THE EVIDENCE STANDARD
Know what the numbers mean.
Augris brings recorded NFL results and source context into a research workspace. Every conclusion starts with a defined sample. Your thesis remains yours.
Sources & identity
Player history uses nflverse data in a validated, retained generation. Directory and schedule responses carry their own provider attribution, including ESPN source listings. Open the evidence attached to a result to inspect its provider, season, phase, generation, and observation time.
Player research follows a canonical player identity and verified provider identifiers. Ambiguous mappings are withheld. A matching name alone does not establish identity, and a roster listing does not establish current availability or active status.
Player identity uses initials and team abbreviations. Augris does not request player photographs or third-party team logos.
Samples & comparisons
History is selected by season and regular season or postseason. An unavailable season stays unavailable. Prior-season results are never substituted into a current-season selection.
The historical population contains source-recorded offensive player weeks with at least one pass attempt, carry, or target. Missing, inactive, and zero-opportunity weeks are not filled with zero. A metric's sample counts only recorded, finite values for that metric.
- Totals and averages
- Additive totals sum the selected recorded values. Averages divide their sum by the displayed sample. Weekly target share and completion percentage over expectation use unweighted means of weekly values; their values are not summed.
- Two-player comparisons
- Both players use the same season, phase, metric definition, and eligible population. Each player keeps their own sample size; the comparison does not imply matched opponents or identical playing time.
- Descriptive findings
- Rolling insight cards compare the final three recorded values with the preceding three and require six observations. Profile findings identify their final and earlier windows separately. Read each finding's displayed sample and supporting values.
- Your threshold
- A threshold counts recorded results above or equal to a number you enter. It is a research input. Historical frequency does not establish a next-game probability or a sportsbook price.
Metric definitions
These 16 definitions come from the same metric registry used by the evidence API. Position eligibility and a recorded sample determine which metrics are available for a player.
Passing yardsyards
Passing yards reported by nflverse at the player-week grain; totals sum the selected recorded weeks.
Formula version: nflverse-player-week.v1
Grain: player-week · Minimum recorded sample: 1
Descriptive recorded results; not a forecast. Source population excludes zero-opportunity weeks.
Pass attemptsattempts
Pass attempts reported by nflverse at the player-week grain; totals sum the selected recorded weeks.
Formula version: nflverse-player-week.v1
Grain: player-week · Minimum recorded sample: 1
Descriptive recorded results; not a forecast. Source population excludes zero-opportunity weeks.
Completionscompletions
Completions reported by nflverse at the player-week grain; totals sum the selected recorded weeks.
Formula version: nflverse-player-week.v1
Grain: player-week · Minimum recorded sample: 1
Descriptive recorded results; not a forecast. Source population excludes zero-opportunity weeks.
Passing touchdownstouchdowns
Passing touchdowns reported by nflverse at the player-week grain; totals sum the selected recorded weeks.
Formula version: nflverse-player-week.v1
Grain: player-week · Minimum recorded sample: 1
Descriptive recorded results; not a forecast. Source population excludes zero-opportunity weeks.
Passing EPA contributionexpected points
Passing EPA contribution reported by nflverse at the player-week grain; totals sum the selected recorded weeks.
Formula version: nflverse-player-week.v1
Grain: player-week · Minimum recorded sample: 1
Descriptive recorded results; not a forecast. Source population excludes zero-opportunity weeks.
Completion percentage over expectationpercentage points
Provider-reported completion percentage minus modeled expectation, in percentage points. Multiweek averages are unweighted means of reported weekly values.
Formula version: nflverse-player-week.v1
Grain: player-week · Minimum recorded sample: 1
Descriptive recorded results; not a forecast. Source population excludes zero-opportunity weeks.
Carriescarries
Carries reported by nflverse at the player-week grain; totals sum the selected recorded weeks.
Formula version: nflverse-player-week.v1
Grain: player-week · Minimum recorded sample: 1
Descriptive recorded results; not a forecast. Source population excludes zero-opportunity weeks.
Rushing yardsyards
Rushing yards reported by nflverse at the player-week grain; totals sum the selected recorded weeks.
Formula version: nflverse-player-week.v1
Grain: player-week · Minimum recorded sample: 1
Descriptive recorded results; not a forecast. Source population excludes zero-opportunity weeks.
Rushing touchdownstouchdowns
Rushing touchdowns reported by nflverse at the player-week grain; totals sum the selected recorded weeks.
Formula version: nflverse-player-week.v1
Grain: player-week · Minimum recorded sample: 1
Descriptive recorded results; not a forecast. Source population excludes zero-opportunity weeks.
Rushing EPA contributionexpected points
Rushing EPA contribution reported by nflverse at the player-week grain; totals sum the selected recorded weeks.
Formula version: nflverse-player-week.v1
Grain: player-week · Minimum recorded sample: 1
Descriptive recorded results; not a forecast. Source population excludes zero-opportunity weeks.
Targetstargets
Targets reported by nflverse at the player-week grain; totals sum the selected recorded weeks.
Formula version: nflverse-player-week.v1
Grain: player-week · Minimum recorded sample: 1
Descriptive recorded results; not a forecast. Source population excludes zero-opportunity weeks.
Receptionsreceptions
Receptions reported by nflverse at the player-week grain; totals sum the selected recorded weeks.
Formula version: nflverse-player-week.v1
Grain: player-week · Minimum recorded sample: 1
Descriptive recorded results; not a forecast. Source population excludes zero-opportunity weeks.
Receiving yardsyards
Receiving yards reported by nflverse at the player-week grain; totals sum the selected recorded weeks.
Formula version: nflverse-player-week.v1
Grain: player-week · Minimum recorded sample: 1
Descriptive recorded results; not a forecast. Source population excludes zero-opportunity weeks.
Receiving touchdownstouchdowns
Receiving touchdowns reported by nflverse at the player-week grain; totals sum the selected recorded weeks.
Formula version: nflverse-player-week.v1
Grain: player-week · Minimum recorded sample: 1
Descriptive recorded results; not a forecast. Source population excludes zero-opportunity weeks.
Receiving EPA contributionexpected points
Receiving EPA contribution reported by nflverse at the player-week grain; totals sum the selected recorded weeks.
Formula version: nflverse-player-week.v1
Grain: player-week · Minimum recorded sample: 1
Descriptive recorded results; not a forecast. Source population excludes zero-opportunity weeks.
Weekly target sharefraction
Provider-reported fraction of team targets in a recorded player week. Multiweek averages are unweighted means of reported weekly shares.
Formula version: nflverse-player-week.v1
Grain: player-week · Minimum recorded sample: 1
Descriptive recorded results; not a forecast. Source population excludes zero-opportunity weeks.
Player measures & ranks
Player research pages use these 32 measures (formula version augris.metric.v4.7.0). Each one has a minimum sample. Below that minimum, the value is shown with its sample and is not ranked.
- Qualification cohort
- A player is compared only with players at the same listed position, for the same season, phase (regular season or postseason), venue, opponent and last-N-games window. A player joins the cohort only when the measure has a value and its sample meets the minimum.
- Rank and ties
- Rank = 1 + the number of qualified players with a strictly better value. Better means higher, except for measures marked “lower is better”. Descriptive measures such as average depth of target are ordered from highest to lowest and do not imply that higher is better. Players with exactly equal values share a rank, shown as T-#6. The next distinct value skips the shared places (1, 2, 2, 4). Values are compared at full precision and rounded only for display.
- Percentile
- Percentile = 100 × (players with a worse value + half of the players with an equal value, including this player) ÷ cohort size. The median and quartiles use linear interpolation over the same qualified values.
- Recent change
- The last three recorded games in the selected season and phase are compared with the three games before them, ordered by source game ID. At least six recorded games are required, and both windows need a value.
- Missing and zero
- Inactive and missing weeks are not filled with zero. A measured zero shows as 0. A missing value shows as —.
Passing yards / gameyards
sum(passing_yards) / recorded games
Minimum sample: 2 games · Positions: QB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.passing_yards
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Passing TD / gametouchdowns
sum(passing touchdowns) / recorded games
Minimum sample: 2 games · Positions: QB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.passing_tds
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Passing yards / attemptyards
sum(passing_yards) / sum(pass attempts)
Minimum sample: 40 attempts · Positions: QB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.passing_yards, nflverse.stats_player_week.attempts
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Completion ratepercent
100 × sum(completions) / sum(pass attempts)
Minimum sample: 40 attempts · Positions: QB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.completions, nflverse.stats_player_week.attempts
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Passing EPA / attemptEPA
sum(passing EPA) / sum(pass attempts)
Minimum sample: 40 attempts · Positions: QB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.passing_epa, nflverse.stats_player_week.attempts
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Rushing yards / gameyards
sum(rushing_yards) / recorded games
Minimum sample: 2 games · Positions: QB, RB, WR, TE, FB, HB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.rushing_yards
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Carries / gamecarries
sum(carries) / recorded games
Minimum sample: 2 games · Positions: RB, FB, HB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.carries
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Rushing TD / gametouchdowns
sum(rushing touchdowns) / recorded games
Minimum sample: 2 games · Positions: RB, FB, HB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.rushing_tds
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Rushing yards / carryyards
sum(rushing_yards) / sum(carries)
Minimum sample: 20 carries · Positions: QB, RB, WR, TE, FB, HB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.rushing_yards, nflverse.stats_player_week.carries
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Rushing EPA / carryEPA
sum(rushing EPA) / sum(carries)
Minimum sample: 20 carries · Positions: QB, RB, WR, TE, FB, HB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.rushing_epa, nflverse.stats_player_week.carries
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Receiving yards / gameyards
sum(receiving_yards) / recorded games
Minimum sample: 2 games · Positions: RB, WR, TE, FB, HB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.receiving_yards
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Targets / gametargets
sum(targets) / recorded games
Minimum sample: 2 games · Positions: RB, WR, TE, FB, HB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.targets
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Receptions / gamereceptions
sum(receptions) / recorded games
Minimum sample: 2 games · Positions: RB, WR, TE, FB, HB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.receptions
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Receiving TD / gametouchdowns
sum(receiving touchdowns) / recorded games
Minimum sample: 2 games · Positions: RB, WR, TE, FB, HB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.receiving_tds
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Catch ratepercent
100 × receptions / targets
Minimum sample: 15 targets · Positions: RB, WR, TE, FB, HB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.receptions, nflverse.stats_player_week.targets
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Yards / targetyards
receiving yards / targets
Minimum sample: 15 targets · Positions: RB, WR, TE, FB, HB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.receiving_yards, nflverse.stats_player_week.targets
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Yards / receptionyards
receiving yards / receptions
Minimum sample: 10 receptions · Positions: RB, WR, TE, FB, HB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.receiving_yards, nflverse.stats_player_week.receptions
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Receiving EPA / targetEPA
receiving EPA / targets
Minimum sample: 15 targets · Positions: RB, WR, TE, FB, HB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.receiving_epa, nflverse.stats_player_week.targets
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Target sharepercent
100 × mean(provider-reported weekly target share)
Minimum sample: 2 games · Positions: RB, WR, TE, FB, HB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.target_share
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Receiving TD ratepercent
100 × receiving touchdowns / targets
Minimum sample: 15 targets · Positions: RB, WR, TE, FB, HB · Higher is better
Source: nflverse weekly player statistics · nflverse.stats_player_week.receiving_tds, nflverse.stats_player_week.targets
Descriptive source-recorded results; inactive and missing weeks are not zero-filled.
Receiving success ratepercent
100 × successful recorded targets / targets with success labels
Minimum sample: 15 targets · Positions: RB, WR, TE, FB, HB · Higher is better
Source: charted play-by-play (nflverse participation and FTN charting) · nflverse.pb_participation, nflverse.play_by_play.success
Descriptive charted plays only; availability follows the source charting coverage.
Explosive target ratepercent
100 × targets gaining at least 20 receiving yards / targets
Minimum sample: 15 targets · Positions: RB, WR, TE, FB, HB · Higher is better
Source: charted play-by-play (nflverse participation and FTN charting) · nflverse.pb_participation, nflverse.play_by_play.receiving_yards
Descriptive charted plays only; availability follows the source charting coverage.
Average depth of targetyards
mean air yards across targets with charted air yards
Minimum sample: 15 targets · Positions: RB, WR, TE, FB, HB · Descriptive (ranked highest to lowest)
Source: charted play-by-play (nflverse participation and FTN charting) · nflverse.pb_participation, nflverse.play_by_play.air_yards
Descriptive charted plays only; availability follows the source charting coverage.
YAC / receptionyards
sum(YAC) / receptions with charted YAC
Minimum sample: 10 receptions · Positions: RB, WR, TE, FB, HB · Higher is better
Source: charted play-by-play (nflverse participation and FTN charting) · nflverse.pb_participation, nflverse.play_by_play.yards_after_catch
Descriptive charted plays only; availability follows the source charting coverage.
Catchable-target ratepercent
100 × catchable targets / targets with catchability labels
Minimum sample: 15 targets · Positions: RB, WR, TE, FB, HB · Higher is better
Source: charted play-by-play (nflverse participation and FTN charting) · nflverse.pb_participation, nflverse.ngs_passing.catchable
Descriptive charted plays only; availability follows the source charting coverage.
Drop ratepercent
100 × drops / catchable targets with drop labels
Minimum sample: 15 catchable targets · Positions: RB, WR, TE, FB, HB · Lower is better
Source: charted play-by-play (nflverse participation and FTN charting) · nflverse.pb_participation, nflverse.ngs_passing.drop
Descriptive charted plays only; availability follows the source charting coverage.
Red-zone target sharepercent
100 × player red-zone targets / team red-zone targets
Minimum sample: 5 targets · Positions: RB, WR, TE, FB, HB · Higher is better
Source: charted play-by-play (nflverse participation and FTN charting) · nflverse.pb_participation, nflverse.play_by_play.yardline_100
Descriptive charted plays only; availability follows the source charting coverage.
Third-down yards / targetyards
third-down receiving yards / third-down targets
Minimum sample: 5 targets · Positions: RB, WR, TE, FB, HB · Higher is better
Source: charted play-by-play (nflverse participation and FTN charting) · nflverse.pb_participation, nflverse.play_by_play.down
Descriptive charted plays only; availability follows the source charting coverage.
Success rate under pressurepercent
100 × successful pressured pass plays / pressured pass plays
Minimum sample: 15 pressured plays · Positions: QB · Higher is better
Source: charted play-by-play (nflverse participation and FTN charting) · nflverse.pb_participation, nflverse.pbp_participation.was_pressure
Descriptive charted plays only; availability follows the source charting coverage.
Clean-pocket success ratepercent
100 × successful non-pressured pass plays / non-pressured pass plays
Minimum sample: 40 clean-pocket plays · Positions: QB · Higher is better
Source: charted play-by-play (nflverse participation and FTN charting) · nflverse.pb_participation, nflverse.pbp_participation.was_pressure
Descriptive charted plays only; availability follows the source charting coverage.
Play-action EPA / playEPA
mean EPA on charted play-action pass plays
Minimum sample: 15 play-action plays · Positions: QB · Higher is better
Source: charted play-by-play (nflverse participation and FTN charting) · nflverse.pb_participation, nflverse.pbp_participation.play_action
Descriptive charted plays only; availability follows the source charting coverage.
Yards / carry by box sampleyards
rushing yards / carries with charted box count
Minimum sample: 20 charted carries · Positions: RB, HB, FB · Higher is better
Source: charted play-by-play (nflverse participation and FTN charting) · nflverse.pb_participation, nflverse.pbp_participation.defenders_in_box
Descriptive charted plays only; availability follows the source charting coverage.
Model estimates
Game Intelligence shows two model estimates. Each is a statistical estimate with a range. It is not a guarantee, a probability of winning, or a betting recommendation. Both use only games completed before kickoff, so the target game never informs its own estimate.
- Game margin (augris.game-margin-rolling-1)
- Each team's last eight completed games before kickoff are used. These can cross seasons and phases. A team's expected points = the average of its points scored and its opponent's points allowed over those games. The estimate is the home team's expected points minus the away team's. The 80% range adds the 10th and 90th percentiles of past errors to the estimate. Those errors come from rerunning the same rule at every earlier game (rolling origin, prior games only). The sample is the number of team games used, up to 16. If either team has fewer than four prior completed games, the estimate is withheld.
- Player production (augris.player-production-rolling-1)
- The estimate is the median of the player's last eight regular-season games before kickoff. It covers passing yards for quarterbacks, rushing yards for running backs, and receiving yards for other positions. Only games with a recorded value and a recorded opportunity count are used. The 80% range comes from the 10th and 90th percentile errors of the same median rule, rerun over the player's earlier games. The lower bound is never below 0. The sample is the number of games used, up to eight. With fewer than four games, the estimate is withheld.
- Model history label
- “Deep”, “moderate” or “limited model history” describes only how many past estimates and prior games exist. For the game margin: deep = at least 100 past estimates and eight games per team; moderate = at least 40 and six. For players, moderate = at least 12 past estimates and eight games. It is not a statement of accuracy.
- What is shown about error
- Each game estimate shows its own count of past estimates, mean absolute error, and the share of past errors that fell inside the 80% range. These figures are recomputed from prior games for that game. Augris makes no calibration claim beyond them.
- Not inputs
- Injuries, confirmed availability, snap share, weather, individual defender assignments, and market prices are not inputs to either estimate.
The touchdown opportunity model in the codebase (td-opportunity-logistic-1) is a research baseline. Augris does not publish its probabilities.
Freshness & limits
Evidence separates the time a source was observed from the period its data describes. For market observations, retrieval time and the book's update time are separate. Fetching a quote again does not make the underlying quote new.
- Fresh
- A source response meets its freshness check. This label alone does not establish live market coverage or an executable offer.
- Retained
- A historical generation or earlier source snapshot is preserved for inspection. Check the stated season, phase, and timestamps before using it.
- Delayed or unavailable
- A usable update, observation time, identity, or required field is missing. The response explains the limitation; a missing value is not a measured zero.
Market context is limited to attributed pregame observations when supplied. A player prop requires a verified player, named book, market period, side, line, price, and observation times. Unsupported props, pressure, route, and coverage datasets remain unavailable.
Research measures describe recorded results. The model estimates above are labelled as estimates with a range and a sample. Augris does not claim a validated predictive edge.
Keep the original reasoning.
Saving research preserves the original evidence snapshot beside your thesis and invalidation condition. Later edits create revisions. A new source generation does not silently replace the evidence you originally saved.
A saved-versus-current review is a comparison of the supplied snapshots. Inspect the changed values and source context before deciding whether your thesis has changed. A change in generation alone does not establish a material change in the game.
Journal odds, stake, conclusions, and results are your recorded inputs. They do not become independently verified market observations or settled outcomes through being saved. Use Notebook export to retain a copy of your research.
Return to Augris