WAR in Baseball, Explained
WAR in Baseball, Explained
What wins above replacement is trying to do, what replacement level means, and why Baseball-Reference and FanGraphs give different numbers for the same player in the same season. This page computes no WAR, because any figure it produced would be neither of theirs.
Which WAR you are reading
Explainer, not a calculatorA Baseball-Reference figure of 5.0 for a pitcher
What the two mainstream implementations do differently
| Baseball-Reference (bWAR, rWAR) | FanGraphs (fWAR) | |
|---|---|---|
| Pitching starts from | runs actually allowed | Fielding Independent Pitching |
| Fielding, modern seasons | Defensive Runs Saved, from 2003 | Ultimate Zone Rating, from 2002 |
| Fielding, earlier seasons | Sean Smith’s Total Zone | Total Zone |
| Replacement level | .294 winning percentage, a 48-114 record | .294, the same convention since 2013 |
| Park factors | a player-specific factor for pitchers, weighted by appearances in each park | park-adjusted, by its own method |
| Is the full method reproducible from published steps? | no — the site calls WAR “necessarily an approximation” | no — the defensive input is a proprietary model |
The same player-season, valued twice
| Player and season | Baseball-Reference | FanGraphs | Gap | Why, per the source |
|---|---|---|---|---|
| Sandy Alcantara, 2022, pitcher | 8.0 | 5.7 | 2.3 | ERA 2.28 against a FIP of 2.99 — runs allowed beat the fielding-independent line |
| Kevin Gausman, 2022, pitcher | 3.0 | 5.7 | 2.7 | the reverse case: the fielding-independent line beat the runs allowed |
| Brendan Rodgers, 2022, position player | 4.3 | 1.7 | 2.6 | the two defensive models disagree |
| Manny Machado, 2022, position player | 6.8 | 7.4 | 0.6 | a small gap, which is the more usual case for a hitter |
| Kyle Gibson, 2019, pitcher | 0.0 | 2.4 | 2.4 | runs allowed against FIP again, on a different season |
What this page will do with WAR, and what it refuses to
| Here | Elsewhere | |
|---|---|---|
| Explain what WAR is for | yes | — |
| Say which implementation you are looking at | yes, from the fields above | — |
| Compute a WAR figure | no, and it never will | the two sites’ own leaderboards and data downloads |
| Convert bWAR into fWAR or back | no — there is no mapping | nowhere; they are different quantities |
| Compute the components a box score can support | no, but the pages below do | the ERA, WHIP, wOBA and fielding pages on this site |
| Turn a WAR figure into a price or a probability | no | out of scope across this entire site |
What WAR is trying to do, and why this page will not compute one
Wins above replacement tries to answer a question no single traditional statistic can: how many more games did a team win because this player was on it rather than whoever was freely available? Everything a player does with a bat, on the bases, in the field and on the mound is converted into runs, those runs are converted into wins, and the total is expressed relative to a replacement-level player. The appeal is obvious — one number, comparable across positions, so a shortstop and a starting pitcher can be put on the same scale.
Replacement level is a construction, not a measurement. It is deliberately set below the worst regular player: Baseball-Reference uses a “.294 winning percentage”, which it describes as a 48-114 record, calibrated so that “2013 MLB has 1000 WAR in the entire major leagues”, and FanGraphs uses the same convention. That is why a negative WAR is not simply bad play: it means a team would have done better with someone off the waiver wire. It is also why the figure is a valuation rather than an observation — change the replacement baseline and every player’s WAR moves.
There is no single WAR. Baseball-Reference’s own explainer opens by saying that “There is no one way to determine WAR”, and the two mainstream public implementations — Baseball-Reference’s, usually written bWAR or rWAR, and FanGraphs’, written fWAR — return different numbers for the same player in the same season. They are not two estimates of one published quantity; they are two different quantities that share a name and a unit. The two sites say so themselves, which is the strongest evidence available. FanGraphs’ library states that “FanGraphs uses Fielding Independent Pitching (FIP) as our baseline” while “Baseball-Reference uses a pitcher’s runs allowed”, and that “Pitchers, however, are valued very differently by the different systems”. For fielders: “For fielding since 2002, we use Ultimate Zone Rating, while B-Ref uses Defensive Runs Saved.” Baseball-Reference’s comparison chart puts the general case as “there are dozens of implementation details that have multiple reasonable approaches to take”.
How large is the disagreement in practice? Two to three wins on a single season, repeatedly. Samford University’s Center for Sports Analytics printed four 2022 player-seasons with both figures: Sandy Alcantara at 8.0 bWAR against 5.7 fWAR, Kevin Gausman at 3.0 against 5.7 — the same gap in the other direction — Brendan Rodgers at 4.3 against 1.7, and Manny Machado at 6.8 against 7.4. Twinkie Town, independently, printed Kyle Gibson’s 2019 as 0.0 bWAR against 2.4 fWAR. Alcantara’s 8.0 was checked against Baseball-Reference’s own player page and is correct there, alongside an ERA of 2.28 and a FIP reported elsewhere as 2.99 — which is exactly the mechanism: one system credited the runs he did not allow, the other did not. The FanGraphs halves of these pairs come from the secondary sources and were not verified against FanGraphs itself, and that is said here rather than glossed over.
So this page computes nothing, and the reason is not caution. There are two separate obstacles and either alone would be enough. First, neither site publishes a complete reproducible specification: Baseball-Reference calls its own WAR “necessarily an approximation”, and both systems’ defensive inputs are proprietary fitted models rather than arithmetic anyone can repeat. Second, even a perfect specification would need inputs that are not on any box score — league-wide seasonal run constants, a runs-to-wins converter for that season, and a park factor for the specific stadiums a player appeared in, weighted by appearances. A figure produced here would be neither bWAR nor fWAR, and it would be read as “WAR”. That is the same position this site takes on the expected goals (xG) difference calculator, where every provider fits its own model and none publishes the coefficients, and on the Duckworth–Lewis–Stern explainer, where the method’s own resource tables are not public.
Nothing here is a price or a probability of anything you can stake on. This page does not convert odds, imply a market probability or evaluate a wager, and no page on this site does. An expected win total from runs scored and runs allowed is a description of a season that has already been played, not a forecast and not a quotation.
What can legitimately be computed instead, and is. Every input a published box score actually contains has an exact, definitional answer somewhere on this site: the earned run average calculator for run prevention under Rule 9.21(e), the WHIP and pitching rates calculator for baserunners allowed, the wOBA calculator for weighted offence on FanGraphs’ own published annual run values, the batting average, OBP and SLG calculator for the three percentages Rule 9.21 defines, the fielding percentage and range factor calculator for the fielding arithmetic that is in the rulebook, and the Pythagorean expectation calculator for the team-level runs-to-wins question that WAR asks at the player level. None of those is WAR, and all of them are exact.
Frequently asked questions
Can I calculate WAR myself?
Not in a way that would match either published figure. Neither Baseball-Reference nor FanGraphs publishes a complete reproducible specification, both use proprietary fitted defensive models, and both need league-wide seasonal constants and a per-park factor weighted by where a player actually appeared. None of that is on a box score.
Why do bWAR and fWAR disagree?
Mainly two choices. For pitchers, Baseball-Reference starts from runs actually allowed while FanGraphs starts from Fielding Independent Pitching, so a pitcher who outperforms their strikeouts and walks scores better at the former. For fielders, the two use different proprietary models — Defensive Runs Saved against Ultimate Zone Rating.
How big is the gap between bWAR and fWAR?
Published examples run to two or three wins on a single season. Sandy Alcantara’s 2022 was 8.0 bWAR against 5.7 fWAR; Kevin Gausman’s the same year was 3.0 against 5.7, the gap reversed; Brendan Rodgers’s was 4.3 against 1.7. Manny Machado’s 6.8 against 7.4 is the more usual size for a hitter whose defensive metrics agree.
What does replacement level actually mean?
A deliberately chosen baseline, not a measured one. Baseball-Reference sets it at a .294 winning percentage — a 48-114 record — calibrated so that the whole of the 2013 major leagues sums to 1,000 WAR, and FanGraphs uses the same convention. A negative WAR means a team would have done better with a freely available player.
Why does this page not have a WAR calculator?
Because any number it produced would be neither bWAR nor fWAR and would be read as WAR. The inputs are not available and the methods are not fully published, so the result would be a third quantity wearing a familiar name. What the page does instead is tell you which implementation you are reading and point to the components that can be computed exactly.
Is WAR the same as expected goals in football?
Not in substance, but the honesty problem is identical: both are families of fitted models rather than single published formulas, so there is nothing for a calculator to implement and no way to compare one provider’s figure with another’s. This site takes the same position on both, and on cricket’s Duckworth–Lewis–Stern method for the same reason.
Related calculators
References
- Baseball-Reference. WAR Explained — “There is no one way to determine WAR”; WAR is “necessarily an approximation”; fielding uses “Baseball Info Solutions Defensive Runs Saved from 2003-present”, Total Zone before; replacement level is a “.294 winning percentage”, a 48-114 record, set so that “2013 MLB has 1000 WAR in the entire major leagues”; pitchers carry “a player-specific park factor for pitchers weighted by actual appearances in each park”. baseball-reference.com (read 12 Oct 2026).
- Baseball-Reference. WAR Comparison Chart — compares five systems on park factors, fielding metric, pitching starting point, replacement level and downloadability, stating “there are dozens of implementation details that have multiple reasonable approaches to take”; replacement level “.294 (48-114)” for FanGraphs and current B-R WAR. It contains NO player-season WAR figures (read 12 Oct 2026).
- FanGraphs Library. fWAR, rWAR, and WARP — “FanGraphs uses Fielding Independent Pitching (FIP) as our baseline” while “Baseball-Reference uses a pitcher’s runs allowed”; “Pitchers, however, are valued very differently by the different systems”; “For fielding since 2002, we use Ultimate Zone Rating, while B-Ref uses Defensive Runs Saved”. It gives no numeric size for the disagreement and names no players (read 12 Oct 2026).
- McLeod J. “Sabermetrics 101: Understanding the Calculation of WAR.” Samford University Center for Sports Analytics, 11 April 2023 — quotes both systems for the same 2022 seasons: Sandy Alcantara “only valued at 5.7 fWAR compared to 8.0 bWAR”; Gausman 5.7 against 3.0; Rodgers 1.7 against 4.3; Machado 7.4 against 6.8. “The major difference between fWAR and bWAR comes from the measurement of fielding runs.” samford.edu (read 12 Oct 2026). The bWAR side of the Alcantara figure checks out against Baseball-Reference; the fWAR side was not verified against FanGraphs directly.
- Baseball-Reference. Sandy Alcantara, standard pitching — the 2022 row reads W 14, L 9, ERA 2.28, IP “228.2” and WAR “8.0”. It corroborates the 8.0 bWAR quoted by the Samford article, and the printed “228.2” is a published instance of the notation this batch is built around: 228 innings and two outs. baseball-reference.com (read 12 Oct 2026).
- Twinkie Town (SB Nation). Analysing the Kyle Gibson WAR discrepancy — a second, independent set of player-seasons with both figures: Gibson’s 2019 is “a 2019 pitching WAR of 0.0” at Baseball-Reference and “2.4” at FanGraphs, with Berrios 3.0 against 4.0, Odorizzi 3.2 against 4.0 and Rogers 2.6 against 2.2. The high infield-fly rate is offered as a hypothesis, not a conclusion. twinkietown.com (read 12 Oct 2026).
- Fish Stripes (SB Nation). Sandy Alcantara 2022 stats — writes the same workload as “228 2/3” and as “228.2” on one page, which is direct published evidence that the decimal is a thirds notation. Gives his 2022 FIP as 2.99 against an ERA of 2.28, which is the mechanism behind the two WAR systems valuing that season differently. fishstripes.com (read 12 Oct 2026).