Sources and scoring
League facts
Sleeper’s public API supplies profile identity, leagues, rosters, scoring settings, matchups, trades and picks. A public username finds a profile; it does not prove ownership of that Sleeper account.
Upstream league facts refresh on a 60 second source cache. Your workspace retains its latest view while refreshing in the background every five minutes. The full player directory refreshes daily, as Sleeper requests. Shared caching keeps repeated player visits light.
Player production
nflverse supplies historical regular season box scores under CC BY 4.0. Identity joins use the DynastyProcess player ID database when Sleeper omits an NFL ID. Game logs include available current and prior seasons. Career research retains available history from 2016 and earlier career seasons where the source supports them.
The last completed season is always labeled. Postseason records are excluded from regular season totals.
Projections and lineup plans
Weekly projections blend fresh Sleeper and ESPN raw stat lines with equal provider weights where identity, season, week and core scoring coverage match. Each included stat is multiplied by your league’s scoring weight, including position reception premiums. Only scoring events shared by the included providers enter the blend. Missing bonuses and incidental events stay labeled as unmodeled. ESPN default fantasy totals never substitute for your settings. Both feeds are supplemental public endpoints and can be unavailable.
Source caches refresh after ten minutes. Open workspaces recheck league data every five minutes and on returning to the tab; the existing hourly maintenance task also refreshes shared projection feeds. Player pages show source inclusion, weights and retrieval times. The matchup and trade planners use Juncture models where ready and the fresh provider blend elsewhere. They find the highest total among eligible players. Confirmed byes, IR, taxi and unavailable players are excluded. These are estimates, not guaranteed outcomes, and kickoff locks still need checking in Sleeper.
News, injuries and discussion
ESPN national and team RSS feeds plus official NFL team news and video feeds supply dated reporting without paid keys. Matched reports are saved to the player history. Team context is separate. Interview titles are not transcripts. Optional approved X and Reddit connections remain community opinion.
nflverse’s injury feed stopped after 2024. The assessment also checks ESPN’s current structured availability feed by exact provider ID. Sleeper directory flags remain separately attributed. A missing injury entry is not clearance, and disagreements require review. Confirm availability with the official NFL injury report. Source details show actual fetch dates and outages. Team feeds refresh on use after ten minutes and ESPN after thirty minutes. Saved reports retain their publication dates.
Independent market reference
Values by RosterAudit.com supplies an additional dynasty player and pick market perspective. Its units remain separate from Juncture Value and weekly points. The basic feed uses one QB or superflex PPR; differences from your league are labeled. Future picks use early, middle and late scenarios. Incomplete or conflicting pick tiers are excluded. Complete source tiers retain their published values; ranges use their minimum and maximum. Values are retrieved on an hourly source cache with visible retrieval times.
How player assessments work
Player dossiers hold saved production, availability, news and change history. Profiles, Team, Matchup and trade comparisons use the same released league forecast. Provider event projections are scored under the same supported league rules. Historical workload and efficiency remain separate research. Only adjustments that pass the prospective forecast checks can change the shared forecast. Prior season workload is not treated as current momentum, and new model revisions cannot borrow old validation results. Confirmed absences and byes exclude a player from the weekly lineup model.
Prebuilt market baselines match current CBS Sports and Rotoworld draft boards, plus ESPN and RotoBaller dynasty boards. They are stored by exact player identity and refreshed separately from news. Rank ranges show outside market context, not Juncture trade prices.
Juncture combines dated player reporting, current availability, regular season production, comparable workload samples and league scoring. Only narrowly matched, recent reports create availability or role review actions. Stale reports, social posts, questions and possible outcomes cannot establish a player status.
Historical production remains labeled by season. Missing sources, partial scoring and a lack of current season games lower the evidence rating. That rating describes coverage, not a measured probability of being correct. News never adds arbitrary points to a projection. A roster or trade assessment can flag new risk without inventing a market price.
Measuring forecast accuracy
One forecast per player, league and week is recorded before the game date. Later game records are scored with the same modeled league rules, then compared with that saved forecast and provider baseline. Profiles show sample size and average absolute error. The model is experimental; no measured accuracy rate is claimed until scored outcomes exist.
No invented certainty
Juncture Value compares all positions on one scale. 100 means perfect; leading a ranking never automatically earns 100. Overall market value anchors the assessment, with bounded production and current evidence adjustments. League quarterback demand affects scarcity. Health and youth alone cannot create value. Production only assessments are labelled Provisional when no compatible market board is matched. Source identity matching normalizes suffixes and equivalent team abbreviations, and ambiguous matches remain unavailable. Defenses and kickers have streaming value far below elite offensive players. The score is a model judgment, not a measured probability or trade currency.
Roster depth counts are not talent grades. Expert rank disagreement is displayed instead of averaged away. All play records describe results, not future odds. Trade advice uses the selected goal, individual Juncture Values, current role and injury evidence, supported prospect upside, positional depth, both eligible lineups and planned drops. Scores on the relative 0 to 100 scale are never added together as trade currency. Draft picks retain their exact year, round and original team, but unmodeled pick prices keep any fairness verdict provisional. A three point gap between the strongest players is a review flag, not a trade price. A ten point season value loss cannot be justified by a weekly gain alone. Recommendations say accept, counter, pass or wait and explain which value, lineup, evidence or roster issue controls the call. Weekly comparisons cover only starting slots the deal can change; missing projections in those slots remain unknown. Unrelated unchanged slots do not block the comparison. No acceptance probability or guaranteed winner is claimed.
The core tools are free. Sleeper requires a licensing discussion for commercial API use. No paid plan is enabled.