.
.1 innings is definitely a fractional inning. It’s 1/3 of an inning.
It definitely maths as 1/3 of an inning as well. A pitcher who pitches .1 innings in 3 games has pitched 1 inning.
I don’t dislike it as a crude rule of thumb, I’d be in favor
This is certainly better math to use expected runs, but just for simplicity I always wondered about if you credited it by base instead of by run, i.e. you charge the pitcher .25 runs for each total base at the time of pitching change. So if an inherited runner scores from 3rd the new pitcher is assigned .25 run, and old pitcher .75, .5 each from 2nd, etc.
Issue with this is it erases the number of outs recorded when a relief pitcher comes in the game. There’s a big difference in expected outcomes between coming in with 2 outs & RISP vs. 0 outs & RISP
That’s already accounted for by IP in ERA calculations. In the event that the pitcher is allowed to finish the inning in both cases, the former situation is easier, but he only gets credit for 1/3 IP as opposed to a full IP in the latter case.
For example, if the relief pitcher allows an inherited runner (who is not there due to any error, passed ball, catcher’s interference, or extra innings) to score from 2B with 2 outs and then gets the third out, he would be charged with 1/2 earned run across 1/3 IP for a game ERA of 13.51.
In the latter, more difficult situation where he comes in with no outs and allows the runner from second to score while also getting 3 outs (1/2 ER over 1 IP), his game ERA is only 4.50.
If a reliever comes in with the bases loaded and no outs and allows two inherited runners to score, that reliever would get charged with a 6.75 adjusted ERA. Run expectancy in that situation is 2.08, so the reliever was actually a shade better than average. Do you see the disconnect here?
I knew Way Back Weissert would be on this list…
Currently allowing 50% of inherited runners to score this season. He’s well below average for his career at this – 34% when league average is ~28%. And that’s including an excellent 2/20 from his Yankees days, so he’s been even worse since joining the Red Sox.
I like it and like the suggestion from others about fractional runs based on which base the runner was on.
The other ERA thing that I would change is: The amount of unearned runs can only be the amount of hitters that got on/scored via error. I.E., I hate it when a fielder makes an error & then the pitcher allows 4 hits & 2 walks with 5 runs scoring & is credited with only 0 or 1 earned run.
I get it- the fielder made an error & no runs would have scored without it,but, that doesn’t excuse the pitcher then falling apart..but, ERA does just that. I know other stats account for this phenomenon, but, ERA is so ubiquitous, would be better if it did so.
You can’t really change ERA because we can’t go back and adjust everyone’s ERA throughout baseball history. This is a good idea but it has to be a different stat and not ERA.
I’m sure they could, actually. We have WAR retroactively calculated throughout history. This is more involved, yes, but with technology and algorithms at the level they are now, it’s probably not that difficult for the right people and programs to go back and re-calculate ERA back to its beginning
You mean, like starting a runner on second in extra innings?
I don’t think the zombie counts against ERA.
Correct, the zombie runner is an unearned run.
Its not really an unearned run in the normal sense – its an inherited runner (and thus isn’t charged against the extra inning pitcher), but since there is no prior pitcher to charge the run against, there is no one who “earned” it by virtue of allowing the runner to reach base.
Clearly, these runs should be charged to the one responsible for them. Hence two new stats – MRA and MRAA. The second A is Abetted.
MRA is a counting stat added to the karmic debt of the party responsible for this atrocity. MRAA is a rate stat based on MRA and charged to the pitcher as a completely separate category affecting no other stat.
For fairness sake, the initial run expectancy should be subtracted from MRAA and added to the karmic debt.
Every runner in the past that started on second in extra innings also got treated as an unearned run.
It’s not unprecedented, as the save rule has been modified since it became an official stat. The exact definition of an RBI and even a home run (which once counted balls that bounced over the fence, as well as different rules concerning the foul poles) have changed over the years, as well.
The important thing is that the stats are counted under the rules played under at the time, while any rule changes have to be considered when comparing stats across different eras.
So MLB very well could modify how ERA is calculated going forward. They just wouldn’t adjust anyone’s ERA from previous years.
Pitcher making an error not counting for him always seemed odd to me also. Well…who earned it then
I’ve been complaining about ERA calculations not adjusting for inherited runners for years. Drives me nuts!! I’d love a solution like this. Let’s make it happen 👏🏻 👏🏻
I too like this. Assigning a pitcher negative runs seems odd at first, but getting out of a bases loaded jam really is more difficult than pitching a clean inning, so giving the pitcher credit for that makes sense.
The starting pitcher tables are a clear indication of which players are liked by their teammates and which are not.
RE24 is the best!
If we had a RAA value on pitcher pages, we’d be able to calculate this ourselves.
* We’d be able to translate this into a true WAR analogue.
Great work. Now I’m curious who the greatest historical changes would be. Would looking at it this way have affected any Cy Young races? HoF cases?
I haven’t read this yet but isn’t this just RE24 . will read and find out
ah so you’re converting RE24 to an ERA scale?
So it seems!
No.
No, it is not RE24. RE24 uses a different reference point.
Excellent idea, Ben. You are the best!
This has me wondering if LOB% is an ability some relievers have or just noise year to year – can you prioritize building a bullpen of “stoppers” that clean up these messes more regularly than average?
For example, for much of Pierce Johnson’s career, he’s bee in the 80% range for LOB%, so he seems strong in this respect YoY.
Regardless of the stickiness, its a foundational component of reliever value…which both forms of WAR explicitly ignore.
Devil’s advocate though, is it kind of like RBIs? We see RBIs as valuable, but opportunity driven, and not sticky, so we don’t use it. LOB had those characteristics too.
It’s “kind of like RBIs” in that Ryan Howard’s RE24 outperformed his wRAA every single year, making WAR an inaccurate description of his value.
Unfortunately, FG doesn’t break down pitcher RAA the way B-Ref does, but a quick check shows me:
Jansen and Hoffman’s RE24 clears their RAA by 50 runs.
Wagner and Kimbrel’s did so 40 runs.
Chapman and Roberston’s have done so by 30 runs.
Miller’s did so by 25 runs (as a reliever).
Rivera’s did so by 20 runs.
Joe Nathan and Rollie Fingers actually underperformed theirs by 50 runs!
Someone could probably run a Stathead query on it.
I was just reacting to your comment beginning with “regardless of stickiness”. I think stickiness matters a great deal. But your comment about Howard shows that you do too.
I also thought you were talking about LOB%.
I think omitting a reliever’s success in stranding inherited runners from an assessment of his value is closer to regressing every hitter’s BABIP.
I’m team “stickiness matters for evaluating talent” and team “value is value,” haha!
We’re not going to solve this in the comment section (indeed nobody has put this to bed yet in several years of debate). But I think even when discussing value, something has to be shown as a repeatable skill or else it’s just a hollow form of value. Like saying a guy is valuable at playing craps for rolling several 11s. Well OK… but was that attributable to him or just randomness?
It boils down to whether you want WAR to be descriptive stat or a predictive one.
I’m happy to leave questions of stickiness to projection systems.
I disagree that this is what it boils down to. I want it to describe only things within the player’s control. It doesn’t have to predict anything.
So should offensive WAR be entirely based on xwOBA, because hitters don’t control fielders’ performances?
How exactly do we delineate between a pop up that’s caught up in the infield as opposed to the outfield – do we just treat the latter as random variation?
You’re gonna think I’m nuts. But I say yes. With the caveat that it wouldn’t be perfect, e.g. some players’ base running skill allows them to consistently beat xwOBA so you need to account for that. And other stuff too I’m sure.
I’d think it’s one of those baseball things where it’s a very real skill that pitchers have (or don’t have), but extremely difficult to accurately quantify because of how much noise you’d need to sift through
Not surprised to see someone like Grant Anderson fare well by this methodology.
Going back to 2023 Brewers pitchers are tops in all of MLB at (2nd in parentheses)…
75.7 LOB% (74.8 SDP)
+368.63 RE24 (+257.64 ATL)
+30.84 reliever WPA (+25.66 CLE)
-0.40 ERA/FIP gap (-0.28 CLE)
87 ERA- (91 PHI)
3.85 RA/G (3.99 SDP)
So 3.85 RA/G over their last 556 team games, 3.70 RA/G over 70 games so far this year (with a corresponding improvement from a 98 FIP- between 2023 to 2025 versus an 83 FIP- so far this year), 3.91 RA/G in 2025, 3.96 RA/G in 2024, 3.99 RA/G in 2023…and depth charts projects 4.23 RA/G rest of season (which to its credit is the second lowest projection of the 30 teams).
This kind of highlights another issue that Grant Anderson is representative of. Dan’s article mentioned that ZiPS didn’t assign enough playing time to Brewers players over-achieving their projections, but a bigger issue would seem to be that Brewers players change their true talent level faster than the projection systems are programmed to recognize…
2025-26 Grant Anderson
99 IP of 3.09 ERA | 3.85 FIP
FG DC: 4.18 ERA | 4.24 FIP
2024-26 Aaron Ashby
139 IP of 2.53 ERA | 2.77 FIP
FG DC: 3.51 ERA | 3.54 FIP
Abner Uribe career
146 IP of 2.53 ERA | 3.03 FIP
FG DC: 3.20 ERA | 3.38 FIP
Misio career
153 IP of 2.65 ERA | 2.52 FIP
FG DC: 3.22 ERA | 3.18 FIP
Chad Patrick career
173 IP of 3.48 ERA | 3.65 FIP
FG DC: 4.24 ERA | 4.21 FIP
2025-26 Kyle Harrison
101 IP of 3.02 ERA | 3.29 FIP
FG DC: 3.99 ERA | 3.96 FIP
[wasn’t with the Brewers in 2025, but his FIP and xERA both took nice jumps in his small sample that year before his big jump this year]
Logan Henderson career
48.1 IP of 2.23 ERA | 2.72 FIP
FG DC: 3.97 ERA | 3.92 FIP
[obviously a small MLB sample due for big regression, but hard for me to see Logan doing that much worse than what Chad Patrick has already done throughout his career]
2025-26 Brice Turang
959 PA of 128 wRC+
FG DC: 111 wRC+
2025-26 Jake Bauers
464 PA of 131 wRC+
FG DC: 119 wRC+
2025-26 Andrew Vaughn
365 PA of 149 wRC+
FG DC: 109 wRC+
[get that the sample sizes are still varying degrees of small here for these three hitters, but all have seen noticeable improvements in their underlying StatCast data too]
Jackson Chourio career
1331 PA of 120 wRC+
FG DC: 121 wRC+
[projecting no real improvement rest of season for a 22 year old off to his best start ever with a 157 wRC+ so far this year?]
This is great—an issue I hadn’t thought about with ERA but is so clear once you point it out.
I’ve been thinking lately that I’d love to see a pitcher’s “BaseRuns9,” still incorporating fielding/groundballs finding holes, but strip out sequencing. Given that we calculate BaseRuns for teams, shouldn’t be that hard. And for hitter stats (wOBA) we’ve leaned hard towards sequencing independence—I can see being interested in how a pitcher did “getting out of jams,” but I doubt there’s much predictiveness there.
Would make for a cool article like this one—whose ERA vs BaseRuns9 gap is largest. Could also call it “wOBA/wRC+ against”
I mean, there’s absolutely a skill component to LOB.
Pitching from the stretch is, prima facie, different to pitching from the windup – look at Javier Vazquez’s career numbers.
Here are leaders and laggards for RA9 – BaseRuns9 using a playing time cutoff of ~1 out per team game:
(1) Enmanuel De Jesus | 5.60 RA9 | 3.45 BR9
(2) Juan Meija | 6.75 RA9 | 4.82 BR9
(3) Tatsuya Imai | 6.43 RA9 | 4.78 BR9
(1) Jason Adam | 1.78 RA9 | 4.51 BR9
(2) Richard Lovelady | 3.45 RA9 | 5.91 BR9
(3) Kyle Finnegan | 1.87 RA9 | 4.27 BR9
You can also do it in terms of RA9-WAR minus BaseRuns WAR. The number in parentheses is the difference:
(1) Nick Martinez (1.22)
(2) Chase Burns (1.20)
(3) Aaron Civale (1.12)
(1) Bryan Woo (-1.33)
(2) Nolan Mclean (-1.20)
(3) Reid Detmers (-1.02)
This is great, Ben. Thanks for looking into this.
I’m not sure if you were already working on something like this, but I was one who asked you the question about this stat a few weeks back. This is exactly the kind of formulation I was looking for. Is this now available in the Lab or somewhere?
Also, I totally get not wanting to worry about scorer decisions, but if you were to revisit that question and work on a RE-eRA9, I think this methodology would actually lead to a better way to evaluate errors. Some runs are more unearned than others, and I think official scoring aways explicitly references the expected counterfactual – eg a single and then a two base error via an errant throw. And that means you could simply adjust by the run expectancy differential between the two scenarios (a runner on 1st vs a runner on 3rd, in that example) instead of the often distorting binary of earned vs unearned.
I remember running a similar exercise to help me figure out who is more deserving for 2021 NL Cy Young.
Bottom line is that Burnes left 15 baserunners and 3 of them scored while the expected run was 7.6.
(Notably, he exited a game with bases loaded and 0 outs, and nobody scored with 1.9 expected runs)
For comparison, Wheeler left 14 baserunners and 6 of them scored while the expected run was 4.3.
(Notably, he exited a game with men on 1st and 2nd with 2 outs twice, and both runners scored both times with 0.4 expected runs.)
Substituting expected runs for the inherited runners, their ERA changes to
Burnes 2.43 -> 2.67
Wheeler 2.78 -> 2.71
Given how close the voting ended up being, if this was a stat more people cared about, good chance that Wheeler would have won the Cy Young.
That year still frustrates me based on an analysis I saw in the comments here. The big debate was whether or not Burnes’ better ERA should overcome Wheeler’s extra innings. Someone did the math and discovered that Wheeler had essentially the same ERA as Burnes for the first 5 innings or so. And then (on average) Burnes would come out of the game and Wheeler would stay in and pitch 2 or so (on average) above average innings (but not as good as Burnes and Wheeler’s first 5). It seems crazy to penalize him for continuing to pitch better than most pitchers rather than being done for the night, but that’s essentially what voters did.
I think I’m mainly annoyed because Wheeler will probably never win the cy young despite being (so far and probably eventually) the best pitcher of the decade.
I’ve spent a long time thinking about this over the years, but I always end up arriving at WPA or something similar. To me that’s the ultimate measure of situational impact for a reliever, and if we’re trying to adjust for inherited baserunners we might as well just start from scratch.
I just don’t like a measure of value that says coming in with the bases loaded and one out in the 5th isn’t as important as coming in with two outs and a man on third in the 8th, assuming the same score.
WPA is very much a “just for fun” stat. As in, it is literally there to quantify fun. The “swords” stat on Baseball Savant is another one.
Fully agreed – it’s the batting average of “value” stats.
Interesting that Nats have 2 relievers on each of the best and worst list. Feels right with respect to Ribalta, Varland, and Poulin. The latter 2 aren’t firemen. The Nats have one of the worst bullpens by rep and many measures, but Butera has navigated around that pretty well, moving Beeter to high lev stopper use and relying more on Ribalta for getting out of deep voodoo. Poulin is the primary opener, for which he seems suited
Not a huge fan of this on the starter side (as a former pitcher, it just feels like those runners are your fault because you didn’t do your job), but I do like the general concept on the reliever’s side – at least as a positive metric when they get out of the inning with minimal to no damage.
Maybe we need a statistic like Inherited Runner LOB%? BLOB% (Baserunner)? SLOB% (Stranded)? SHELOB% (Stranded Hitters Eliminated)?
From what I understand, the “starter side” is blaming you the correct amount for leaving the runners on.
Yes. Seems alderman missed the point.
A starter’s ERA can depend wildly on what happens with the runners they left on base and how the offense performs against other pitchers.
This brings up a similar if less common and maybe less important situation as well where you could apply a similar metric (and I’m curious if it is included in the analysis).
You go through the scenarios where there is greater run expectancy than normal, but what about relievers who enter with an objectively easier task? Should we account for the fact that when a reliever enters with one or two outs instead of zero, he has to record fewer outs before allowing runs?
If you come in with two outs and nobody on and proceed to load the bases, you can still just record an out, any out, and come away with a “clean” performance.
Would this metric punish that reliever? Should that also be considered?
Every time you add a runner, you’re increasing the run expectancy…and then if you get that out, you’re erasing the run expectancy you’ve created yourself, so it ends up a 0 R 0.1 IP.
Was something like this used to re-evaluate Kershaw’s playoff “struggles” or am I old-man hallucinating that?
This is an excellent article. So well thought out and made me say “how did I not think of this before?” Well done and thank you
I agree that it’s not fair to assign all inherited runs to either pitcher.
Your method may work as an advanced metric, but like all advanced metrics it’s overly complicated in order to be more precise, and of course it doesn’t account for errors, which are still important to consider even in an otherwise defense-neutral context.
I also don’t agree with you giving default negative value to extra inning runners who aren’t any pitcher’s fault for being on base. Therefore, they should either be treated neutrally like they are as unearned runs for ERA, or as positive value whenever a pitcher prevents him from scoring.
As a simpler solution (the OPS to your wOBA, if you will), I suggest just assigning half of each inherited earned run to each pitcher and then calculate their ERAs like usual. We already have fractional innings pitched (while the NFL has half-sacks), so the idea of half-runs shouldn’t be an issue.
Didn’t Michael Wolverton or someone create something like this decades ago at BaseballProspectus? Sadly, it never really caught on, but I don’t know if we have to reinvent it,
While Googling for it, I did find this FanGraphs Community post from 10 years ago which basically does this too: https://community.fangraphs.com/wera-rethinking-inherited-runners-in-the-era-calculation/
Here’s the Wolverton article – darn I’m apparently old:
https://www.baseballprospectus.com/news/article/257/what-the-r-column-doesnt-tell-you-evaluating-relievers-by-prevention-of-expected-runs/
RP comes in with a runner on first. He scores, 3/4 of a run goes to the RP, 1/4 of a run to the SP. Runner on second? 50/50. Third base? 1/4 RP vs 3/4 SP.
It’s not perfect because the number of outs plays a major role, but at least it’s easier math to figure out when scoring a game.
The version of this that I imagined when I young was a much simpler and I wonder how it would compare – just assigning runners a fractional weight based on which base they occupied at the time the reliever came in. So, for instance, if a runner on 3rd was inherited and then scored, 3/4 of that run would be assigned to the original pitcher and 1/4 to the reliever. It’s entirely vibes based, admittedly, but I liked that it mapped on to the bases themselves and felt more intuitive than the existing system.
Does this correlate with itself better than era? How well does it or era correlate better with estimators like fip. I like this much more than ERA. Also you could use this methodology to handle unearned runs much better, just back out the added run expectancy from the error from the pitchers ledger.
I’d be curious how to (or if you even could) account for fielder’s choice runners in these instances. If someone comes in with a runner on first and no outs, get a ground ball that results in the lead runner out a second, and then the resulting runner on first comes around to score, it’s all attributed to the reliever, even though that runner would not have been there if the original pitcher hadn’t left someone on base.
Obviously a niche scenario that probably doesn’t happen all that often, but it’s an edge case where the reliever is actually punished by coming in with runners on base.
I’ve always thought the each pitcher should only be liable for how many bases a runner advanced on that pitcher. In the bases full example, if the guy on 3rd scores on a relief pitcher, the relief pitcher gave up .25 runs. The initial pitcher .75. If the guy on 2B scores. .50 for each pitcher and so on. It works for situations with more than two pitchers . Each pitcher is responsible for the bases advanced to scoring.
I have always thought about another aspect that unfairly favors the relief pitcher; he can record outs from the inherited runners, but if those same runners score they are credited against the starter. I always thought, at least in your example with the double play, that at least one of those outs should have gone to the starter. The other big element is getting to come in in relief mid-inning, and not having to get three outs in the inning which is a major advantage to your ERA statistics.
“I’m not saying that you should stop using ERA”
But, we should stop using ERA. What value does it provide as a stat?
I feel like I read an article like this like a decade ago. Maybe it was here at FG. Maybe it was Tango? Maybe somewhere else?
I thought the Nats starters had so many runs credited to them last season because of the bull pen was bad at holding inherited runners. Weirdly this season its does not seem as glaring. It would interest me to find out if I was right last year. How much of the starters era is from runs while the relievers are on the mound?
Oh my god i feel heard for the first time ever.
I have been screaming this from a mountain for years. Justice for Jacob degrom, whose mid 2010s would be that much more impressive!
The worst difference relievers are a murderers row of guys whose fans hate them, yet somehow inexplicably still get high(ish) leverage innings.
Great idea
I always try to use Inherited Runners Scored to measure the effectiveness of Relievers, but I can never find the stat on FG. Am I just missing it? or is it called something else?
Love to see this coming back around. This is nearly identical to Runs Prevented, a stat I invented in 1994, and which was published in STATS Baseball Scoreboard from 1997-2001.
Awesome idea for one new valuable statistic. I hope every agent will take this into account in their negotiations on behalf of their clients. I also the Cardinals do the same at least when it comes to to players not yet playing with the Cardinals😎. Once again, this is the type of articles FanGraphs should support, encourage and incentivize..
I like this alot. Do the numbers for relievers in your table include a runs debit if they are pulled mid-inning?
Yep!
FWIW, I think “ReeRah9” would be a great fantasy stat. I know I’m not alone at feeling frustrated when one of my pitchers is removed mid-inning with runners on, but then the next pitcher allows those runners to score (frustration increases with number of outs when my guy was removed :)).