The projections the sharpest DFS players were using. Now yours.
Built for a select group of elite professional DFS players and now open to everyone on Pro. Every player on every DraftKings and FanDuel NFL slate: points, floor, ceiling, ownership, and the full 74-column stat line underneath — from a model that keeps the box score physically consistent and knows your QB and his WR1 rise together.
State of the art, in DFS terms.
No single-number guesses. Four things this engine does that the CSV you are paying for right now does not.
Anchored to the market.
The closing spread and total set each team's implied points and play count — 64 plays at the base, never outside 58 to 72. De-vigged player props (passing, rushing, and receiving yards, receptions, anytime TD) anchor the stars. The model never argues with a number the book is willing to take money on.
The box score adds up.
Receiving yards sum to the QB's passing yards. Receiving TDs sum to passing TDs. Carries sum to team rush attempts, capped at 0.42 per snap, targets at 0.30. No WR3 catching 900 yards from a 250-yard QB — so your optimizer never builds a lineup that cannot happen.
Stacks are measured, not assumed.
The simulation resamples whole real games, so correlation is inherited from what happened on the field. A QB and his WR1 move together (+0.37). A shootout lifts both QBs (+0.20). Your DST fights the opposing QB (−0.29). Measured on 2,269 game-sides, 2021 through 2024.
Ownership you can ride or fade.
Projected ownership with a low/high band, volatility, and a chalk-fragility score. On 16 held-out 2025 DraftKings contests the call landed within about two ownership points of the real field — half the miss of a salary-based guess.
Every number you would want on a player card.
Not one projected score. The full projected stat line, the 27 scoring components that add up to it, and the distribution around it — for DraftKings and FanDuel scoring, on every slate.
- ✓Projected points, floor (p15), ceiling (p85), and nine percentiles from p05 to p95
- ✓Snaps, snap share, carries, targets, catch rate, and each player's share of the team's volume
- ✓Passing, rushing, and receiving yards and touchdowns, plus interceptions and fumbles
- ✓27 scoring components — every yard, touchdown, reception, and bonus scored separately
- ✓Projected ownership, low/high band, volatility, and chalk fragility
- ✓Value (points per $1k of salary) next to the slate salary
- ✓The player's own props attached — anytime TD, yardage, receptions
- ✓Kicker distance buckets and DST points-allowed brackets
- ✓A team layer: all 32 teams each week with implied total, spread, projected snaps, pass/run split, and the passing and rushing totals every player draws from
Five stages, no shortcuts.
Market anchor
Closing spread and total, de-vigged player props, implied team points, and a play count bounded between 58 and 72. The market sets the ceiling on volume before any player gets a share of it.
Opportunity model
Gradient-boosted, trained walk-forward season by season on play-by-play, snap participation, Next Gen Stats (separation, aDOT, CPOE, rushing yards over expected), depth charts, practice status (DNP/LP/FP), and weather decomposed against the stadium's orientation — domes read 0 mph. Snap-share error runs about a third lower than a recent-usage average.
Conservation
Positional touch budgets, touchdown conservation, and per-snap touch caps. Every team's projected box score is one that could actually be played.
Simulation
2,000 draws per player per week, resampling whole real games from a 1,137-game library. Out come the percentiles, floor, ceiling, standard deviation — and the stack correlations above, inherited rather than tuned.
Ownership
A boosted model blended with a crowd of optimizer-built lineups, normalized so each position's ownership adds up the way a real contest field does.
Two calls to the slate.
Plain REST over HTTPS, JSON out. player_stats is the 74-column line per player; projections is the same array under its older name. The team call needs season and week.
API=https://api.stat-api.com/api/v1
AUTH="Authorization: Bearer YOUR_PRO_API_KEY"
# 1. one slate: every player — points, floor, ceiling, ownership
curl "$API/dfs/slates/244335/projections" -H "$AUTH"
# → { "slate": { "id": 244335, "operator": "DraftKings" },
# "source": { "updated_at": "..." },
# "player_stats": [ {
# "name": "...", "position": "RB", "salary": 9200,
# "fantasy_pts": 24.86, "floor": 17.1, "ceiling": 33.4,
# "projected_ownership": 28.4, "value": 2.7,
# "quantiles": { "p05": ..., "p50": ..., "p95": ... },
# ... 74 columns
# } ],
# "computed": { "projected_players": 658, ... } }
# 2. all 32 teams for the week — implied total, snaps, pass/run
curl "$API/nfl/team-projections?season=2026&week=1" -H "$AUTH"
Projection questions, answered.
Who was using this before?
A select group of elite professional DFS players. The engine was built for them and run for them. It is now available to everyone on the Pro plan, with nothing held back.
Which operators and sports are covered?
DraftKings and FanDuel, with each operator's own salaries, positions, roster slots, and scoring rules. NFL today — every slate on the schedule, classic and showdown.
How do you know the numbers are right?
Three receipts. The scoring engine reproduces DraftKings' own settled points on 99.78% of graded player-games across 61 settled contests. Ownership landed within about two points of the real field on 16 held-out 2025 DraftKings contests it had never seen. Snap-share error runs about a third lower than a recent-usage average. Every backtest is point-in-time — week t uses only what was known before week t kicked off — with one exception we state out loud: weather is what was observed at kickoff, not a forecast, so backtests are slightly optimistic on weather.
What do floor and ceiling mean here?
Floor is the 15th percentile and ceiling the 85th of 2,000 simulated games for that player that week. The nine percentiles from p05 to p95 and the standard deviation come along, so you can size a cash play off the floor and a GPP play off the tail.
How current are they?
Projections are published per slate ahead of lock for the week's slates, and re-published when lines, injuries, practice reports, and depth charts change. Each response carries the timestamp it was built.
What plan do I need?
Pro — $3.20 per day, billed monthly at $99/mo (was $199/mo), 50% off for a limited time. Subscribers grandfathered with the DFS add-on keep full access. Commercial use is included on every paid plan.
Get the projections the pros were using.
Included in Pro at $3.20 per day — billed monthly at $99/mo (was $199/mo), 50% off for a limited time — alongside 20M records a month and the full sports API.