Starting lineup fills automatically as you log your picks. Dashed gold = still unfilled.
Ranked by value over replacement for your lineup, adjusted for how likely each player is to survive to your next pick, positional runs, roster needs, and market mispricing.
Enter every pick as it happens — yours and your opponents'.
| ★ | Player | Pos | VBD | ADP | Status |
|---|
Best available now versus what you can realistically expect at this position when you're next on the clock. Big number = take it now; small number = safe to wait.
Backups to running backs you already own, and quarterback/pass-catcher pairings from your roster.
Players you've starred. Shows survival odds to your next pick so you know who to grab now versus who can wait.
Click × to remove a mis-logged pick from anywhere in the order.
Every team's picks so far. The two or three teams drafting between you and your next pick are the ones whose needs actually decide whether your target survives.
Tiers are cut where value drops off a cliff, so you can see when it's safe to wait and when it isn't. Edge compares expert rank to market ADP — green means the market is letting him fall, red means you'd be paying up.
| ★ | # | Player | Pos | Tm | Bye | Floor | Proj | Ceil | VBD | Opp/G | Role Δ | ADP | Edge | Split | Tier |
|---|
Runs thousands of full snake drafts from your slot. Opponents draft to market ADP with realistic noise; you draft by each strategy's rules. Because value (expert projections) and price (ADP) are separate inputs, strategies can actually beat the market here rather than just tracking it. Rosters respect real position limits, so no team ends up with six quarterbacks.
Everything downstream — replacement levels, VBD, roster slots, recommendations — recalculates instantly when you change these.
Pulled from Sleeper's public API — free, no account, built for outside apps, so it stands a real chance of working directly in your browser (unlike the ADP scrape, which sites block). Injury designations and depth-chart positions refresh automatically; your manual flags always override the feed.
These control how much each factor sways the recommended picks. Be clear on what they are: hand-chosen starting values, not coefficients fitted to historical draft results. Nothing here has been backtested against real outcomes — if your instincts differ from mine, your instincts are as valid as these defaults. Adjust freely.
Scraped once and baked into this file. Nothing re-downloads when you open the app, so it works offline and never changes under you mid-draft. When it goes stale, ask for a fresh build.
Projections blend two independent estimates. The first comes from last season's actual usage — targets and carries per game from Pro Football Reference's 2025 data — converted to points using league-average efficiency at that position, regressed toward the positional mean (harder for small samples), then adjusted by an age curve and expected games played. Volume is used rather than points because opportunity is far stickier year over year than efficiency. The second is the expert consensus curve, which captures everything last season can't: team changes, rookies, depth-chart moves, coaching changes.
Crucially, last season's usage is re-based onto this season's roster before it's used. Raw 2025 volume is misleading the moment a teammate leaves or arrives — Jahmyr Gibbs split carries with David Montgomery last year, and Montgomery is now in Houston. So for every team the app pools last season's carry and target volume, works out who is actually competing for it in 2026, and reallocates: departures free volume up, arrivals dilute it. Claims are weighted by both prior role and the market's expectation of this year's role, converted to implied opportunity so a receiver's higher points-per-touch doesn't let him outbid a running back for carries. That shows as ROLE ↑ / ROLE ↓ and the Role Δ column. Detroit is instructive: Montgomery's departure helps Gibbs, but the Lions signed Isiah Pacheco, so the gain is small rather than the windfall it looks like at first glance.
The two are weighted by how much the usage signal deserves trust — a full 17-game season of heavy volume earns the most, a 4-game sample earns little, and a rookie with no NFL history gets none at all and falls back entirely on expert consensus. Where the two disagree sharply, that shows as a USAGE ↑/↓ flag; it's a genuinely independent third opinion, since it comes from neither ADP nor the expert boards.
Floor / Proj / Ceil replace the old single number. Width comes from measurable uncertainty: games missed, small samples, age, how far the two projection methods disagree, and how split the expert sources are. The recommendation engine uses this positionally — in rounds 1–2 players are valued nearer their floor, because an early bust sinks a season; from round 11 on they're valued nearer their ceiling, because the downside there is already replacement level.
Still honest about the limits: this is a transparent statistical model, not a trained one. Nothing here was fitted to historical outcomes or backtested — the age curves, regression strengths and blend weights are reasoned choices. The roster-context model also only knows about players it has 2025 data for: a rookie or a free agent outside the top 225 who wins a job will not appear in the competition pool, so treat unsettled camp battles as something to override manually. It's a real improvement on the pure rank curve it replaces, because actual measured usage now enters the calculation, but it isn't a machine-learned forecast.