FWI How it works

FWI — Full Wide Receiver Index — summarizes regular-season receiving performance across target value, catch efficiency, yards after catch, and explosiveness. It evaluates NFL players listed as wide receivers in the player-stat feed; restored 2006–2008 target-only games require a verified same-season, same-team position. Tight ends and running backs are excluded.

FWI uses 50% opponent-adjusted target EPA, 25% catch rate over expectation, 15% YAC over expectation, and 10% explosive target rate. Each component is standardized within the eligible WR cohort for that season. The blend is standardized again, scaled to 50 + 15 × z, and bounded between 0 and 100. Version 0.1 is a descriptive beta; the weights were chosen to represent these dimensions, not fitted to a proven forecasting advantage.

Before the final 0–100 bounds, the qualified season cohort has a mean of 50 and standard deviation of 15. A higher FWI indicates stronger performance under this recipe. The score is not a percentile or a percentage chance of success. Cross-season comparisons reflect performance relative to each season’s qualified receivers.

Expected points added is the change in the offense’s expected points on a play. Target EPA averages that value over targeted passes with available EPA. It includes quarterback, blocking, scheme, opponent and interception effects. It does not assign every outcome solely to the receiver.

Within the same season and regular/postseason phase, the model compares a defense’s EPA allowed on other WR targets with the league reference. The entire game being evaluated is excluded from both reference samples. The defensive difference is shrunk by n / (n + 200) and capped at ±0.08 EPA per target. Tougher observed defenses produce a positive adjustment. With no usable other-game sample, the adjustment is zero. Historical full-season adjustments are retrospective, not pregame projections.

Catch OE is actual catches minus the sum of expected catch probabilities, divided by targets with valid model probabilities, multiplied by 100. It is expressed in percentage points. For example, +5.0 means five percentage points above the source model’s expected completion rate on the covered targets. This is not a charted drop rate or a measure of receiver separation.

YAC OE compares yards after catch with the source model’s expected YAC on completed receptions that have both values. It is expressed in yards per matched reception. A +1.0 means one yard above expectation on average. Quarterback placement, blocking and defensive context can still affect it.

The v0.1 component is the number of primary receptions gaining at least 20 yards divided by targeted passes. Lateral-only receiving yardage is retained in production totals but is not an additional targeted reception.

Completed regular seasons require 50 targets. The current threshold is ceil(50 × completed week / 18), or 12 targets through Week 4. Seasons through 2020 use 17 weeks. A receiver also needs verified target counts and matching catch and receiving-yard totals, complete EPA coverage, at least 90% catch-model coverage, at least 80% expected-YAC coverage, and at least five matched receptions for YAC. At least five eligible WRs are needed to form a scoring cohort. Unqualified rows remain available when Qualified only is unchecked.

A dash means the score is unavailable because the sample or model coverage did not qualify, a source disagreement remains unresolved, or the row is postseason data. The details show the reason. Missing metrics are not replaced with zero.

Regular-season and playoff statistics are kept separate. FWI v0.1 assigns scores only to regular-season season records. Individual game rows and playoff records retain production and contextual receiving metrics without invented FWI grades.

Lateral plays can credit receiving yards to a player without a target or reception. The dataset preserves those yards and keeps a zero-target game record. Games recorded counts rows in the player-stat feed, not all appearances; games with no recorded stats require a separate participation source to establish that a player appeared.

The numerator is WR targets. The denominator is all-position team targets in the games with a recorded WR stats row. It is not route participation and may differ from a full-team-season target share for players without records in every game. aDOT is air yards per target with air yards available. Yards per target uses canonical receiving yards, including lateral credits.

Low leverage is a pre-play win probability at or below 2% or at or above 98%. Competitive and low-leverage target EPA and the low-leverage target share are displayed separately in receiver details. They do not change the scoring weights. Plays missing win probability are excluded from the split denominator.

Receiving yards and fantasy points are production totals that depend heavily on opportunity. Target EPA describes the value of targeted pass plays. FWI adds catch and YAC expectation models and explosiveness, then compares the resulting blend within the season’s qualified WR cohort. Targets and raw production remain beside FWI so role and sample are visible.

Not fully. The inputs include shared offensive context, and some components overlap. The public sources used here do not provide verified routes, yards per route run, separation, contested-catch grades or charted drops. FWI should be read alongside film, role and other statistics.

Nineteen complete adjacent-season comparisons were examined from 2006–2007 through 2024–2025. The balanced formula did not consistently exceed the value-only baseline in next-season adjusted target EPA rank correlation. Checks use returning qualified WRs and have survivor-selection limits. Current-season 2026 was excluded. These results support a transparent descriptive beta, not claims of proven predictive accuracy.

The archive includes 2006–2025 regular seasons and playoffs, plus 2026 through the completed-week cutoff shown above the table. Current refreshes use nflverse inputs and match the published FQI completed-week cutoff. A newer source file does not automatically mean a later week is complete. The existing snapshot remains available if a refresh fails validation.

The older weekly feed has incomplete target codes and can omit games with targets but no catches. FWI uses nflverse’s parsed receiver IDs to restore identified incompletions and interceptions, excluding nullified plays, spikes, sacks and two-point attempts. Targets and air-yard totals come from those plays. Missing target-only game rows require a verified same-season, same-team player identity and position. Weekly catches, yards and other production totals are preserved and checked independently. Unknown receiver identities are never guessed; sample, model-coverage and reconciliation requirements still apply.

Source providers may correct player credits, plays or model fields. Those corrections can change both the affected player’s inputs and the cohort used for standardization. Published snapshots keep source URLs, update times and SHA-256 identities. Formula changes require an explicit model-version change.

Stable player IDs join games across teams within a season and phase. Season totals combine those game records and display the teams represented. Individual games retain the team and opponent for that game.

Use the + button beside a receiver in Current or History, then open Compare. Different seasons of the same receiver can be selected. Download CSV exports the currently filtered and sorted table. These exports include the model version and distinguish regular season from postseason.

FWI is a proprietary Full Press Coverage metric built from public nflverse inputs. It is not an NFL, ESPN or PFF rating and is not endorsed by those organizations. The public methodology describes the current recipe and limitations. The metric will evolve through versioned changes and documented validation.