Bring this to the group chat

Show your work.

Every leaderboard has a rule. Here are ours, including the parts the data cannot settle.

Weekly awards and lineup ceilings

Highest and lowest commissioner-adjusted team score among completed paired matchups. Ties share the award. Byes and unpaired rosters do not qualify. Win margin is team score minus opponent score. Starter points come from Sleeper's ordered starter scores, with player-score maps as a fallback.

Bench points sum the nonstarters in the weekly roster snapshot. The snapshot ceiling is the maximum-scoring assignment of those players to the league's starting slots, with every player used at most once and FLEX eligibility respected. Possible bench gain is ceiling minus actual starter points, floored at zero. Team score overrides are not added to player totals.

This is a hindsight estimate. Public weekly snapshots do not fully preserve taxi status, IR status, historical position eligibility, or kickoff locks. Explicit reserve/taxi IDs are excluded where provided. Missing positions or scores make the ceiling unknown. A potential lineup swap is not proof that a manager could legally have made it at kickoff.

Power rankings

Power = 100 × [0.50 × percentile of season points per paired week + 0.30 × percentile of the last three paired weeks + 0.20 × average all-play win fraction]. All-play compares each score with every other paired team that week. Ties count half, and percentile ties use midranks. This is an editorial ranking, not playoff seeding.

Historical records and identities

Every linked previous_league_id season is included once backfill finishes. Regular-season head-to-head wins are calculated from weekly pairings; ties count half in win percentage. League-median bonus games are excluded from career head-to-head records. Sleeper's official season records and points-for totals appear separately. Rankings by record then points for are display order, not a reconstruction of division or commissioner seeding.

Weekly records include regular season and playoffs. Multiweek playoff rounds are shown as scoring weeks; advancement and championships use the winners bracket. Only a first-place bracket match with p=1 and an explicit winner creates a championship. Playoff appearances require a recorded playoff week, so preseason brackets do not count.

Manager careers use stable Sleeper user IDs, not team names or matching roster numbers. A season/team page remains tied to its exact league ID and roster ID. Midseason changes are not fully archived by Sleeper; the commissioner can supply week-specific ownership overrides in config/identities.json. Historical team names are the names available when the season was captured.

Win/loss streaks shown here are within a regular season and require consecutive observed scoring weeks. A tie or missing week breaks the streak. This avoids joining incomplete history into an invented run. Best team-season tables require a completed regular season. Player totals count only weeks in which the player appears on a saved league roster, including bench points. They are not full NFL career statistics.

Waiver value

Track each completed waiver/free-agent addition from the next scoring week until the first known drop, trade, ownership change, or missing roster membership. The acquisition and disposal weeks are excluded because the public API does not reliably establish the player's kickoff timing. Holding periods may continue into a renewed season when the same manager and roster membership can be established.

Roster points include started and benched points. Started points count only actual starts. Replacement is the median observed bench score of other players with compatible fantasy positions in our league that week. Realized points over replacement sum player points minus that median for starts only. Values may be negative. An unused pickup has zero realized lineup value.

Rank by points over replacement, then started points, then roster points. Pickups with missing baselines or known scoring gaps are shown separately and remain unranked, rather than treating missing values as zero. This measures use against a league-local bench proxy, not the best available free agent or a projection. Latest-week rankings measure returns this week, including pickups made earlier; season and career rankings use their respective return windows.

Trade value and winners

For each side, add started points produced by received players while they remained with the receiving manager. Once an asset leaves, its later scoring stops counting for that trade. Trade-week scoring is excluded. Multi-team deals keep every side separate. Net edge equals that side's started points minus the mean of the other sides' returns.

A scoring winner needs at least three observed post-trade weeks, no unresolved picks, and no known missing scoring samples. Otherwise the page names only a provisional scoring leader or says too early. A tie stays a tie. The leaderboard sums net edges and shows pending-trade counts. Wins are based on fantasy production so far, not an assertion that a dynasty rebuild was a mistake.

Future picks stay pending. A pick is resolved only when its year, round, original roster lineage, draft-slot mapping, and receiving roster identify a unique actual selection. That player's subsequent starts may then count. Picks traded again remain unresolved here; downstream trade-tree assets are not guessed. FAAB is shown separately and has no invented points exchange rate.

Dynasty market-value changes are unavailable. The documented free Sleeper API has no historical consensus dynasty-price series. Age curves and pick-round units in the newsletter remain explicitly labeled planning heuristics. They are not market-price changes and are not mixed with fantasy-point returns.

Drafts

Evaluate each completed draft selection by the player's started points for the drafting manager in that first season, ending at departure. Subtract the median return of selections in the same round of the same draft. Sum those surpluses for team draft results. Historical completed seasons populate best/worst drafts; current-season pick results remain provisional.

This measures realized lineup production and use. It favors immediate contributors and does not measure full dynasty talent, future development, price paid in an auction, or return from later trades. Missing ownership and draft-slot data are reported as unknown.

What refreshes

Completed weekly score snapshots and older transaction buckets are frozen JSON. Historical seasons are fetched once. Normal runs refresh the active scoring/transaction bucket plus league, roster, user, bracket, and draft-status metadata needed to detect changes. The full player catalog is refreshed at most once daily. Advancing to a new week finalizes the preceding snapshot once.

Late stat corrections and moves added to old buckets require an explicit refresh/backfill. This is intentional: immutable history and automatically changing old results cannot both be guaranteed. Data coverage is shown separately. No missing value is silently substituted for a confirmed score.