How the
paper is made

The season leaves a mark

The page carries a subtle stock tint that follows the streak. Compare the original five paper conditions in this specimen; it does not change the archive or the current streak.

A win resets the paper. Losses progress through one–two, three–four, five–six, and seven or more games. A tie uses the first worn condition. Text describes every state, so the meaning survives without the color or texture.

Follow the team, with each source in view

The news desk carries headlines and original publication times from the official Jets news feed. Each headline opens the original story. Publication time describes the article; the successful-check time describes when this edition fetched the feed.

The Sunday briefing compares validated offensive and defensive samples. Each side shows EPA per dropback or non-scramble rush, its included play count, and completed-game count. Defensive rates retain their “EPA allowed” label. The two units faced different opponents, so the comparison describes recorded performance rather than a predicted matchup result. News labels describe explicit headline wording; they do not add reporting.

The game-day desk uses separately reviewed official viewing guidance and dated team availability reporting. Each curated briefing belongs to one fixture and expires at kickoff. Its notes keep their publication and review dates, and warn when the review is more than 24 hours old. Practice participation and roster status do not establish game-day clearance.

The roster uses the latest regular-season week in the nflverse roster files. Active, practice-squad, and reserve labels describe source roster membership. They do not establish injury status or availability for a game. Profiles are matched by GSIS identifiers, with validated ESPN identifiers as a fallback. Entries without either identifier are counted visibly as awaiting identifiers. Headshots come from the source and fall back to initials when unavailable.

Passing, rushing, and receiving totals come from nflverse weekly player statistics, restricted to Jets regular-season games confirmed final by the schedule. Each game's passing completions, yards, and touchdowns must reconcile with its receiving totals before publication; replacements cannot silently lose previously verified player-game rows. Yardage supplies the leader ordering. A player's recorded-game count includes games with a source statistics row, rather than all roster appearances. Defensive, special-teams, and preseason production are outside this display. Missing player statistics are described as missing, rather than presented as zero production.

News, roster, and player statistics have independent successful-check times and cutoffs. If one source fails validation or fetching, its previous valid data remains visible with a warning. Without a previous snapshot, that section is marked unavailable. Previous-season roster records keep their source-season label; they cannot supply statistics for the current edition. A reader-side clock marks source checks older than 24 hours as overdue.

Follow a football season across the calendar

The season directory groups recorded finals, usable game cases, sourced moments and explicitly tagged media by football season. January 2011 playoff games belong to 2010. The regular-season and playoff controls change the statistical sample; search narrows the displayed games, facts and reporting without changing that sample's totals.

Archived scores come from nflverse play-by-play final-score fields; newly ingested analysis is checked against schedule finals. Current-season schedule-confirmed finals take precedence. A valid final can remain visible when its play-by-play analysis is held. Points per game use recorded final counts; scoring margins subtract points allowed from points scored. Games within eight points are a final-margin count, not a measure of in-game tension. Historical game coverage begins in 1999; earlier entries contain selected verified moments. Historical player volume comparisons are available where the complete league sample passes checks; historical EPA splits are not included.

The same season, against the league

Team scoring ranks use every schedule-confirmed league final in the selected phase. Yardage ranks additionally require both teams' complete weekly statistics for every included game. Net passing adds the source's negative sack-yardage field to gross passing yards. Defensive yardage mirrors each opponent's offense. Per-game values compare exact totals and game counts before display rounding; equal values share competition ranks, such as 1, 1, 3. The population is the actual participating teams, including 31 teams in 1999 and only playoff participants in postseason views.

Individual volume ranks come from nflverse weekly player statistics, reconciled against team statistics. Attempt, carry, target or credited-event requirements are shown beside each metric. Missing identities, fields or unreconciled totals withhold the affected metric. A player's full all-club total determines the league rank; Jets stint production is shown separately for traded players. Combined-season totals include extra postseason games and are labeled accordingly. Rank snapshots retain their original successful-check time if a later refresh fails.

NFL Next Gen Stats measurements are distributed through the nflverse tracking release under CC BY 4.0. Available coverage begins in 2016. This edition uses published regular-season aggregates and actual attempt or target samples, rather than inferred tracking coordinates or a whole-roster sample. Traded-player season aggregates can include other clubs. Completion over expected is measured in percentage points; rushing over-expected fractions are converted to percentages once. The source's qualification rules determine which players appear; no unverified tracking percentile or rank is generated.

Selected PFF figures are small, attributed excerpts from public pages, with the football season, published scope, comparison population where stated, source date and manual-review date attached. Grades use PFF's 0–100 assessment scale; they are distinct from counting-stat ranks and NGS model estimates. A 2026 article reviewing 2025 does not become a 2026 performance grade. Comprehensive historical facets or a continuous grades feed require authorized access and display rights. Publication timestamps use Eastern time; date-only publisher datelines retain their stated calendar date.

Original reporting, with attribution and dates

The Media Room is a manually reviewed collection of beat reporting, sports television, radio, official film and independent analysis. Each selection carries an outlet, author, original URL, publication date where verified, explicit football-season tags and a brief original synopsis. The collection check time describes the editorial review; it is not a continuously refreshed news-feed timestamp. Source coverage bars count items in the current selection, rather than measuring reach, engagement or audience sentiment.

Video thumbnails belong to their original YouTube publishers. Supported YouTube players and original X posts load only after a reader asks; their providers control playback and post availability. Missing or restricted embeds retain a direct original-source link. Comparing two selections places their actual metadata and summaries together, without generating an attributed opinion or imitating an author's voice.

The official Jets RSS desk refreshes separately. Validated entries with future publication times are withheld until a later check; their dates are never rewritten to appear current.

Selected cases, with sources attached

The classic case files are editorial selections, separate from the probability rankings. Their dates and scores must match the checked archive before a game link appears. Each case links to an official account supporting its historical context. The Super Bowl III feature links to the Jets’ historical account; the play-by-play archive begins in 1999 and supplies no tape for the 1968 season.

The rivalry ledger covers regular-season results against today’s AFC East opponents: Buffalo, Miami and New England. It gives the included game count and season range, rather than claiming an all-time series record or complete historical division record. Playoffs are excluded. Current schedule-confirmed finals replace older analysis scores; unconfirmed fixtures do not count. An unreconciled archive score is omitted unless a confirmed current result supplies the score. Game-analysis links additionally require matching, eligible play-by-play.

The personal game-day ticket saves user-selected scores, conviction and ritual in this browser when storage is available. It starts blank and has no public poll or community tally. Fixture changes require a new review; the form closes at kickoff, or at the start of the listed game date in America/Chicago when kickoff is unconfirmed. A saved call can be compared with a schedule-confirmed final, including a tie. Browser storage is editable, so the ticket is a personal record rather than a verified prediction contest. “Copy saved ticket” copies text to the clipboard; it sends no message.

Film study, with the source in view

The Film Room pairs four curated archive plays with official NFL or Jets replays and gamebooks. Down, distance, field position and scoring context come from those books. Attributed source accounts identify a reported route or a player’s explanation; they do not establish the full coverage, blitz or protection call.

Win probability is the archive’s pre-play estimate; the displayed change is the reported model delta. Miami’s older touchdown row repeats a clock from the preceding play, so its header shows the quarter and its notebook gives the separately documented scoring clock. The overtime kick retains its pre-kick estimate rather than replacing it with the final outcome.

The coverage board is an independent teaching schematic, with illustrative positions, routes and eleven assigned defenders. It supplies no All-22 footage, player tracking or verified video offsets. Coverage and pressure controls change the lesson, not our classification of the selected play. Photo-style scene recreations are visibly labeled as AI-generated and linked to historical references; they are editorial illustrations, never film evidence.

Your playbook, in motion

The Playbook Lab is a diagramming tool. Its offensive presets illustrate common alignments with eleven players, five interior linemen, two eligible ends and four players in the backfield. Personnel notation counts running backs and tight ends; it does not identify a formation. Defensive presets describe personnel and alignment, rather than establishing a coverage or how many defenders rush. See the NFL formation guide and the 2026 NFL rulebook.

Preset concepts and custom movement paths are teaching examples. Players travel along those paths on a six-second diagram timeline, optionally within a chosen movement window. The ball follows the selected carrier and an illustrative transfer, or a staged sequence of flights, loose-ball movement and possession changes. This motion does not calculate player speed, blocking, collisions, pass completion, quarterback reads or the result of a play. A defender following a route illustrates a chosen assignment; it does not solve real man coverage.

The Jets archive keeps recorded actions separate from supporting study choices. Full-snap study gives all 22 players movement and a position-specific note: protection, rush, releases and defensive reactions. Supporting assignments are illustrative; no historical coverage, blitz or protection call has been verified. Source-supported action limits movement to the documented central actors and leaves unknown assignments blank. A stationary marker in that mode does not assert that the real player stood still. Both modes use schematic positions, geometry, spacing, ball flight and diagram time.

Broad actions were observed in the official Jets Wilson breakdown at 1:15 and 1:18: blockers engaging rushers, other eligible receivers releasing and defenders moving with routes. Those frames do not establish all 22 identities or individual responsibilities. Elliott’s official gamebook establishes play-action and his eligible status; the fake back’s identity remains unknown. Sources sit beside the board. The Butt Fumble study includes Gregory’s illustrative approach, recovery and return without collision physics or a judgment of Moore’s blocking. Edited movement is labeled separately from the original study plan; Undo or Restore recovers the selected mode’s original diagram.

Moving a player freely can produce an illegal formation. The editor preserves eleven players on each side, but imported and custom alignments are not certified as legal NFL plays. No historical play in the source notebook inherits an assignment from the lab.

Saved diagrams remain in this browser when storage is available. Export downloads a play file; import validates its structure and field coordinates before loading. A shared link contains the diagram itself and loads it into the editor. These actions do not publish a community playbook or send a message.

Team efficiency, with the sample in view

Expected points added (EPA) measures how a play changes the offense's expected scoring position. EPA per play is the sum divided by the number of included plays. Higher offensive EPA is better; lower defensive EPA allowed is better. Success rate is the share of those plays with EPA above zero.

The sample includes current-season regular-season games confirmed final by the schedule, with completed, score-reconciled play-by-play. Included plays have a valid offense and defense, down 1–4, a run or pass play type, and finite EPA. No-plays, kneels, spikes, and two-point attempts are excluded. These are pooled play rates, rather than averages of game averages.

Dropbacks include passes, sacks, and quarterback scrambles marked as dropbacks in the source. The rushing split uses the remaining included non-dropback plays; it excludes quarterback scrambles and may include aborted run snaps. Empty samples show a dash. League ranks compare eligible team samples using unrounded rates and give tied values the same rank; the next rank skips the tied places. Displayed EPA rates use three decimals in the Sunday briefing and league/unit comparisons; each rate carries its own play count.

The efficiency snapshot includes 65 completed games and 32 teams through Oct 8, 2026. Week 5 is the highest included week, rather than a claim that every game in that week has been analyzed. 0 confirmed league finals await usable analysis. Last analysis update: Oct 9, 2026, 1:22 AM CDT.

League averages pool team rates by their included play counts, counting each offensive play once. Zero EPA is the model baseline; the observed league average can differ from zero. All game situations are included and rates are not opponent-adjusted. Offense-versus-defense tables describe what each unit has recorded against its own opponents; they do not calculate a predicted matchup advantage.

Early-season samples are small. The comparison describes recorded performance and does not predict who will win the next game. Read the nflfastR guide to EPA and win probability.

The ranking is a calculation

Heartbreak ranks losses by the highest Jets win probability reached in the second half. Miracle ranks wins by the lowest probability reached in the second half. Overtime is included.

The latest confirmed current-season final supplies the lead. Its probability analysis appears only when the play-by-play passes integrity checks and agrees with the confirmed score. A result awaiting analysis still counts toward the record and streak. When the season has no confirmed final, an older archive feature is clearly labeled.

The curve shows the Jets’ estimated chance to win before each recorded play. “Change on this play” is the source WPA oriented to the Jets and expressed in percentage points, not a relative percentage change or an individual player grade. Inspect the nflfastR probability field definitions.

These are model estimates at recorded moments. The analyzed archive covers the 1999–2026 seasons, through Oct 4, 2026; the paper is a postgame edition.

The useful finding was a bad number

A quarter-ending administrative row in the 2000 Oakland game carried a 99.4% win probability while nearby real snaps were around 4%. Checking only for a non-null possession team did not remove it: the row contained an empty string.

The extraction now requires a possession team that is neither null nor empty, plus a play type, before reading probability. The probability curve and second-half extrema retain recorded rows such as nullified plays, kneels, and spikes when the possession and probability are valid. Selecting the decisive play excludes no-plays, kneels, and spikes. The source query keeps those rules beside the calculation.

Keep the exception visible

The current dataset contains 453 games; 452 qualify for the two rankings. 1 game is flagged because the running play-by-play score does not reconcile with the recorded final score.

• 2002_04_NYJ_JAX2002-09-29 · at JAX · Jets 3, opponent 28

Flagged records remain in the dataset for inspection but cannot win a probability ranking. A flagged latest result can supply the confirmed score on the front page, with its probability analysis held for review. No replacement probability is invented.

A small system with a visible chain of decisions

The results and schedule are checked separately from the play-by-play analysis. DuckDB reads the source Parquet files during data preparation and writes static JSON. Next.js prints the edition from that checked data. Small React controls handle archive filters and play-by-play exploration. Editorial photographs come from official Jets coverage. Captions identify the game, practice, or archive context, and link to the original source.

Results last checked Oct 9, 2026, 1:22 AM CDT. Analysis last updated Oct 5, 2026, 11:55 AM CDT. Inspect the results and schedule source.

A temporary current-season play-by-play outage retains the last verified analysis while independently checked results and team coverage advance. The successful analysis check, analysis content update, and pending game counts describe different things. Invalid analysis still stops publication for review.

Each eligible game and current roster player has a dedicated page and share card. The interactive archive and roster keep their filters and play controls. Vercel Web Analytics and Speed Insights provide audience and performance measurement when enabled for this project; page URLs are stripped of searches and fragments before sending. Personal game-day tickets stay in browser storage.

There is no runtime language-model call deciding the headline. The source data, rules, and output can be inspected independently.

• Original data: nflverse
• Editorial and paper rules
• Source and local setup

Return to The Back Page →