Methodology

How it works

Most dynasty rankings are vibes with a spreadsheet. Ours are a model with the receipts. Every number on this site comes from the same pipeline, and this page explains exactly what it does, including where it's weak. If a rankings site won't show you its math, that's a reason, not a mystery.

How the rankings are built

We tried to beat the market with a pure model. On quarterbacks, we did. On everyone else, the market beat us, so we stopped pretending otherwise.

Non-quarterbacks start from market ADP. We use preseason dynasty ADP (DraftSharks) as the baseline for RBs, WRs, and TEs. This isn't surrender; it's measurement. Across 2018–2023, when our discounted-career model disagreed with ADP on a non-QB, ADP was right more often at every age band: 53% vs 35% for ages 25–30, 63% vs 25% for ages 21–24. The market prices youth curves, breakout odds, and situation better than our equations do. So we let it.

Quarterbacks get our corrections. Where our model disagrees with QB ADP by 12 or more spots, we trust the model, it won 63.2% of those disagreements from 2018–2023. Quarterback value is mostly job security and career length, which is exactly what our survival-based model measures well.

Bridge quarterbacks get a situation haircut, not an age penalty. A 33-year-old fringe starter isn't automatically worthless, but a 33-year-old fringe starter on a team projected to pick in the top 10 that already rosters a 23-year-old quarterback is. We price the probability he's starting next year from the situation (draft position, young QB on roster, contract structure). In October 2026 that math put Jacoby Brissett at roughly a 15% chance of starting in 2027.

ADP fades; production takes over. ADP is the preseason starting point, not a permanent anchor. Every week, actual year-to-date production flows through the same career-DCF engine described below, and the preseason baseline matters less with each game played. The quarterback job-security corrections stay active all season.

The track record. 112% better than the market across 111 model-vs-market QB disagreements (2018–2023), wins and losses published on the track record page. Our earlier 54.7% figure described the pure career-DCF model; that model is no longer the published board, so the 54.7% is retracted as a standalone claim.

The engine inside the hybrid

The quarterback corrections and in-season updates below run on our career-DCF model:

1. Project each player's career

For every relevant player we project fantasy points per game for each season of their projected remaining career:

• Base: weighted average of the last three seasons (50 / 30 / 20), regressed toward replacement level when the sample is small. We don't trust eight games the way we trust thirty.
• Age curve: every position has its own peak and decline, fit from a decade of history. Running backs fall off a cliff after 26; quarterbacks cruise into their early thirties. Elite producers get a softer decline. Superstars really do age better.
• Career length: how many seasons are left? We estimate it from historical survival data: what fraction of players at each age and position were still playing one, two, five years later. High producers get longer projected careers, because historically they stick around. A 38-year-old quarterback gets two or three seasons; a 25-year-old gets the better part of a decade.
• Rookies: no NFL data yet, so they get a draft-capital prior: the historical hit rate of their draft slot, by position. A top-5 pick has earned more optimism than a sixth-rounder.
• In-season: once games are played, current-season production blends in with weight √(games/17). Week 2 barely moves the needle; Week 12 dominates. Players on season-ending IR score zero this year. The model doesn't pretend otherwise.

2. Convert points to championships

This is the part that makes us different. We don't value points. We value championship equity: how much does this player raise your odds of winning a title over their projected career?

We ran 60,000 simulated seasons across roster archetypes and measured it directly. The finding: title value is convex in scoring. A 24-point-per-game stud isn't worth two 12-point players. He's worth about four and a half of them. Mild concentration wins championships; perfect balance is mediocre; extreme stars-and-scrubs loses. Each future season's equity is discounted 12% per year, a rate fit from historical data. The fitted curve:

The exponent 2.16 isn't a guess. It's fit from the simulation (R² = 0.993). And because values already carry the stud premium, our trade math stays simple: just add up each side. No black-box "trade formula" doing secret adjustments behind your back.

3. Subtract replacement

A value only means something against what you'd get for free. Every player's final number is their championship equity over replacement: the freely available player at their position in a 12-team league (roughly QB18, RB45, WR55, TE20 in 1QB; QB30 in superflex, because quarterbacks are actually scarce there). A waiver-wire streamer is worth zero. That's not an insult, it's arithmetic.

4. Normalize to 0–1,000

The top player in each scoring profile is set to 1,000 and everything else scales from there. Bigger number, better player. That's the whole scale.

Settings profiles

• 1QB: the base values. Quarterbacks are replaceable here and the math knows it.
• Superflex: same engine, but replacement quarterback is QB30. Scarcity is priced in structurally, not with a fudge factor.
• TE Premium: tight end equity gets a 1.35× boost for full-PPR scoring (v1 approximation; we're calibrating this properly next).

Rookie picks

A pick's value is the expected championship equity of the rookie drafted there, computed over the full historical distribution of rookie outcomes at that draft tier, not just the average. The convex value curve rewards the right tail, which is exactly where pick value lives: most picks bust, a few become stars, and the math prices both. Future picks are time-discounted (a 2027 pick pays off a year out).

Update cadence

Full refresh weekly during the season (after Monday Night Football), with a light daily pass for injuries, depth-chart changes, and trades. Monthly in the offseason, plus event-driven updates for the draft and free agency.

Known limitations

We'd rather tell you than have you find out:

• Young quarterback breakouts are our #1 miss category. The model can't see a second-year leap coming. Nobody's can, but we're working on draft-capital-adjusted growth rates.
• Team context (offensive quality, quarterback play, coaching) isn't modeled yet.
• TE Premium and Superflex adjustments are first approximations, pending calibration.
• Forward injury risk isn't predicted: we discount after injuries, we don't forecast them.

Backtested, not vibes

The model is validated walk-forward across six seasons: tuned parameters beat the default spec by 17%, hold up on seasons the tuning never saw, and beat "just use last year" by 38%. The full report, including our biggest misses and what caused them, is how we keep ourselves honest.

Changelog

Frequently asked questions