DELTA evaluates dynasty players over a 2–3 year window using what they've actually done — production history, age relative to their position's prime, opportunity depth, and contract situation. It deliberately ignores market prices as an input, so the model can find gaps between what a player is worth and what the market pays.

Most dynasty tools are self-referential — they reflect what other managers think, which reflects what other managers think. DELTA starts from first principles. Market value appears in the app as a comparison point and yardstick, never as an input to the score or verdict. The gap between model value and market value is the signal.

Quick reference for DELTA's terms:

• DELTA Score (0–100) — the proprietary grade of demonstrated dynasty value: production history, age vs. positional prime, opportunity, contract. League median ≈ 60.
• Market Value — consensus trade value (FantasyCalc): what the market pays. Used as a yardstick only, never a model input.
• Model Value — DELTA's own asset price, built from projected production. The gap between Model and Market is the whole product.
• Verdict (Buy / Sell / Hold) — how a player's model-vs-market gap compares to other players at a similar price. So "buy" means cheaper than comparable assets cost, not simply cheaper than the league average. A Strong verdict needs a 25% gap in either direction, and is withheld from players who have not yet had 20 games' worth of chances (seasons in the league × 17) — a first-year player cannot reach that in a 17-game season.
• Proj PPG — DELTA's projected points per game for the upcoming season, at your league's scoring.
• Start Profile — per-start outcomes across recent seasons: hit (startable week), elite (slate-breaking week), miss (hurt your lineup). Weekly dependability, not season totals.
• Alpha / Workhorse Score — role size grade (targets/touches share): how much of the offense actually flows through the player.
• System Score (0–100) — offensive environment quality: coordinator, scheme fit, role clarity. League median ≈ 60. Descriptive context only — it does not enter the projection.
• ⚠ OC change — a new coordinator: everything demonstrated came under the old scheme. Flagged for context only — the model no longer discounts for it. Rookies never had a prior scheme, so the flag never appears.
• Offense style factors — validated play-by-play tendencies (2022–25): pre-snap motion helps RB production, heavy 12-personnel squeezes WR3s, pass-lean identity (PROE) lifts WR rooms. Small (±2–4%), explained on each affected card.
• EPA — Expected Points Added per play: quality of play independent of volume. (Yards per route run is not shown anywhere in DELTA — routes-run data is not available in any free source.)
• Ripple Effects — measured value shifts on incumbent players when teammates arrive or depart, calibrated on eight seasons of history.
• Format sensitivity — how much a player's value shifts between standard / half / full PPR, from their reception volume vs. position average.
• 5th-year option — first-round rookie deals carry a team option for a fifth season; shown as "+opt" until exercised.
• Season stat line — raw counting stats on the full player page, toggleable by season back to 2022. Position-aware, and a rushing row appears for a receiver only when he actually carried the ball. These are the plain facts behind every derived metric on the page.
• † stale — a market value not refreshed in the latest fetch; excluded from graded records until it updates.

The college arm tracks production facts plus a small number of DELTA analytical metrics. Nothing here feeds an NFL projection.

• dDOM (0–100) — DELTA Dominator: competition-adjusted production expressed as a percentile within his own position. A 90 means top 10% at that position — identically for a QB, RB, WR or TE, so the numbers are comparable across positions even though the underlying metric differs. Ranked among qualified players only (6+ games plus a volume floor: 20 catches for WR/TE, 60 carries for RB, 120 attempts for QB); below that, no dDOM is shown. An analytical signal to read, not the ranking order.
• What dDOM adjusts for — the strength of competition, using the team's 247 recruiting-talent percentile. A 30% target share at a loaded Power-4 roster is harder than 50% at a thin one, so the former is adjusted up and the latter down. It does not adjust for a player's own teammates — a QB's efficiency still reflects the weapons he throws to.
• Dominator — the raw fact underneath dDOM for receivers: the average of a player's share of team receiving yards and receiving touchdowns. The classic college dynasty signal. Rush dominator is the same idea on team rushing yards and TDs, used for backs.
• ANY/A — adjusted net yards per attempt: (passing yards + 20 × TD − 45 × INT) ÷ attempts. The QB base metric for dDOM. It rewards touchdowns and punishes interceptions, so a quarterback cannot look elite on completion percentage alone by throwing checkdowns.
• PPA — predicted points added per play, the college equivalent of EPA (from CollegeFootballData, garbage time excluded). Roughly 0 is average and higher is better; the player page splits it into all plays, passing, rushing, a season total, and a season-by-season view.
• ◆ Early breakout — the player posted a top-20% dDOM while still an underclassman. Dominating young is the strongest single college-to-NFL indicator, so the badge flags the profile. It records what already happened; it is not a prediction.
• Reception share (RecSh) — receptions as a share of the team's. A stand-in for target share, which CollegeFootballData does not publish.
• Usage — share of all team offensive plays, not target share. It runs roughly half the size of a true target share; read it as relative, not absolute.
• P4 / G5 — Power 4 conferences (SEC, Big Ten, Big 12, ACC) plus Notre Dame, versus the Group of 5. A blunt competition tier; dDOM's talent percentile is the finer-grained version.
• Age (~) — always an estimate, derived from recruiting class, because no free college source publishes birth dates. Never scored, and absent for unrecruited players.
• Trajectory — dDOM by season, so a rising freshman-to-sophomore line is visible at a glance. The next season sits greyed until games are played.
• Team construction — the teammates competing for the same touches. A dominator only means something next to the players it was earned against.

dDOM was validated against real NFL outcomes: college classes 2020–2023 matched to their first three NFL seasons. The competition adjustment improved prediction at every position, but modestly — and at 94% statistical confidence, just under the 95% bar DELTA requires before a metric may influence ranking order. So dDOM is displayed as a signal and is deliberately excluded from the college ranking. College production is a genuinely noisy predictor of NFL success; a 99 means a player dominated his position against real competition, which shifts the odds rather than settling them.

Every figure traces to one of these. Where a number is DELTA's own calculation rather than a third-party feed, it says so.

• NFL production, snaps, EPA, game logs — nflverse via nflreadpy: the open play-by-play project. Also the source for rosters, draft picks, combine testing and depth charts.
• Market value — FantasyCalc dynasty consensus, refreshed daily. The yardstick DELTA is measured against, never an input to the score or projection.
• Contracts — OverTheCap, plus a small hand-maintained override file for deals signed since the last refresh.
• Injury designations — Sleeper roster tags (IR / PUP / NFI / SUS). Only a separate hand-maintained list of confirmed season-ending injuries ever zeroes a projection: an IR flag means "unavailable now", not "done for the year".
• College production, recruiting, PPA — College Football Data (CFBD). Regular season only; bowl and playoff games sit in their own section and are excluded from every rating, because participation is not uniform.
• Play-callers and coaching tendencies — built and verified by hand at DELTA, 2017–2026 at week level, because no public dataset tracks in-season play-calling changes reliably. Of 354 team-seasons: 307 confirmed against contemporaneous reporting, 32 projected for the coming season, 8 likely, 7 unverified. Building it removed 11 coaches other sources list who never actually called plays.
• Offense style factors — derived from nflverse play-by-play and FTN charting (2022–2025): pre-snap motion, personnel groupings, pass rate over expected. Each factor was validated against production before it was allowed to move a projection.
• Scarcity curves and ripple effects — not sourced, derived. Scarcity is fit on nine seasons of nflverse finishing data; ripples are calibrated on eight seasons of teammate arrivals and departures. Both are DELTA's own work.

Every pipeline that produces these files is in the public repository, so any figure can be traced to the code that made it.

DELTA is free, ad-free, and needs no account. It is built and maintained by one person. Support DELTA on Ko-fi if it is useful to you — it keeps the data flowing.

• Recent injuries or transactions not yet reflected in production data — check the data freshness badge.
• How a new offensive coordinator will actually deploy a player. The ⚠ OC flag marks the change for context, but the model can't predict scheme fit.
• Any future performance. DELTA scores what a player has demonstrated, not what they might become.
• Situational upside from teammate changes — if a player signed with a new team this offseason, the model scores their historical production and new system score, but hasn't seen them in that role yet.

Strong Buy / Buy / Hold / Sell / Strong Sell reflect market pricing relative to DELTA's model — not absolute player quality. A Sell on an RB35 means the market is paying more than the model thinks the player is worth at that tier, not that the player is useless. The Read section in each player card explains the specific gap driving the verdict.