Win the Numbers
Dynasty player values, a trade calculator, and trade ideas tuned to your team's direction. Import your Sleeper league and make every move by the numbers.
Everything a dynasty manager needs, in one place.
Dynasty values
Every player and pick valued from historical production, projections, advanced stats, age, draft capital, and your league's roster and scoring settings.
Trade calculator
Pick any team, drop in players and picks on both sides, and see how the values stack up instantly.
Trade suggestions
Ideas tailored to your team's direction — contenders get win-now moves, rebuilders get deals that pay off later.
Team analysis
See your roster's needs, surplus, overall rating, and injury status at a glance.
Player research
A fast, filterable, sortable list of every player, plus deep player detail pages: value, bio, historical and advanced stats, and projections.
Scouting
College prospects with draft-eligible year, school, status, stats, and value, all filterable and sortable.
Start/sit optimizer
Your matchup simulated 20,000 times. See your win probability, the lineup that maximizes it, and which bench swap moves the number most.
Waiver wire
Every available free agent ranked against your actual roster holes — who to add, who to drop, and whether it's a value pickup or a points upgrade.
Not a gut-feel ranking. A formula you can inspect.
Every value is computed from the same set of inputs, weighted by position.
- Recency-weighted history
- Up to three seasons of production, weighted 5/4/3 toward the most recent, with a minimum games-played bar before a season counts.
- Position-specific age curves
- Running backs, receivers, quarterbacks, and tight ends age differently — each position gets its own prime window and decay rate, not one curve applied to everyone.
- Draft capital
- For players early in their career, where the NFL actually drafted them still carries signal — folded in alongside production, not instead of it.
- Advanced stats
- Per-play efficiency — EPA, target share, air yards, yards after catch — composited by position, not just counting stats.
- Your league's settings
- Positional scarcity and value over replacement are computed from your league's actual roster requirements and scoring, not a generic default.
- Simulated matchups
- Start/sit win probability comes from 20,000 Monte Carlo simulations per matchup, not a single point estimate — so the lineup recommendation reflects a full range of outcomes, not just the median one.
The measurements we have today — and exactly what they prove.
College prospect grades — historical performance (p20)
A disk-only archive replay rebuilt the pre-draft board for four mature draft classes and followed the same 260 covered players into the NFL. Every model and baseline below is scored on that identical cohort; every consumed source response is timestamped and hashed.
| Position | Players | Pre-draft grade | Public consensus | College-only | Settled draft slot (diagnostic) |
|---|---|---|---|---|---|
| QB | 37 | 0.560 | 0.484 | 0.371 | 0.626 |
| RB | 70 | 0.509 | 0.584 | 0.453 | 0.619 |
| WR | 108 | 0.570 | 0.586 | 0.477 | 0.648 |
| TE | 45 | 0.257 | 0.428 | 0.212 | 0.414 |
| All | 260 | 0.492 | 0.513 | 0.416 | 0.600 |
The honest result: our pre-draft grade does not beat public consensus overall on this sample. The paired difference is -0.022 with a 95% interval of -0.115 to +0.070. It does beat the college-only grade by +0.076, with a 95% interval of +0.021 to +0.131. That isolates the value of the consensus blend without overstating a head-to-head win.
Source: docs/backend/prospect-grading-backtest.md, algorithm version p20.
The replay is drafted-only because the historical source attaches pre-draft consensus ranks
to eventual draft picks; it does not claim coverage of ranked-but-undrafted prospects.
Dynasty values — historical performance control (v15)
We reconstructed three preseason dynasty cohorts, ranked the same 1,375 players with our value, public expert consensus, and trailing production, then followed their best two of the next three NFL seasons. Historical projections were removed because no trustworthy point-in-time archive exists, so this measures v15's documented projection-ablated path.
| Position | Players | Projection-ablated value | Public consensus | Trailing production |
|---|---|---|---|---|
| QB | 164 | 0.4610 | 0.7505 | 0.6180 |
| RB | 430 | 0.6028 | 0.7163 | 0.3888 |
| WR | 548 | 0.6175 | 0.7467 | 0.4877 |
| TE | 233 | 0.5945 | 0.7014 | 0.5261 |
| All | 1,375 | 0.6006 | 0.7036 | 0.4926 |
The honest result: public consensus is better by 0.1026, with a 95% interval of 0.0697 to 0.1350. Our value beats trailing production by 0.1081, with a 95% interval of 0.0502 to 0.1654. Because projections carry 40% of the complete dynasty formula, this control does not establish the performance of the exact projection-inclusive values users receive.
Source: docs/backend/player-value-backtest.md, algorithm version v15.
The report labels this result projection-ablated and includes cohort-level join attrition,
paired uncertainty, and the exact reproduction command.
Redraft values — historical fallback control (v15)
Five preseason cohorts let us test the projection-ablated legacy composite retained as v15's fallback. The same 2,190 players are compared with public expert consensus and trailing production, then scored against the following regular season. This is a diagnostic control, not the league-specific projection/VOR model users receive.
| Position | Players | Projection-ablated fallback | Public consensus | Trailing production |
|---|---|---|---|---|
| QB | 256 | 0.4052 | 0.7589 | 0.5604 |
| RB | 691 | 0.5601 | 0.6906 | 0.4709 |
| WR | 853 | 0.5129 | 0.7080 | 0.5122 |
| TE | 390 | 0.5263 | 0.6880 | 0.5486 |
| All | 2,190 | 0.5213 | 0.6757 | 0.5164 |
The fallback trails consensus by 0.1545, with a 95% interval of 0.1244 to 0.1865. Its +0.0047 difference from trailing production has a 95% interval of -0.0417 to +0.0522. The exact v15 projection/VOR path still requires complete, point-in-time projections and a frozen viable roster shape before it can make a historical performance claim.
Source: docs/backend/player-value-backtest-redraft.md, algorithm version
v15. The report explicitly marks this as a projection-ablated legacy fallback/control,
not evidence for the shipped offensive projection/VOR model.
What we don't publish a number for.
- Start/sit win probability — the model is documented in full, but no calibration figure exists yet; nothing checks that a stated 65% wins ~65% of the time.
- Waiver suggestions — every threshold is hand-picked and uncalibrated.
- Trade fairness and trade suggestions — no outcome label exists for "was that trade fair."
- Team analysis and overall rating — same.
- IDP values — the available public consensus baseline is offense-only, so the IDP half of the value engine has no equivalent comparison.
- Exact projection-inclusive dynasty superiority and shipped redraft projection/VOR performance — the historical controls above omit inputs that were not archived point in time.
No purchase required.
A few advanced features are unlocked by watching an optional rewarded video — no purchase required, and rewards never depend on sharing tracking data. Sign in to sync your watchlist, settings, and leagues across devices, or continue as a guest.
- NFL statistics provided by the nflverse project (CC-BY 4.0).
- College football data provided by CollegeFootballData.com.
- League and roster data accessed via the Sleeper API.
Not affiliated with or endorsed by Sleeper, the NFL, or any college or professional league. For informational and entertainment purposes only.