How our values work

Every player and rookie pick sits on one scale so a quarterback, a receiver and a pick can be added up in the same trade. The best asset in your league settings is 9,999. In each win-window column only that window’s best asset reaches 9,999, and no player a window favours is shown below his overall value.

A value is a player’s points a game above what a team could get from its bench and the waiver wire (roster depth, about 1.3 starters deep at each position), added up over the seasons he is likely to play and priced for your scoring and lineup.

We start from what a player scores in a typical healthy game now: his recent usage in games he actually played (recent games count most), times his career efficiency. Then we look at how much of that level players of the same position, age, tier and draft capital kept in each later season since 1999, and whether they kept their role (age, production trend, availability, rookie-deal years, team record and quarterback play all count). A young player’s chance of a starting job starts from his draft slot and moves with what he has shown, where he sits on his club’s depth chart and how many young players are chasing the same job; an undrafted player is judged the same way, on his own production. A young player’s production counts as fully proven after two seasons’ worth of starter output, not one breakout season.

An elite producer with two or more seasons in the last three has shown a floor: while he keeps his role and plays at or above it, his worst of those seasons is the least we project him for, until he turns 30.

Time missed counts. How much of his team’s schedule a player has been available for over the last three seasons (games missed through injury, not games he sat as a backup) scales what we project him to score in later seasons, and a pattern of long absences, four games or more in a season, costs more than one bad break. His floor is scored the same way, in the games he is likely to play.

Values at the very top are sticky: a top-24 player’s value falls at most 3% a week unless he goes on injured reserve, is ruled out for the season, is suspended, becomes a free agent or reaches the age his position falls away, and a young player breaking out is shown among the top 24 only after holding that level for four weeks.

Each season is priced over a bad, an expected and a big year, so a high ceiling counts and a bust only costs down to the bench. His injury record, depth chart, offence, market momentum, the crowd’s calls and the news adjust it, and the projections are checked every season against what actually happened.

All In pays for this season, with the fantasy playoffs counted twice, and less for every season after it. Playoff Gamble pays for this season and next. Soft Rebuild and Hard Rebuild pay for two to five years out, with Hard Rebuild paying a premium for young players who already produce. Contenders pay a little more for spike weeks; rebuilds pay for long-run upside. The very best players move least between windows.

Floor and ceiling are the 20th and 80th percentile outcomes of comparable careers. Weekly range is a player’s 20th to 80th percentile week over the last three seasons. Spike weeks are the share of his weeks among the best at his position (top 6 QBs and RBs, top 8 WRs, top 4 TEs). Volatility is how far comparable players ended from this value.

Redraft value counts this season only: a player’s projected points a game above a replacement-level starter at his position, times the fantasy weeks left, for your scoring and lineup. A player on injured reserve counts for none of the regular season that’s left. If he’s elite and could be back for the fantasy playoffs, those weeks count from the week he’s expected back, at his chance of making it, with nothing at all for a player out for the season. Dynasty values price his current season the same way, in every win window. Weekly projections come from usage, efficiency, the opponent, the betting market’s expected score, the weather forecast at kickoff and the injury report, and a player’s line freezes at his own kickoff. College projections are half-PPR points a game with recent games weighted up, scaled by his team’s expected score from the betting market, or team ratings where there is no line.

A rookie pick is worth the average career of the players taken in that slot in rookie drafts since 2000, discounted for the risk of a miss. College prospects are priced by their projected NFL draft slot and production; the draft slot leans on the consensus of public mock drafts and big boards, and once that consensus is three weeks old it moves only with new evidence from the stats.

The rest of the NFL season is played out 20,000 times from each club’s results so far. The playoff chance is the share of those seasons a club makes it; a pick is shown where it lands on average, with the range it falls in four times in five. Compensatory picks are left out until they are awarded in March.

Need is read from the current roster: how each unit is playing this season ranked among the 32 clubs, starters past the age their position usually falls away, players who are out, our view of the next three seasons at the fantasy positions, and first- or second-round picks spent there in the last two drafts. Contracts are not counted.

An article can only move what the numbers can’t see yet: a role, depth chart, coaching, scheme, trade, contract or off-field change. A player’s own injury, box scores, matchups and weather are already counted, so they never move. One story moves a value by 5% at most and a weekly projection by 10%; all of a player’s stories together by 12% and 20%. The nearer a player is to the top of the scale, the less he moves.

News about this season counts most for contenders, news about later seasons most for rebuilds. A move holds for a week, then fades as the games take over: gone after 28 days for this-season news, 42 for both, 56 for later seasons, and 21 days for a projection.

Signed-in votes nudge values, never projections. Ranking arrows and the daily question (a start/sit call, who you would rather own, or which player belongs higher) count once at least 20 people have weighed in over the last 30 days, move a value by at most 10% in all, and by no more than one percentage point a day. A start/sit call counts toward redraft value. Choosing a player over one valued higher counts for more than choosing the favourite. Over/under calls move a value up to 3% a week and 15% in all; a split crowd counts for less than a united one.

A player’s value over time is rebuilt from the data available on each date, every two weeks for the last two years, in a 12-team superflex PPR league with a TE catch bonus. It is not a record of trade-market prices.

We rebuild ten past boards, each October from 2015 to 2024, from only the data that existed that day, and compare them with what happened. The score is a rank correlation: 1 would be a perfect order, 0 no better than chance. No model gets close to 1, because injuries, trades and coaching changes can’t be seen in advance. So the fair test is against the simple approaches anyone could use.

Dynasty values beat both simple orders in all 7 seasons we can grade four years on (2022 onwards still have seasons to play). All In beat this season’s points a game in 8 of 10 seasons. Each table takes the top 200 by each method on the week 5 board; the simple orders can’t price rookies or picks, so they are graded on veterans only.

Dynasty is graded on points above a replacement starter over the next four seasons, nearer seasons weighted more; All In on points above replacement over the rest of that season, fantasy playoff weeks counted twice.