AI DFS PROJECTIONS WITH A CEILING · WHO GOES NUCLEAR THIS WEEK?
DK classic, full PPR, bonuses in. Tap a column to sort, a name for his expected line, the props that moved him and his injury call. 3X/4X = the chance he scores 3 or 4 times his salary in thousands. Kickers are showdown-only.
| PLAYER | SAL | MED | CEILTOP 20% | NUKETOP 5% | 3X | 4X | CEIL/$K |
|---|---|---|---|---|---|---|---|
| Loading the board… | |||||||
Every game on the board, with who is assumed playing and who is out (tap a name for the call and its source), the QB pairs and 3-man stacks that boom together most, and two things you can only get from a simulation: the ceiling world (tap a name and the whole game is re-priced from the simulations where he had his top-20% game; type a number to ask “if he scores 35+”) and the Q toggle (flip a questionable call and his teammates move). Tick players to price a stack of your own.
The Ceiling Board at twogungpt.com/projections/ lists every DraftKings main-slate player by his CEILING, the top-20% game his simulation gives him, a week that happens about one time in five. Sort by NUKE for the top-5% game, about once a season. It is free: the top three players show without an account, and the full board needs a free TwoGunGPT account (Google sign-in or an emailed magic link, no password).
Not the median times a number. TwoGunGPT simulates each game 10,000 times at the box-score level, both teams together, off the betting line, and splits every simulated box score among the players by their usage. A player’s CEILING is the 80th percentile of his 10,000 DraftKings scores: a top-20% game, about 1 in 5. His NUKE is the 95th percentile: a top-5% game, about 1 in 20, roughly once a season. Both frequencies are printed next to the numbers so nobody mistakes a ceiling for a prediction.
Public play-by-play and box scores, the injury report, and the betting line, plus the sportsbooks’ player props where they exist: each prop line is devigged and sets that player’s center without breaking his correlations. No vendor’s projections appear on the page or feed it. Every number is TwoGunGPT’s own, and the page says which props moved each player. TwoGun doesn’t monitor these numbers. For projections a human actually oversees, we recommend StatRankings.com.
It was tuned on 2022 to 2024 box scores and checked once on a held-out 2025 season. On that season 19.2% of player games beat their CEILING line and 4.8% beat their NUKE line, against targets of 20% and 5%. At the CEILING level its edge over a plain median board was about 1% for running backs and receivers, which is not decisive. Live weeks are being scored now; until a few are in, the page claims nothing about beating anyone.
Every player is either PLAYING or OUT; nobody’s number is shaded for the chance he sits. A questionable player gets a call from the injury news, printed on his game card with the reason and the source. An OUT player’s targets and carries go to the teammates who play. The Q toggle on each game card flips a call so you can see how his teammates move. Calls come from the Friday and Sunday-morning news; a late scratch after the last update is not reflected until the next one.
Twice a week in season: Thursday before the Thursday game, and Sunday morning before the early games. The stamp at the top of the page says exactly when it was last updated.
Yes, since Week 3 of 2026. The chat’s lineup desk builds on the Ceiling Board’s numbers whenever you have not loaded projections of your own: a main-slate build uses each player’s CEILING and his 3x odds (give it one name in or out and it builds the rest, a different random world every time), a showdown build follows your game story with the board’s numbers as the guide. Your own projections file always wins. The chat also answers any player’s median, ceiling, nuke, 3x odds, injury call or stack odds straight off the board.