Data sources
What this analysis is built from, what is missing, and which parts are measured rather than inferred.
Full analysis
Statistics, the forecast and the schedule are available, and every player that matters is matched to them.
Inferred: GM Mode decides the mode from the coverage figures below. Full needs your whole roster and every starter matched, 90% of the league, and 80% of the waiver pool.
Where each input comes from
| Status | Input | Source and state | Basis |
|---|
| ✓Available | League data | Demo fixture: invented league, rosters and players. | Measured |
| ~Partly available | Weekly forecast | GM Mode's stat-based model. None needed: every weekly value comes from the model. | Inferred |
| ✓Available | Scoring rules | Demo league rules. | Measured |
| ✓Available | Player statistics | Invented demo statistics for weeks 1, 2, 3. | Measured |
| ✓Available | Prior-season baselines | Invented prior-season output for the demo players. | Measured |
| ✓Available | NFL schedule | An invented demo schedule with bye weeks. | Measured |
| ✓Available | Identity matching | Demo players carry their own statistics, so every player is matched. | Measured |
| ✓Available | Your current lineup | Set in the demo league, so current-versus-recommended swaps are shown. | Measured |
| ✓Available | Free-agent pool | Demo free agents. | Measured |
| ✓Available | This week's matchup | Your demo opponent this week. | Measured |
| ✓Available | Matchup adjustments | Fantasy points allowed by position, computed from completed weeks; applied only to GM Mode's own model. | Inferred |
| –Unavailable | Ask GM | No AI key configured. Preset questions are answered from the analysis; every recommendation is computed without AI. | Explains only |
| –Unavailable | Prospective ledger | The demo league is never recorded. | Status |
| ✓Available | Last refresh | Analysis computed Sep 22, 2026, 1:16 AM UTC. Model 0.5.0+w87a93b91. | Status |
- Demo League: fictional players and statistics.
- Trade packagesExperimental. Projected lineup changes only; realized benefit is not validated, and acceptance is never predicted.
- ForecastingGM Mode does not claim to out-forecast its projection source. Its intended edge is in decisions: league-specific value, roster construction and the discipline to do nothing.
| Format | 12-team · PPR · 1QB |
| Starting lineup | 9 starters · QB, RB, RB, WR, WR, TE, FLEX, K, DEF |
| Quarterbacks | 1QB |
| Receptions | 1 per catch (PPR) |
| TE premium | None |
| Passing | 4 per TD · -1 per INT · 1 pt per 25 yds |
| Rushing / receiving yards | 1 pt per 10 rush yds · 1 pt per 10 rec yds |
| Bonuses | None |
| Kicker | Flat per field goal |
| Team defense | Points/yards-allowed tiers plus events |
| Roster | 6 bench · 1 IR |
| Waivers | FAAB ($100) |
| Scoring support | Every configured rule is scored. |
How this format sets player value
Replacement level is calculated from this league, not a generic ranking. Starter demand counts every team's dedicated slots plus the flex slots each position actually fills. A player's trade value is measured against the replacement starter; add and drop decisions against the waiver wire.
How recommendations are made
Rankings, projections, lineup optimization, waiver ranking and trade matching are computed in code. A language model is used only to explain and to answer questions.
- 1ForecastWeekly points come from Sleeper's published player projections, re-scored under your league's rules, whatever platform your league is on. GM Mode does not claim to out-forecast them; historical testing found them more accurate than its own model, which fills gaps and drives rest-of-season value.
- 2League-specific scoringEvery stat line is scored with your league's own rules where they can be read, so a reception is worth what your league says it is.
- 3League-specific replacement valueReplacement level comes from this league's team count, starting slots, flex and superflex demand, and what is left on waivers.
- 4Lineup optimizationFilling your starting slots is solved as an assignment problem, respecting real position eligibility including flex and superflex.
- 5Action thresholdA claim must beat the cost of making it: the dropped player, the bid or priority, the roster spot and the upside surrendered. The bar is set per position. Low-confidence claims are shown to watch, not to make. Most weeks, most moves do not clear it.
- 6One consistent planClaims are applied in order and every later claim is re-scored against the roster the earlier ones leave, so no player is dropped twice, budgets are never overspent, and lineup advice reflects the claims above it.