Methodology & important information

Fantasy tools

Fantasy Overdrive provides NFL rankings, player comparisons and lineup-building tools. Projections are estimates conditional on playing. We do not accept wagers, hold entry fees, enter contests or guarantee results. We are independent of the NFL, its teams, DraftKings, FanDuel and our data providers. Platform names identify supported scoring and file formats; they do not imply endorsement.

How projections work

The baseline is a weighted average of up to five recorded appearances before the selected week. Each older appearance receives 75% of the next newer appearance’s weight. Missed games are not zeroes. For QB/RB/WR/TE players with at least eight prior-season appearances with their current team, we blend that recent estimate with their prior-season average. The recent weight is n/(n+2), where n is the number of recent appearances (up to five). After four appearances this gives 67% current-season weight. For RB/WR/TE with at least two appearances, a change of at least 50% and three weighted carries/targets versus the prior season raises current-season weight to at least 80%. Component averages use the same blend before market adjustments. RB/WR/TE touchdown components use projected opportunities times current-season scoring rates smoothed toward position rates from earlier data (100 carries or 50 targets of smoothing). This is not a goal-line opportunity model. Without qualifying history, the recent estimate remains in use. Small samples, rookies and role changes can make estimates less reliable. Historical low/high scores are observed results, not forecast floors and ceilings.

Available odds and player props are processed on the server as statistical inputs to fantasy estimates. Raw prices, lines and bookmaker quotes are not published. For covered statistical components, the model blends 65% of a price-adjusted estimate with 35% baseline production. It uses paired over/under prices, removes their combined margin and uses a normal approximation for yards and a Poisson count approximation for attempts, catches and other counts. Only uniquely matched upcoming games and sufficiently fresh inputs qualify. Missing or stale inputs fall back to production history or a supported role estimate. Prop coverage is partial: collection considers games up to seven days ahead and rotates small batches across games within the existing allowance. Not every player or component will have a fresh prop. Depth sources are checked every six hours; their original chart timestamps still determine whether evidence is usable.

Paired passing, rushing and receiving touchdown props inform their respective scoring components, using four points per passing touchdown and six per scored touchdown. We remove the paired bookmaker margin and infer an expected count with a Poisson assumption before blending 65% market estimate and 35% statistical baseline. For scored touchdowns, paired anytime Yes/No prices are a fallback proxy when rushing-plus-receiving totals are unavailable; those markets may include return scores and do not measure passing touchdowns. One-sided, stale, mismatched or ambiguous quotes are excluded. This count conversion and blend are not calibrated accuracy claims. Team scoring expectations adjust only touchdown components without usable player TD props, by at most 15%, avoiding a second adjustment to components already covered by props. Kickers use paired kicking-points props when available; otherwise field-goal and extra-point props can adjust historical production, retaining historical field-goal distance mix. With no usable kicker props, the team scoring adjustment is capped at 15%. Defenses use sacks, turnovers, defensive/special-teams touchdowns and reconstructed points allowed; their scoring-environment adjustment is capped at ±3 points. Rare defensive recovery and points-allowed attribution can differ from official contest scoring. DraftKings yardage bonuses use historical bonus frequency. These assumptions are not calibrated forecasts.

Workload, matchups and availability

Evidence includes target share, offensive snap share, air-yard share, red-zone opportunities, team pace, depth charts and opponent performance. Targets are not routes; red-zone opportunities do not guarantee touchdowns. Missing measurements remain unavailable. Snap counts describe past appearances, and depth charts may lag role changes.

Attempts, completions and interceptions can inform workload and scoring. Direct yardage props take precedence over estimates inferred from workload; overlapping combined markets only fill missing components. Targets are inferred from projected catches and historical catch rate when possible, since no targets market is currently listed by this provider. Alternate paired lines are a fallback. Longest-play, first/last-scorer and individual-defender props remain research context rather than extra fantasy points. Changes in prices are tracked privately from our first observation; movement does not receive an additional projection boost.

The optional usage/pace adjustment is capped at ±10%, excludes baseline components already covered by props and is off by default. A chronological 2025 diagnostic over 5,402 scored appearances had PPR mean absolute error of 4.612 points with the layer and 4.605 without. The optional defensive matchup adjustment uses up to six earlier weeks, four games of league-average smoothing and a ±10% cap. Its 5,427-appearance diagnostic measured 4.598 versus 4.595 points. Lower is better: neither demonstrated improvement. These are heuristics, not trained XGBoost or LightGBM models.

The rankings directory keeps known players visible, including injured players, backups and players without a usable projection. Recommendation ranks apply only to supported upcoming estimates. Other rows show availability, baseline-only or projection-pending labels; this does not make those players eligible for automatic lineups. Injury reports are not game-specific confirmation of active status. An absent report means unverified, not healthy. Out, inactive, doubtful and reserve reports are conservatively excluded, including older reports. For a confirmed new running-back absence, fresh statistics and an unambiguous depth chart can trigger a role-first replacement estimate: 65% of vacated carries to the next available back, the remainder to other backs, and half of vacated targets shared by role. Team and league rates estimate efficiency and touchdowns, with RB touchdown totals capped at the previous baseline. These affected backs start with the role estimate. Qualifying paired props then blend 65% price-adjusted estimate with 35% role baseline for covered components. Quotes must be no older than six hours and timestamped at or after both the depth-chart evidence and the confirmed absence report. This timing safeguard does not prove the market has incorporated the news. Missing, older or one-sided props leave that component on the role estimate. Overlapping props are not added twice; inferred targets use the role catch rate. Separate usage and team-total adjustments are skipped for these backs. Continuing absences use observed workloads since the teammate last played, for absences up to four weeks old, without adding vacated touches a second time. Doubtful teammates permit an explicitly conditional scenario; these estimates are labeled and excluded from automatic lineups until the absence is confirmed. Missing reports do not establish health, and uncertain or stale role situations remain visible under review. Goal-line-specific rates and limited participation are not modeled. Explicit historical mode cannot recreate pregame injuries or market snapshots and may include players who did not play.

Simulations and lineup search

The builder evaluates sampled candidates against 3,000 shared hypothetical outcomes with a fixed seed. Offense uses positively skewed fantasy-point distributions; defense can score below zero. Assumed shared team, game and passing factors create correlations, including opposing offense/defense effects. These are not play-by-play simulations; passing and receiving statistics are not reconciled.

Cash favors a modeled 25th-percentile lineup score and GPP the 90th percentile. These preferences do not estimate cash rates, ownership, contest winnings, profitability or win probability. Search is approximate and does not guarantee a global optimum. Player exposure counts all appearances, including Captain/MVP, with fractional limits rounded down. Missing projections require a manual estimate or exclusion.

The dart-throw screen looks for at least three appearances, an observed high of at least 1.6 times the estimate and a low no greater than half the estimate. It describes historical volatility, not a calibrated ceiling. Model diagnostics use corrected historical data and are not prospective backtests. Weather and AI explanations are not connected. Running-back role redistribution is an uncalibrated heuristic; a 13-player-game conditional diagnostic had PPR MAE 4.479 versus 4.271 for the unchanged baseline, and RMSE 5.393 versus 5.835. Historical depth timestamps were unavailable, so these results are not a validated pregame backtest. Separate formula-based ownership estimates are described below.

Salary files and exports

Supported standard full-game Classic contests use nine players, including one defense; kickers are excluded. Showdown uses one Captain/MVP and five FLEX players and permits kickers and team defenses. Platform caps, multipliers and team constraints are applied. Special, partial-game and in-game formats are unsupported. Confirm the exact contest rules and entry limits.

Automatic DraftKings salaries depend on supported slates being available from the provider. FanDuel currently uses CSV import. Provider IDs are never substituted for platform upload IDs. Export needs the original complete contest template and matching player reference table; automatic salaries additionally require exact game-date, salary and featured-role matching. Unsupported layouts are rejected.

CSV files are processed in your browser tab. Final injury and kickoff checks run before generation and export. Downloads expire after at most two minutes or at kickoff. New full-lineup generation and export are blocked after lock; late swap is not supported. Downloading does not reserve, enter or edit any paid contest. Actual platform acceptance must be checked with your contest. DraftKings bulk CSV upload is desktop-only.

Data sources and freshness

NFL statistics, play-by-play, depth charts and snap counts are sourced through nflverse, including PFR snap data. Used under CC BY 4.0; Fantasy Overdrive filters, summarizes and transforms the data. Source providers do not endorse these tools.

Injury, schedule and salary data: BALLDONTLIE. Internal market inputs: The Odds API. Data can be delayed, incomplete, revised or incorrect. The providers are not official league feeds. Their permissions do not clear every potential third-party right.

Odds, selected-game props and statistics use six-hour caches. A separate twice-daily refresh requests due feeds and one upcoming game’s props, when available. The collector discovers available player markets and rotates market batches within the existing allowance. This produces sparse coverage, not full-slate or real-time market tracking. Coverage of all player props is not guaranteed. Injury checks normally run every two hours, every five minutes Sundays 10 a.m.–9 p.m. Eastern, and from three hours before other game-day kickoffs through kickoff. Final lineup checks may refresh reports directly. Salary catalogs cache for 30 minutes and pools for 15 minutes. Usage data caches for 12 hours.

Freshness timestamps and actionable errors remain beside the tools. Always confirm availability, contest requirements and lock time on your platform before using a lineup.

Use and limitations

These tools are informational and do not guarantee performance, prizes or profit. Use only where you are eligible and where the contest is lawful; follow the platform’s age, location and entry rules. You remain responsible for your decisions and entries. For support with gambling-related harm in the United States, call 1-800-GAMBLER.

Account information

Sign in or create a free account with email; no password is needed. Public rankings and comparisons remain available without an account. A free account supports one Classic cash lineup per run and saved builders. Compare Free and Pro plans for paid features and current pricing. Account pages are not cached publicly.

Choosing Save builder & lineups stores a private slate, settings and lineup snapshot in your account. The sign-in handoff preserves your current builder so you can resume after verification. You can export saved account data and request account closure through Account. Saved snapshots do not establish current player availability or contest eligibility.

Subscriptions and paid account management are available through Account > Billing. Read the Terms, Privacy Policy and Refund Policy for purchase, data and account-management details.