How Fantasy Wins Added works

Baseball's WAR asks how many games a team won because it had a player instead of someone freely available. Fantasy football is head-to-head, one opponent a week, so FWA asks the same thing about matchups: each week a player plays, how much more likely was his team to win with him in the lineup than with the best player it could have picked up off waivers? Add up the weeks and you get matchups won. A season of 3.0 FWA is three extra wins; the best seasons since 1999 are worth about four.

Points don't win matchups one for one. In a 12-team PPR league a week's scores spread widely, so the first points above average turn a coin flip into a likely win, and the next ones mostly pad a win you already had.

Value-based drafting and points over replacement count both twenties the same. That's the difference FWA makes: a steady player who wins close weeks gets more credit than a boom-or-bust one with the same points, because the boom weeks are often blowouts and the bust weeks losses.

The baseline is the player you could actually get. Every format's lineup is filled for 12 teams from that season's best players by points per game, FLEX spots going to the best running backs, receivers and tight ends left over (superflex to the best player of any position left). Benches come on top, as deep as real ones: we counted end-of-season rosters in 12-team Sleeper redraft leagues. Past the rostered players is the waiver wire, and replacement level is the points per game of the 5 best players there. It moves every season: running backs filled more FLEX spots in 2003 than they do now.

1. The week's average lineup. For each week, the season's starters at each position who played that week give the average score at that position and how widely it varies. Added up over the lineup, that's an average team's score and spread. Real lineups spread a little wider than this (managers start the wrong player, scoring rules differ), so the spread is widened by 20%: at that width a team's starters' FWA predicts its record in real Sleeper leagues about one for one.
2. His chance to win. With the rest of his lineup average and an average opponent, a player's score that week gives his team a chance to win. The bigger his score, the higher the chance, flattening out toward 100%.
3. Minus the replacement's. The same chance with the replacement's score instead. The difference is his wins added that week: usually between −.3 and +.45.
4. Add up the weeks. Weeks 1–17, the fantasy regular season and playoffs. A week he doesn't play, the replacement plays: zero. So an injury or a suspension costs FWA, just as it cost the teams that had him. A bye week does too.

Every number is for a 12-team league. The site leads with 1QB full PPR; the leader page switches between five.

The 1QB lineup is Sleeper's default and the most common one among the 12-team 1QB leagues we've seen; the superflex lineup is the most common superflex one. Kickers and defenses aren't included.

• The best seasons are the ones you'd expect. LaDainian Tomlinson's 2006 is first, then Marshall Faulk's 2000 and 2001, Priest Holmes's 2002 and 2003 and Christian McCaffrey's 2019. Cooper Kupp's 2021 is the best by a receiver.
• Real matchups follow the curve. In 475 real 12-team Sleeper leagues from 2024 and 2025 we took every starter's week (659,000 of them) and looked at how often his team won. When a starter scored 25 or more above the average starter, his team won 79% of the time. FWA's curve says 81%; counting points in a straight line says 85%. At 35 or more: 86%, 88% and 96%.
• Teams that start high-FWA players win more, by about what FWA says. A team's starters' FWA, against its opponents', predicted its record in those leagues at 0.96 wins per FWA, close to one for one.
• It describes, it doesn't forecast. A player's FWA one year and the next correlate at 0.72, the same as points over replacement and a little less than points per game (0.78). Use it to say what a season was worth, not what the next one will be.
• Over a whole season it mostly agrees with points over replacement (a correlation of 0.998). The differences are in the extremes: the biggest boom-or-bust seasons give back up to four tenths of a win (Marshall Faulk's 2000), the steadiest gain one or two tenths.

• Points are nflverse's: four-point passing touchdowns, minus two for an interception or a lost fumble, no bonuses. Your league's rules may differ.
• It's context-neutral: every team is average apart from the player. It doesn't know who your other starters were, or who you actually played.
• A game counts as played when he recorded a stat. A starter who played without a touch shows as a missed week (zero), not a bad one.
• Seasons start in 1999, so a career that began earlier (Marshall Faulk, Jerry Rice) counts from 1999.

WAR comes from baseball. For the NFL itself, see nflWAR (Yurko, Ventura and Horowitz, 2018) and PFF WAR. Fantasy versions have been written up before, by Jeff Henderson (Stathole), The Fantasy Footballers and Fantasy Points (for subscribers). What's new here: every QB, RB, WR and TE season since 1999 in one open table, five formats, a replacement level set by real rosters, and a win curve checked against real matchups.