Game Day
Roster 5 · Live league connection
Live NFL week is not verified yet. The uploaded Week 9 snapshot is historical; verified scoring settings remain available offline.
Data quality & verification
Data confidence
Verified: uploaded 2026 Sleeper league settings and scoring format.
Live when available: Sleeper roster, weekly matchup and league week, with timestamp.
Not yet verified: official 2026 NFL defensive scoring, player projections, injury feeds and win probabilities.
Historical research archive
Research archive
Previous dashboard's 2026 statistics, historical analyses and experimental models are retained separately to avoid mixing them with the live roster.
Open research archiveMy Roster
Sync & lineup verification Details
Historical PPR research Not weekly projections
Archived historical PPR estimates
Weeks 1–6 CSV-derived estimates, not current-week forecasts. The source PPR convention charges −2 per interception and may differ from Berry’s Dynasty scoring.
Player identities and status labels come from a timestamped Sleeper player-directory snapshot. Injury designations can change after retrieval.
Weekly Matchup
Scores are recorded points returned by Sleeper, not projections. A zero does not establish whether a player has played. No automatic winner declaration before final status is confirmed.
Uploaded Week 9 snapshot (historical)
This is the Week 9 matchup file you previously provided. It is not a live feed and does not replace a current Sleeper sync.
Player Compare
About this historical comparison
These figures use Weeks 1–6 of 2026 only. They exclude injury status, opponent, snaps and current usage. A higher number is not automatically a better start.
Research methodology & limitations
Data provenance
The Research panel shows the corrected Weeks 7–9 comparison for the archived CSV-derived model. Outcomes require player-level offensive snaps or recorded events; the earlier gms_active-only report was invalid and has been superseded.
This result requires player-level snaps or events and does not evaluate availability. In the corrected report, 208 candidate rows lack stats, 128 rows lack player-level participation, and 8 identities lack qualifying training records. It is not current-week start/sit advice.
Lineup Lab
Model limitations & validation
The archived CSV-derived V14 estimate had MAE 4.503 versus 4.513 for the simple baseline after correcting the participation filter; the margin is small and bias moved upward. It is not evidence of a reliable forecast. No current injuries, matchup, bye weeks or live opportunity changes are included.
Weekly recommendation verification
Weekly recommendation readiness
Awaiting verification. Historical optimizer remains research-only.
Saved roster snapshot is an explicitly labeled historical example, not your verified current roster. The calculation never fills missing player values with zero.
Hypothetical lineup scenario lab
Enter your own PPR assumptions. These values are NOT verified forecasts. A live Sleeper roster is required; reserve and taxi players are excluded.
Research & Models
Chronological opportunity baseline evaluation
Rolling mean, last-active-game and position-average baselines are calculated from earlier Sleeper weeks only. This retrospective test is conditional on player participation; it does not estimate availability or produce a current-week projection. Includes coverage exclusions and source hashes.
Open reproducible evaluation reportRebuild locally with npm run backtest:rolling-opportunity.
V33: prior-only opportunity-to-points decomposition and baseline comparison. Historical only; missing fields and source conflicts are documented.
Open efficiency bridge backtestOpen source discrepancy auditOpen roster PPR score reconciliation (127 rows)Open role-signal backtestOpen air-yards and role-change backtestOpen multi-season air-yards validation (2022–2025)Open four-season full-PPR baseline evaluationOpen 2026 Week 1–9 baseline validationOpen rolling residual interval coverage (historical)Open roster availability evidence auditThe archived Sleeper directory has player identity coverage but lacks per-player news timestamps. Its six raw injury labels are unverified and are not used to adjust projections.
Walk-forward evaluation applies live-verified Berry scoring to earlier player records. Same-season mean MAE 4.524 vs cross-season mean 4.692 on 10,702 common historical rows; no availability, opponent or role adjustment is included. Historical baseline only, not current-week start/sit advice.
Sleeper availability labels
Sleeper directory labels are not NFL team availability reports. Missing player news timestamps are unverified; timestamps older than seven days or directory snapshots older than 24 hours are marked stale. A recent timestamp does not confirm game availability. The full Sleeper directory is fetched at most once per 24 hours.
Live roster stats
On-demand Sleeper weekly stats for the verified roster. This is recorded statistical activity, not a projection. Missing fields and player rows stay missing; Sleeper's generic PPR value is not Berry's scoring.
League activity & dynasty picks
Official Sleeper league transaction and traded-pick records. Transactions are shown only for their recorded week and status; player labels come from the dated Sleeper directory snapshot. Pick data reports original/current roster ownership, not draft order or player value.
Traded picks
Archived V14 CSV model · exact-scoring evaluation
These results evaluate the former CSV-derived opportunity model, not the current Sleeper-ID component baseline. The holdout requires player-level offensive snaps or recorded events. Availability and start/sit quality are not evaluated.
Select a player to inspect expected opportunities.
Legacy CSV PPR series · Weeks 1–6
Archived CSV model: 85% season-to-date average plus 15% last-three-game average. Its old source PPR scoring is not Berry’s Dynasty scoring, and this metric is separate from the exact-scoring Weeks 7–9 backtest above. No injury, matchup, weather or depth-chart adjustments.
This archived MAE is measured against the uploaded CSV’s PPR column, not Berry’s Dynasty scoring. It is not evidence of current league-scoring performance.
Weekly Sleeper stats explorer
Historical 2026 Weeks 1–9 Sleeper snapshots. Player name, position and team are joined by Sleeper ID to the dated player-directory snapshot; those labels may differ from the historical game. Missing fields display as — and are not treated as zero.
Team weekly stats
Historical Sleeper team records are joined by NFL team ID and week to the separately retrieved Sleeper schedule. These are team-level totals, not player metrics or league-specific projections. Schedule linkage does not infer why a team row is absent. The schedule endpoint is live-verified but not listed in Sleeper’s official API reference; its retrieval time and errors are shown.
Historical dashboard
Access the existing 2026 research modules, D/ST reliability analysis, scoring rules and player statistical tables.
League D/ST historyD/ST reliability labPlayer statisticsForecast quality gate
Do not use historical D/ST proxies as official Sleeper fantasy totals. Lineup optimization and win probabilities remain disabled until live player availability and independently validated forecasts are integrated.
Opponent defense · experimental holdout study
Position-level fantasy points allowed by each opponent, based on uploaded Weeks 1–6. Includes backups; not official D/ST scoring. This research does not automatically adjust current projections.
| Adjustment | MAE | Bias |
|---|
Caution: selecting a weight using the same Weeks 7–8 test set makes its reported error optimistic. No opponent adjustment has been approved for live lineup decisions.
Dynasty Trade Scenario Lab
Enter values for players and draft picks on each side. No market prices are invented. A positive net means your assumed incoming value exceeds your assumed outgoing value.
Use unique IDs, give/receive sides, player/pick types, and nonnegative numeric values. Picks require season and round.
League & Diagnostics
Live sync diagnostics
No request yet
App version
Development v65 · local review only. No Cloudflare upload required yet.
Offline access uses the most recently saved successful Sleeper response, labeled as cached rather than live.
Last live Sleeper validation
Timestamped audit of the most recent successful live API check. It is a snapshot, not proof of current live state.
Open verified Sleeper integration snapshotRefresh from the development project with npm run verify:sleeper; failed checks are logged separately and do not overwrite this success record.
Source inventory
This test distinguishes local files, live API access, and current roster availability. It does not validate player performance statistics.