What WAR means in baseball

Wins above replacement tries to answer one deliberately broad question: if this player vanished and the team replaced him with a freely available minor-league call-up, how many wins would it lose?

That comparison point — replacement level — is the key idea, and it’s what distinguishes WAR from most other statistics. It isn’t comparing a player to the league average; it’s comparing him to the realistic worst case, the player any team could sign for nothing. A replacement-level team would win roughly 48 games in a 162-game season.

Everything a player does that raises the team above that baseline — hitting, baserunning, fielding, pitching, and the positional difficulty of where he plays — is converted into runs, and then runs are converted into wins.

WAR is an assembly, not a formula. For a position player it sums roughly five components:

• Batting — runs created above average, park- and league-adjusted.
• Baserunning — steals, taking extra bases, avoiding outs on the paths.
• Fielding — runs saved or cost relative to an average defender at that position.
• Positional adjustment — a credit for playing a demanding position. A shortstop and a first baseman with identical bats are not equally valuable, because shortstop is far harder to fill.
• Replacement adjustment — the constant that shifts the baseline from average down to replacement level.

Those add up to a run total, which is divided by the runs-per-win figure — around ten — to produce wins.

Pitcher WAR works the same way but starts from run prevention, with the versions differing sharply on whether to use actual runs allowed or a fielding-independent estimate.

The numbers are meaningless without the benchmarks, which are roughly stable across versions.

There is no single official WAR, and the two best-known versions — fWAR from FanGraphs and bWAR from Baseball-Reference — routinely differ on the same player, sometimes by two wins or more.

The main disagreement is pitching. FanGraphs builds pitcher WAR primarily from fielding-independent outcomes — strikeouts, walks, home runs — on the reasoning that a pitcher controls those and not what happens once a ball is in play. Baseball-Reference uses runs actually allowed, adjusted for the quality of the defense behind him. A pitcher with a great defense or unusual luck on balls in play can look very different under the two.

They also use different fielding metrics, and defensive measurement is the least settled area in the sport.

The practical advice: never compare an fWAR to a bWAR. Pick one version and stay inside it, and treat differences under about one win as noise in either.

WAR is genuinely good at what it was built for — a rough, one-number summary for comparing dissimilar players across positions, and for valuing contracts, since the market has a fairly stable price per win.

It is much weaker in a few places. Single-season defensive metrics are noisy, so a big WAR swing often reflects a fielding estimate rather than a real change in the player. Catchers are poorly served, since framing and game-calling are only partly captured. And small samples mislead — a month of WAR means almost nothing.

The honest way to use it: as a starting point that tells you roughly which conversation a player belongs in, not as a final verdict that settles it. A 0.4 gap between two players is not a difference.

WAR is a season-long accounting of value. TwentySeven does the game-scale version: open any game and you get player impact rankings built on win probability, showing who actually moved the needle that night and which five high-leverage plays decided it.

The two are complements. WAR tells you who has been valuable across a season; WPA and impact rankings tell you who won this game.