Historical NFL Data
The project uses nflverse data through the local Python pipeline. Player-game, roster, schedule, injury, depth-chart, and context tables support position-specific modeling and reconciliation.
Methodology
Better Projections is designed around fantasy draft decisions: transparent 2026 player-stat projections, conventional fantasy scoring, VBD, tiers, and disciplined uncertainty ranges.
The project uses nflverse data through the local Python pipeline. Player-game, roster, schedule, injury, depth-chart, and context tables support position-specific modeling and reconciliation.
Models are evaluated with time-aware, out-of-time validation so later-season information is not allowed to leak into earlier predictions. MAE means mean absolute error: the average absolute miss in the same unit as the projected stat. Future game-prediction features must only use information available before the game being predicted.
QB, RB, WR, and TE projections are handled separately because opportunity and scoring are position-dependent. The project has tested model families including LightGBM, CatBoost, XGBoost, Ridge, and related baselines where appropriate.
The projection system applies team/context checks and cross-position constraints before publishing website-ready values. Expected stat projections are preserved from the frozen V3 projection engine.
Floor and ceiling values are realistic downside and upside scenario ranges from the V3 interval layer. They are not exact probabilities and are not simulated rank distributions.
Standard, Half-PPR, and PPR use conventional fantasy scoring. VBD compares each player to a replacement-level player at the same position, turning raw fantasy points into positional draft value.
Website V2 calculates replacement levels from your league size and starting roster. Mandatory QB, RB, WR, and TE slots are filled first. FLEX then allocates the best remaining RB/WR/TE players. Superflex allocates the best remaining QB/RB/WR/TE players, which naturally raises quarterback value.
Larger leagues and deeper starting requirements push replacement ranks farther down the player pool. That changes scarcity and personalized overall rankings without changing the underlying player-stat projection.
Bench spots are saved in league settings. VBD replacement remains based on starter, FLEX, and Superflex demand, while auction values use bench spots to add the highest-valued remaining modeled players to the auctionable pool.
Auction values cover the modeled QB/RB/WR/TE pool. The site reserves a minimum bid for each draftable player, then allocates remaining league dollars by positive dynamic VBD and reconciles rounded dollar values back to the modeled auction budget.
Better Projections is an independent fantasy football analytics site and is not affiliated with, endorsed by, or sponsored by the National Football League.