Methodology

QB Synergy measures a quarterback–receiver connection relative to eligible peers. It is not an individual player grade or an absolute measure of good play. Scores use method 2026.4 in every season and time window.

Score components

Outcome weights are 23, 16, 6, 8, and 7 points, respectively. Trust weights are 14, 7, and 4. Duo lift weights are 12 and 3. These weights are unchanged. Interception and failed third/fourth-down penalties have maximum weights of 3 and 2 points.

Zero events and tied values

Zero successful targets, receiving first downs, explosive plays, third/fourth-down targets, or red-zone targets earn zero points for that component. Stabilization cannot create credit for an event that did not occur. An all-zero usage or event-rate pool earns zero component points.

Remaining comparisons use midpoint percentiles: 100 × (number below + half the number tied) / eligible observations. A group of identical values is at the 50th percentile, including a single-observation group. This avoids awarding the 100th percentile to every tied connection. Small comparison groups are flagged; scores can change substantially as more peers appear.

EPA, CPOE, QB EPA lift, and YAC over expectation can be negative. Zero in these metrics is a valid observation and can outperform negative values. The zero-event rule does not apply to these signed metrics.

Positive-component percentiles are rounded to whole percentiles, then multiplied by their component weights. Penalties use midpoint percentiles of observed failure rates and are rounded to hundredths of a point. Zero interceptions or zero failed third/fourth-down targets incur no penalty. No third/fourth-down targets means no failure penalty and no observation in that failure-rate pool. The calculation adds integer hundredths, subtracts penalties, and rounds the final result half up to a whole score between 0 and 100.

Peers, qualification, and observations

WR and TE connections share one comparison pool. RB and FB connections share the other. Fullbacks retain their listed FB label and appear under the RB / FB filter. Position filters change the displayed rankings, while scores retain these two comparison groups.

Scores begin at one target in all seasons, scopes, and time windows, including 2025. Regular-season and combined scores with 1–29 targets are provisional; 30+ targets qualify for qualified rankings. Playoff scores with 1–4 targets are provisional; 5+ qualify. Rankings default to provisional results until qualified connections with available scores exist. A zero-catch connection can still receive a score. All connections also includes rows whose score is unavailable.

Efficiency inputs are stabilized toward the available eligible-peer mean using observations / (observations + 40) for season-to-date and last-four-week scores. Weekly scores use observations / (observations + 8). EPA, first-down, and explosive rates use targets as observations. CPOE and success use targets with that metric available. YAC over expectation uses modeled receptions. QB EPA lift uses the smaller of the duo's targets and the quarterback's targets to other eligible receivers on the same team. Missing observations are excluded from the corresponding baseline and percentile pool.

Each component's breakdown identifies the observed input, adjusted value where applicable, observation count, eligible peer count, and points. Advanced-profile bars show the fraction of component points earned, alongside the raw metric. Additional metrics such as target depth and third/fourth-down target rate are shown separately.

Opportunity shares and team context

Target share is the fraction of the quarterback's eligible receiver-tagged targets directed to a receiver on the same team. Third/fourth-down target share uses the quarterback's eligible third/fourth-down targets as its denominator. Quarterbacks who play for multiple teams receive separate denominators for each team in the selected window.

Team red-zone target share is this QB–receiver connection's targets from the opponent's 20-yard line or closer divided by the team's total red-zone targets in the same scope and time window. The team denominator includes every quarterback and every targeted receiver position, including eligible linemen and gadget-play receivers. The numerator remains specific to the selected connection. Four targets out of the team's twenty red-zone targets is a 20% share. Rushes, untargeted throws, two-point attempts, and nullified plays do not enter this target denominator. The share is compared within the existing WR/TE or RB/FB peer group for up to four score points; it is not multiplied directly by four.

Third/fourth-down target rate is a different statistic: the fraction of this connection's own targets occurring on third or fourth down. The score uses target share, not this rate. For example, two third-down targets among six total duo targets is a 33.3% rate; if the quarterback had twelve third/fourth-down targets, the scored share is 16.7%.

When the team has no red-zone targets, its red-zone share is undefined and the interface displays “No team red-zone targets.” When the quarterback has no third/fourth-down targets, that share displays “No QB opportunities.” Both undefined shares contribute zero points and are excluded from their comparison pools. A connection getting zero of eight available targets is a measured zero and also earns zero usage points.

QB EPA lift is the duo's EPA per target minus the quarterback's EPA per target to other eligible receivers on the same team and in the same time window. With no other receiver targets, the comparison is unavailable. Very small reference samples are flagged.

Missing metrics and model coverage

Missing metrics remain unavailable. A score is suppressed when required EPA, success, CPOE, or QB EPA-lift inputs or comparison data are missing. Zero targets never produce a score.

YAC is the explicit exception: no receptions or no modeled YAC coverage receive zero of the seven YAC points while the other components are calculated normally. Missing coverage is labeled; zero credit is a conservative choice, not evidence of poor YAC performance. No-catch and unmodeled receptions are excluded from YAC baselines and percentile samples.

YAC over expectation averages completed receptions with both observed and expected YAC. Partial coverage is reported, and only modeled receptions determine its stabilization strength. CPOE and success coverage are also reported when incomplete. If observed YAC itself is missing on a catch, raw YAC per reception remains unavailable instead of substituting zero.

Weekly history, recent form, and rank changes

Each weekly point compares only with eligible connections in that NFL week and position group. Select a point by click, Enter, or Space to see that week's calculation. Later games do not enter an earlier week's population. Missing targets or required inputs produce gaps.

Last 4 weeks uses the four NFL calendar weeks ending at the latest included week. Byes remain gaps and do not extend the window. Raw totals and the comparison population are rebuilt for that window.

Rank changes compare cumulative results through the last two fully completed weeks, or the corresponding four-week windows, with the selected position and target minimum. New qualifiers show “New.” Score ties use target count, then player/team IDs for deterministic ordering. Historical snapshots are rebuilt after source corrections, so these are corrected week-to-week comparisons. Partial-week results are excluded from changes.

Data and daily refresh

Sources are nflverse play-by-play, player metadata, and game schedules. Only valid receiver-tagged ordinary pass attempts to WR, TE, RB, or FB players enter receiving totals. Two-point conversion attempts and nullified plays are excluded. Conversion counts are recorded separately in dataset metadata. The nflfastR statistics implementation also treats conversion statistics separately from ordinary targets and receptions. Travis Hunter's receiving plays are treated as WR plays.

Checks are scheduled daily at 4:00 a.m. Central, automatically adjusting for daylight saving time, and incorporate games and upstream corrections when the source publishes them. GitHub may delay scheduled starts. The active season remains explicitly set to 2026 through postseason games played in 2027. Corrected 2025 datasets remain available through the season selector. Coverage, source-check time, and build time are shown separately.

Before publication, checks validate receiving counts, opportunity denominators, model coverage, duplicate plays, every component and penalty in both position groups, and reconciliation of season/recent/snapshot totals with weekly rows. Invalid downloads, malformed inputs, or lost games stop publication; the last successful deployment remains available. A completed week requires final scores and play-by-play coverage for every scheduled game in the scope.

Definitions

EPA is expected points added. A successful target has positive EPA. CPOE is completion percentage over expectation, in percentage points. YAC is yards after catch; YAC over expectation compares observed YAC with the model's expectation. An explosive receiving play gains at least 20 receiving yards.