How to build projections

A top-down workflow: set what the team should produce, allocate that volume to players, then reconcile so nothing is vibes-only.

Open workbook Browse reference data

Quick start

Create a workbook Click New workbook — no signup required, it's private to your browser. Keep Pre-fill from model checked — every team’s offense and every player gets a model baseline in one click.

Edit where you disagree Pick a team + position and adjust cells. The vs model column and cell shading show where your view diverges from the baseline.

Reconcile and review The team strip shows which teams still need work; Reconcile all scales every room to its team totals in one click. Done? Export CSV to take your projections anywhere.

That’s the whole loop. Seeding alternatives (market, prior season), passing-rate helpers, and distribution tools live under Advanced in the team offense block — useful, but never required.

1 · Workbook setup

Each workbook is your private projection sheet. Rosters are seeded from the current season; prior-season stats appear in the Prior season block as read-only reference.

New workbook — seeds projection rows for every rostered player; with Pre-fill from model checked it also applies the full model baseline to all 32 teams.
Reset roster rows — re-seeds players without wiping team-level offense inputs.
Team / Position filters — work one position room at a time (QB, RB, WR, TE).

Tip: work team-by-team. Finish offense reconciliation for a team before jumping to the next.

2 · Team offense (top-down)

Start here for each team. The Team offense block is where you decide how many plays, pass attempts, rush attempts, yards, and TDs the offense produces in total. If you created your workbook with the model baseline, these totals are already set — the tools below (under the Advanced toggle) are for re-seeding or fine-tuning.

Seed your starting point

Copy prior season — copies last year’s team totals into your projection cells. Good baseline when personnel is stable.
Seed from market — uses pre-season win totals to set expected offensive volume and pass-rate tilt. Coefficients are fit from historical data (not hand-waved).
Seed from model — loads walk-forward team KPI predictions (plays/game, pass rate, etc.) and converts them to season totals.
Apply market scale — multiplies your current team totals by the market-derived volume factor without replacing your pass/run mix entirely.

Passing rates

Enter Cmp% and Y/A, then click Apply rates. Completions and pass yards are derived from your attempt total — you don’t need to guess all three independently.

Distribute to players

Reconcile room — one click scales mismatched player rooms to your team offense targets (passing, rushing, receiving, fumbles), preserving each player’s share within the group.
Distribute to QB room — splits team passing stats across QBs weighted by model pass attempts (falls back to workbook cells, then prior season).
Distribute rushes — splits carries and rush yards/TDs across QBs and RBs by model carry share.
Scale skill targets — splits receiving volume across RB/WR/TE by model target share.

The reconciliation table shows Team (your target), Room sum (what players currently add up to), and a Status flag when they diverge.

3 · Player projections

The Projection block is where you edit individual player stat lines. Model values appear as placeholders in empty cells and in the Model column of room totals.

Fill from model — fills only empty cells from the stat model (safe default).
Overwrite with model — replaces every cell with model output. Use when you want a clean model prior before making selective overrides.

Overrides should be intentional: check the vs model column in room totals to see where your edits diverge from the model benchmark.

4 · Benchmarks and league view

Room totals compares your projected room to four references:

Model — sum of stat-model season lines for players in the room
Prior room — what this position group actually produced last year
Projected — your current workbook totals
League median — typical room production across the league

The League board at the bottom ranks all teams at the selected position by estimated PPR — useful for sanity-checking your ranks after you finish a position.

Principles

Top-down before bottom-up. Team volume constraints come first; player splits second.
Models are priors, not gospel. Seed from model/market, then edit with reasons.
Reconcile constantly. If room sums don’t match team offense, something is inconsistent.
Use multiple benchmarks. Model, prior year, and league median catch different kinds of mistakes.

Reference tables (rosters, prior player stats, team offense) live on the Reference page. Model methodology and evaluation metrics are on Models.