How Football Quant
builds a forecast.

01 · The foundation

Forecast football statistics first.

V1 forecasts 15 underlying player statistics: passing, rushing, receiving, sacks, and ball-security outcomes. It does not begin by guessing one fantasy-point total. A common stat forecast can support different league rules without changing the football prediction.

02 · Translation layer

Then apply fantasy scoring.

The rankings engine supports versioned Standard, Half-PPR, and Full-PPR presets. Those settings translate the same forecast stats into projected fantasy points. The Rankings page applies the selected preset instantly while leaving the football-stat forecast unchanged.

03 · Fantasy Quotient

fQ measures value above replacement.

fQ measures projected weekly fantasy points above replacement level in the selected league format. Required starters and FLEX players are allocated first. Each position's best remaining forecastable player establishes its replacement baseline.

fQ = max(projected points − position replacement points, 0)

League size, roster composition, and scoring format change demand, so they can change replacement level and fQ. The same player can be more valuable in one league than another without changing their underlying football forecast. League context matters.

04 · Starting preset

The default is only the starting point.

Teams

12

Scoring

Half PPR

Quarterback

1

Running back

2

Wide receiver

2

Tight end

1

FLEX

1

This preset powers the static initial snapshot. Rankings can then recalculate weekly points, replacement levels, fQ, and ranks for 8, 10, 12, 14, or 16 teams; Standard, Half PPR, or Full PPR scoring; 1QB, 2QB, or Superflex; 1, 2, or 3 RB starters; 1, 2, or 3 WR starters; 1 or 2 TE starters; and 0, 1, or 2 FLEX slots. FLEX is eligible for RB, WR, or TE. Superflex remains one flexible QB/RB/WR/TE slot; it is not the same as a second required QB slot.

05 · Time horizon

Current fQ is weekly.

It measures value for the upcoming forecast week. Rest-of-season fQ is future work and needs dedicated remaining-season forecasts, schedules, roles, and availability assumptions. A rest-of-season projection requires a true rest-of-season forecast. One weekly projection cannot simply be multiplied by games remaining.

06 · Honest nulls

Unavailable is not zero.

Players without qualifying NFL history remain in the weekly population but may not receive numerical forecasts. The rankings page labels those rows “No NFL history” and leaves projections and ranks unavailable rather than inventing zero production.

07 · Production and research

V1 is live. V2 is in the lab.

FQ Player Forecast Engine V1 is the active production methodology. V2 is an ongoing research program exploring additional predictive methods. Research models are not automatically promoted; they must earn their way past the established baseline and human review.

08 · Model performance

We track forecast error over time.

Football Quant measures forecast error after games are played, both overall and by position. Public performance tracking is planned by week, season, and model/version so changes in accuracy can be seen over time. The planned view will report aggregate error for the overall model and for QB, RB, WR, and TE. Performance should be attributable to the model/version that produced the forecasts.

Mean absolute error (MAE) summarizes the average size of a forecast miss; lower is better. Football Quant uses point-in-time historical testing to compare forecasting methods and measure error. New methods only replace the current approach when the results justify it.