The Number
When a reliever enters a game with runners on base, those runners are now his problem. How often do they score?
Across the full 2025 MLB season: 68.3% strand rate. Roughly 1 in 3 inherited runners scores. The 2026 early season (through April 5) shows 67.4% — virtually identical. This is one of the most stable numbers in baseball.
But the average hides enormous variation — not across relievers, but across situations.
The Outs Cliff
The single strongest predictor of whether an inherited runner scores is how many outs there are when the reliever enters. Not which reliever. Not the inning. Not how many runners are on base. The outs.
responsiblePitcher; IR from boxscore.The gradient is steep and consistent across both seasons:
- 0 outs: 43.8% strand rate (2025, 611 entries). The majority of inherited runners score. The reliever needs to get three outs with runners already in scoring position.
- 1 out: 61.4% strand rate (1,449 entries). Closer to the league average. One out already recorded gives the reliever a fighting chance.
- 2 outs: 82.0% strand rate (1,928 entries). Only 18% of inherited runners score. One out to escape.
The 2026 data tells the same story: 38.5% → 61.7% → 79.0%. The pattern is not a 2025 artifact.
This is a descriptive finding, not a causal one. Zero-out and two-out entries differ in many ways — leverage, base state, which relievers managers deploy, how aggressive the batting team is. We can't say "pulling the starter at 0 outs causes worse strand outcomes." But the association is steep enough to be directionally useful: a manager weighing whether to pull a starter mid-inning should know that the inherited runner situation gets materially harder with fewer outs.
The Inning Doesn't Matter
Unlike the outs, the inning a reliever enters barely affects strand rate:
| Inning | Entries | IR | Strand |
|---|---|---|---|
| 4th | 210 | 347 | 70% |
| 5th | 603 | 993 | 67% |
| 6th | 1,028 | 1,595 | 68% |
| 7th | 982 | 1,518 | 70% |
| 8th | 731 | 1,205 | 69% |
| 9th | 272 | 466 | 65% |
65-70% across innings 4 through 9. Flat. The "high-leverage late innings" narrative doesn't show up in inherited runner outcomes. A 6th-inning fireman faces the same probability as a 9th-inning closer inheriting runners.
The Distribution: How Much Do Relievers Vary?
Before looking at individual names, look at the spread. Among 171 relievers with at least 15 inherited runners in 2025, the standard deviation is 12.6 percentage points. That's enormous variance around a 68.5% mean.
Distribution of individual reliever strand rates, 2025 season. Minimum 15 inherited runners (n=171). The shaded region shows ±1 standard deviation from the mean. The question is whether the tails represent skill or small-sample noise.
The distribution is roughly bell-shaped, but the tails are fat. A handful of relievers strand above 90%; a handful below 40%. The key question: are the extremes real skill, or is this what random variance looks like with 15-45 inherited runners per pitcher?
The Leaderboard (With a Warning)
Who stranded inherited runners best in 2025? The top 10 (minimum 15 inherited runners):
| # | Reliever | IR | IRS | Strand % |
|---|---|---|---|---|
| 1 | Max Lazar | 15 | 0 | 100.0% |
| 2 | Ben Casparius | 29 | 1 | 96.6% |
| 3 | Bryan King | 23 | 1 | 95.7% |
| 4 | Adrian Morejon | 45 | 5 | 88.9% |
| 5 | Tyler Ferguson | 29 | 3 | 89.7% |
And the bottom 5:
| # | Reliever | IR | IRS | Strand % |
|---|---|---|---|---|
| 171 | Joey Wentz | 22 | 18 | 18.2% |
| 170 | Colin Poche | 15 | 11 | 26.7% |
| 169 | Taylor Rogers | 32 | 18 | 43.8% |
| 168 | Andrew Saalfrank | 18 | 10 | 44.4% |
| 167 | Aaron Bummer | 30 | 16 | 46.7% |
The warning: These are season-long descriptive summaries, not skill rankings. Even Adrian Morejon's 88.9% on a large sample (45 IR) has a Wilson 95% confidence interval of [76.5%, 95.2%]. Joey Wentz's 18.2% spans [7.3%, 38.5%]. The intervals overlap substantially with the league average of 68%.
Is strand rate a stable reliever skill? We don't know yet. The cross-season correlation between 2025 and early 2026 is r = 0.098 (essentially zero), but the 2026 sample is far too thin (3-5 IR per reliever) to draw conclusions. This is an open question that needs a full 2026 season — or a multi-year analysis — to answer properly.
What This Means
The outs gradient is the actionable finding. It tells us that the entry situation is strongly associated with strand outcomes — more strongly than which specific reliever enters. A reliever entering with 0 outs faces a fundamentally different probability landscape than one entering with 2 outs.
For fans watching a game: when the manager goes to the bullpen with 0 outs and runners on, the inherited runners will likely score regardless of who comes in. When there are 2 outs, they'll likely be stranded. The reliever's identity matters, but the situation matters more.
Whether individual strand rate is a persistent reliever skill or largely noise is the biggest open question in inherited runner analysis. The data hasn't answered it yet. CalledThird tracks this nightly — by the end of the season, we'll have enough 2026 data to test properly.
Methodology
Data: MLB Stats API play-by-play endpoint for IRS attribution (responsiblePitcher field per scoring event) and boxscore endpoint for IR totals per pitcher-game. Entry context (outs, inning) from playEvents[0].count.outs.
Attribution: IRS derived from play-by-play responsiblePitcher events. IR from boxscore per-pitcher-game totals. The two sources are not perfectly reconciled — approximately 0.1% of entries show discrepancies. The method is substantially more accurate than heuristic computation from pitch-level data but is not a perfect per-entry decomposition.
Sample: 2025 full season: ~4,044 reliever entries with inherited runners across ~2,428 games. 2026 through April 5: 225 entries across 126 games. Six duplicated pitcher-game rows from rain delay/makeup games in 2025 (<0.2% impact).
Strand rate definition: Matches MLB IR-A%: (IR - IRS) / IR * 100.
Confidence intervals: Wilson score intervals on strand counts.
Limitations: Outs gradient is descriptive (association, not causation). Cross-season persistence test has n=44 with 2026 sample too thin for conclusions. Leaderboard uncertainty is wide even at 15+ IR.
Cite this analysis
CalledThird. "The Fireman's Dilemma." CalledThird.com, April 7, 2026. https://calledthird.com/analysis/the-firemans-dilemma
All CalledThird analysis is original research. If you reference our findings, data, or charts in your work, please link back to the original article. For data inquiries: hello@calledthird.com