How FTI works

Full Press Team Indices describe NFL team performance from play-by-play data. FOI is the Full Press Offensive Index; FDI is the Full Press Defensive Index; FTI is the Full Press Team Index, combining offense and defense.

FTI is the short name of this destination. The page presents these three indices and a separate FSI special-teams beta together so you can see whether a team’s overall standing comes from offense, defense, or both.

Each index is standardized within its season and scope, centered at 50 with a 15-point standard-deviation multiplier, and capped at 0–100. Higher is better. A score of 75 is not a 75% win probability, a 75th percentile, or a prediction of the next game.

Component bars are separate 1–99 percentile profiles. Averaging those displayed bars will not reproduce the headline index. Small differences between teams should be read with the sample and schedule in mind.

The grade bands are: 80+ Exceptional; 70–79.99 Elite; 60–69.99 Very Good; 50–59.99 Above Average; 40–49.99 Below Average; 30–39.99 Poor; and below 30 Very Poor.

These describe relative performance in the selected pool. The red and gold framing on downloadable team cards is the card design; use the actual numbers and leaderboard grades to interpret performance.

No. FTI v1.0 combines offense and defense on qualifying scrimmage plays. Kicking, punting and return units are evaluated separately by FSI 0.3 beta and remain outside the overall FTI v1.0 formula. That is a meaningful limit when comparing FTI with wins, point differential or a complete assessment of a team.

FOI uses 60% opponent-adjusted offensive value per play, 20% Consistency+, 10% Impact+ and 10% Avoidance+.

The inputs are standardized. The three secondary traits are measured above the expectation associated with a team’s overall value level, then their combined weighted contribution is capped at ±0.50. The resulting composite is standardized again onto the 50-average, 15-point scale.

• Consistency+: the rate of plays with EPA greater than zero, above expectation for that team’s adjusted value level.
• Impact+: the rate of plays worth at least +1.00 EPA, above expectation.
• Avoidance+: avoidance of plays worth −1.00 EPA or worse, above expectation.

The “+” signals performance beyond the fitted expectation. It does not mean a raw percentage, and Consistency+ is about the distribution of play results rather than week-to-week stability.

FDI uses 60% opponent-adjusted defensive value prevented per play, 20% Stops+, 10% Impact Prevention+ and 10% Disruption+.

The secondary traits are standardized after accounting for the defense’s value-prevented level. Their combined weighted contribution is capped at ±0.50 before final standardization. A higher FDI means better defensive performance.

• Stops+: the rate of opponent plays with EPA at or below zero, above expectation.
• Impact Prevention+: prevention of opponent plays worth at least +1.00 EPA, above expectation.
• Disruption+: the rate of opponent plays worth −1.00 EPA or worse, above expectation.

Disruption can come from more than sacks or turnovers. The metric measures the damaging play result, rather than counting only one event type.

FTI gives 50% weight to standardized FOI and 50% to standardized FDI, then standardizes that combination again onto the same 0–100 display scale. It is not the arithmetic average of the two displayed scores, especially when either score reaches a cap.

Equal weighting makes offense and defense equal contributors to the model. It does not assign credit to individual players, position groups or coaches.

Consistency, impact and avoidance overlap with overall play value. Measuring those traits above expectation and capping their combined contribution helps keep the headline score anchored to value without rewarding the same production repeatedly.

The ±0.50 cap applies in standardized composite units before the final score is rescaled. It is separate from the ±0.12 EPA/play opponent cap.

The starting point is offensive EPA per play: the change in expected scoring value associated with a play. The model subtracts an estimated opponent defensive effect to put the result in schedule context.

A positive adjusted value means positive modeled value on average. Raw EPA/play remains visible so you can see what changed with the opponent adjustment.

Defensive value prevented reverses the offensive EPA perspective and adds an opponent-offense adjustment. Higher value prevented is better; a positive number means more value prevented on the model’s scale.

Raw EPA allowed is shown in the opponent’s offensive direction, where lower is better. Because these columns use different perspectives, check the label before comparing their signs.

The model estimates an opponent’s EPA level relative to the league using its plays against other teams. Games against the team currently being evaluated are excluded from that opponent estimate.

The effect is shrunk toward league average with a 400-play prior. If fewer than 40 other-opponent plays are available, the effect is zero. Applied opponent effects are capped at ±0.12 EPA/play. These safeguards reduce overreaction to small opponent samples; they do not remove every schedule difference.

Adjusted net value per play adds adjusted offensive value/play and defensive value prevented/play. It is a useful two-sided rate summary, separate from the standardized FTI composite.

Full Press Offensive Value (FOV) and Full Press Defensive Value (FDV) sum adjusted value above their respective league play baselines. They measure cumulative production, so workload matters. The displayed total value adds FOV and FDV.

Low leverage is defined using pre-play win probability at 2% or lower, or 98% or higher. Other plays are classified as competitive; if win probability is missing, a play is not flagged as low leverage.

Qualifying low-leverage plays remain in the headline indices. The model does not apply an extra leverage weight or delete them from the score. Low-leverage shares and situational context help explain the sample.

The model includes qualifying passes, rushes and quarterback dropbacks with valid EPA and identified offense and defense. It excludes kneels, spikes, explicit no-plays and two-point attempts.

Turnovers and sacks on qualifying plays remain part of performance. The indices use team play results, so they do not attempt to divide every result among the quarterback, protection, receivers, runners and coaching.

The historical collection begins in 2006. It includes regular-season, postseason and combined team seasons, with the current season tracked separately. The page shows the loaded season and, when available, its completed-week cutoff.

Historical team identities follow the source record for that season, including older location abbreviations. Coverage is a data window, not a claim that every era has identical rules, schedules or context.

Regular Season includes regular-season plays; Postseason includes playoff plays; Combined recalculates from both sets of plays. Combined scores are not an average of the two displayed indices.

Each scope has its own comparison pool and opponent context. A playoff-only grade is based on a smaller pool and much less playing time, so compare it carefully with a full regular season.

Team leaderboards include teams with qualifying offense and defense in the selected scope. They do not use the quarterback dropback thresholds from FQI. Inclusion alone does not make a sample stable.

Early-season and postseason scores can move sharply. The 40-play opponent rule controls one adjustment estimate; it is not a general qualification threshold or a guarantee of reliability. Read the week cutoff, games and underlying play counts alongside the grade.

Top Seasons ranks completed team seasons in the selected metric and scope. The current incomplete season is kept out of that collection. FTI, FOI and FDI each provide a different ranking lens.

These are season-relative scores. They show how strongly a team performed in its own comparison pool, rather than a literal simulation of teams from different eras playing each other.

Game FOI, FDI and FTI use the same component structure, but are standardized against the full set of team-game performances in the selected season and scope. Opponent adjustment is derived from that season context.

A single-game score is not standardized only against the selected team’s games. Season grades are calculated from pooled plays rather than by averaging game grades. Game scores may change as the season’s comparison pool and opponent estimates develop.

Select a team name or an index score to open its profile. The profile contains offense and defense components, situational context, season-card downloads and the individual-game log.

Use Download Team Card for a season PNG, or PNG beside a game for its card. Share uses supported device sharing; otherwise it downloads the image. Cards include the selected season or game, so check the context before sharing.

FQI evaluates quarterback performance using quarterback-attributed value, accuracy, consistency and avoidance. FTI, FOI and FDI evaluate whole-team offense and defense from team play outcomes.

FOI includes the contributions of the full offense. It should not be treated as a second quarterback grade. The two destinations share a visual design and navigation links, while using distinct formulas and comparison pools.

Yards and points summarize production and scoring. Raw EPA adds down, distance and field-position context through expected scoring value. These indices build on EPA by adding opponent context and complementary play-result traits, then putting the results on a season-relative scale.

Wins, scoring margin, injuries, special teams and film remain useful context. FTI does not claim to summarize everything that decides a football game.

Source corrections can change eligible plays, EPA, team attribution or game details. New games also change opponent estimates, fitted expectations and the season’s comparison pool. Updated inputs can therefore change an earlier score without a formula change.

The page’s season, scope, cutoff and refresh information describe the loaded snapshot. A previously downloaded image preserves the values from the time it was created.

The model relies on public play-by-play and its EPA estimates. It cannot fully separate individual responsibility, personnel availability, protection, coverage, coaching, weather or tactical choices. Residual traits and opponent adjustment are statistical estimates, not film charting.

Special teams are excluded from FTI v1.0 and shown separately in FSI beta. Low-leverage plays remain included. Small samples, unusual schedules and different postseason pools can make comparisons less stable. Use the indices as one part of an evaluation.

FTI, FOI and FDI are proprietary Full Press Coverage metrics. This page documents Team Indices v1.0. The plugin’s software version can change for a design or feature update without changing the metric version.

Methodology changes should be versioned and evaluated across historical data before adoption. Full Press Coverage may correct data and improve the model as evidence develops. The scores describe modeled performance; they are not guarantees, forecasts or official NFL grades.

FSI is the Full Press Special Teams Index. It summarizes observed kicking, punting, kickoff coverage, punt returns and kickoff returns. The beta release is version 0.3. FSI appears alongside FOI, FDI and FTI; it has no weight in the published FTI v1.0 formula.

Field goals and extra points compare actual scoring with an expected make probability based on kick distance and roof type. Punt and ordinary kickoff results compare post-play expected scoring value with a baseline using pre-kick field position, time remaining in the half and roof type. Realized kick distance and return yards are outcomes, not inputs to those punt and kickoff baselines.

Regular-season baselines exclude the graded game through five folds of games. Postseason plays use regular-season baselines. Blocked kicks and documented defensive kick-return scores receive explicit credit with a training-sample baseline debit. An ordinary missed field goal does not give the opponent a positive windfall.

The season profile shows contributions in quality points per game. Ordinary unit values are shrunk toward average and converted to a league-average workload. Priors are 20 field-goal attempts, 30 extra points and 40 opportunities for each punt/kickoff side. Block and rare-event values use an eight-game prior. Kicking combines field goals, extra points, blocks and kicking penalties.

The five contributions sum to quality per game. They are not fixed percentage weights or percentile grades. Total recorded value is the unshrunk sum of credited play value; it differs from quality because the latter adjusts for volume and sample size.

Completed regular and combined seasons use 50 plus 15 times the standardized quality score, capped at 0–100. The current season and playoff pools instead use neutral zero quality and the spread of the previous three regular seasons; the first historical postseason uses that year’s regular season as a fallback. This avoids forcing a small early-season or playoff pool to average 50.

The familiar grade colors use the same numerical bands as the other indices. A score is not a win probability or a percentile. Profiles with fewer than four games are marked provisional. Even a full season contains relatively few special-teams opportunities.

Explicit onside attempts have a separate baseline using pre-kick field position, clock, score differential and broad rule-era indicators. Training uses other regular-season games and up to three prior seasons. Onside results are credited equally and oppositely to the receiving and kicking teams, outside ordinary kickoff opportunity counts. A short or pooch kick is not automatically classified as onside.

Accepted penalty-only records in a recognizable special-teams formation receive their source EPA with opposite team credits. The voided kick receives no separate credit. Offset or declined penalties without accepted enforcement receive no added penalty value. Live punt and kickoff penalties are already reflected in their resulting field position.

Plays already owned by the published FOI/FDI eligibility rules, including fakes, scrimmage attempts after aborted snaps and a small number of legacy parser anomalies, remain there. FSI excludes them to prevent duplicate credit while preserving published history. Deleted, ambiguous or incomplete records can remain ungraded.

FSI covers available play-by-play from 2006 onward, with regular-season, postseason and combined views. Select FSI in the Index menu for its leaderboard or Top Seasons. Select a team for five unit contributions and a downloadable season card. Its game log includes Game FSI plus separate FSI PNG and Share controls.

Game FSI standardizes total game value against team-game results in the selected season and scope. Game cards show raw unit value in points, while season cards show shrunk quality contributions per game. Averaging game grades does not reproduce season FSI. Rule changes, especially kickoffs and extra points, limit direct comparisons across eras.

FSI 0.3 is a descriptive beta and has no opponent adjustment. Its context baselines account for the kick situation; they do not isolate every coverage unit, returner, kicker or coach. Wind, precise hang time, tracking, blocking responsibility and tactical intent remain incomplete.

A chronological test selected a stronger forecasting prior on 2006–2017 and evaluated it on 2018–2025. Its reduction in prediction error versus a neutral forecast was 1.27%, with a 95% season-bootstrap interval from −3.18% to +5.45%. The interval includes no improvement. The displayed descriptive priors were kept; FSI should not be presented as a validated forecast. Historical source revisions and EPA models also prevent an as-of-data claim.

Yes. Source corrections, new games, fitted kick expectations, comparison pools and beta methodology updates can change scores. The FSI version is separate from the plugin version and the published FTI/FOI/FDI v1.0 formulas. A downloaded card preserves its original values.

If FSI and the team leaderboard have different current cutoffs, FSI is shown as unavailable until a matching update arrives. Missing data is never treated as zero. Any future proposal to include FSI in overall FTI requires separate historical evaluation and a versioned formula change.