The calculator shows its work.

The trade-target hub compares each player's current dynasty 1QB overall rank with the same player's redraft 1QB overall rank, isolating time horizon without mixing in Superflex quarterback scarcity. A separate dynasty Superflex board establishes the market price ceiling. The two direction screens exclude that leading tier and keep candidates inside a useful top-120 market window.

A larger builder edge means the long-horizon dynasty market ranks the player more highly. A larger contender edge means the current-season redraft market ranks the player more highly. The lists are reproducible format screens—not buy-low predictions, projections, or claims about future news. League scoring, lineup needs, health, schedule, and the price another manager will accept remain outside this screen.

Every head-to-head page loads both players from the same validated release and compares them across dynasty Superflex, dynasty 1QB, Superflex TE premium, and redraft 1QB. Quarterback pages then isolate four- versus six-point passing touchdowns; RB, WR, and TE pages isolate standard, half-PPR, and full-PPR scoring.

The five-percent label is a reading aid, not a new trade-value formula. It prevents small score differences from being described as categorical wins. Displayed values retain their underlying market and scoring math; no editorial sentence changes a score. If the higher-valued player changes across settings, the page says the leader flips and directs the reader to the row matching that league.

The comparison collection begins with deliberately curated player files in meaningful position and value neighborhoods, then expands in reviewed batches. It is not every possible name permutation. Injuries, forecasts, manager preferences, and required trade adds remain outside the comparison answer.

Team pages grade each scheduled game from 0 to 100. The largest input is the opponent's prior-season NFL scoring-defense rank: No. 1 allowed the fewest points and creates the cold end of the scale; No. 32 allowed the most and creates the hot end.

The score describes an overall scoring environment. It does not include positional fantasy points allowed, current injuries, depth-chart changes, weather, or player usage, so it is never presented as a start/sit projection.

The weekly lineup model starts with each player's recorded fantasy-point distribution. Up to 18 prior-season games provide the stable baseline; completed current-season games receive increasing weight as the sample grows.

Recent position-specific volume—attempts plus carries for quarterbacks, carries plus targets for running backs, and targets for receivers and tight ends—combines with offensive snap share. That role adjustment is capped at ±4%. The opponent's prior-season fantasy points allowed to the position adds a separate adjustment capped at ±6%.

Availability changes the estimate only when the injury-report week equals the target week. Older designations remain in the availability archive but never silently reduce a current projection. Once both players finish, the page displays their actual points and whether the original deterministic lean was correct.

The Week 2 holdout included 160 player observations. Half PPR mean absolute error was 5.60 points, compared with 7.29 for using the previous game and 5.64 for prior-season PPG alone. Because that edge over the stronger baseline is small, the initial public cohort remains limited to 20 reviewed decisions.

The NFL model is separate from the fantasy-value models. It creates one team rating from completed games before the target week, blending capped opponent-adjusted scoring margin with the prior season as an early-season anchor.

Current games gain weight gradually and no single result contributes more than a 28-point margin. The scoring projection regresses points scored and allowed toward the previous season, then clamps the game total to a plausible 30–65 range and the margin to ±21.

Moneylines, spreads, spread prices, totals and total prices come from the timestamped nflverse schedule release. They never create the team rating. FTT compares its independent margin and total with the market snapshot, calls differences under one point a pass, and freezes the complete record at kickoff.

The 2025 Weeks 5-18 walk-forward covered 208 games. It finished 125–83 straight up, 81–82 ATS and 81–80 on totals, with 7.63 points of per-team score error. ATS and totals were effectively coin flips.

Straight-up probability also powers the survivor and confidence-pool boards. Survivor ranks the available winners but cannot know which teams an entrant already used. Confidence pools assign the maximum points to the highest win probability and descend one point at a time; close probabilities should be treated as close decisions.

The public release is analysis, not proof of an edge. The NFL model does not manufacture parlays, player props, unit sizes, locks or guaranteed-return claims. Separate calculators perform transparent arithmetic only from numbers the user enters. Final scores grade archived pregame calls without rewriting them.

The odds, parlay, round-robin, no-vig, arbitrage, teaser, hedge and Kelly tools run locally from the values entered. They do not fetch a sportsbook account, transmit a wager, supply live odds or treat an FTT model output as the user's probability estimate.

American, decimal and fractional prices convert through one decimal-odds representation. Invalid prices, negative stakes and out-of-range probabilities are rejected. Kelly is clamped at zero when the supplied estimate has no positive expected return; half and quarter Kelly scale the full result. A hedge or dutching allocation can equalize a loss as well as a profit, so the net result remains visible.

Round-robin maximums assume every leg wins. Settlement grades every generated combination from the entered win, loss and push results; pushed legs contribute a 1.0 decimal multiplier. Arbitrage allocation can solve from either a fixed total stake or a desired equal return. Teaser payouts use the complete-card price entered by the user.