Position-aware composite of process rates. Within-position percentiles are reliability-shrunk by minutes, then display-scaled for the card. Higher is stronger process quality for that position.

Model Score versus current price within position. Positive means the player looks underpriced for the model; negative means you pay a premium. Players with under 90 current-season minutes are left blank.

Expected minutes for the next gameweek (0–90). Locked starters sit near 75–90; rotation risks sit lower. Insights use this as minutes security.

One process tag built from Model Score, Value Edge, and xMins. Green is positive, white is caution or neutral, red is negative or too early. Insights are not the same as content BUY/SELL calls or Creator Calls.

High Model Score, fair or better Value Edge, and locked minutes. Strong all-round profile — a priority squad piece.

Strong Model Score with a clearly positive Value Edge and solid minutes. Quality that looks underpriced relative to the model.

Solid (not elite) Model Score, strong Value Edge, and locked minutes. A cheapish starter who frees budget for premiums elsewhere.

Elite Model Score and locked minutes, but a mildly negative Value Edge. Worth owning for quality — you pay a premium for it.

Strong Model Score and good Value Edge, but minutes look fragile. Track the role before treating as nailed.

Locked minutes and a good/strong Model Score, but a poor Value Edge. Template-tax starter — reliable, expensive for the output.

Decent Model Score and/or Value Edge on paper, but xMins say not nailed. Looks attractive until the benchings start.

Decent Model Score, fair price, and okay minutes. No urgent transfer action from these three metrics alone.

Weakening Model Score with a negative Value Edge and soft minutes. Funds are working poorly — consider moving the money.

Weak Model Score and/or bad Value Edge with poor minutes security. Low process appeal on the core trio of metrics.

Model Score exists, but there is not enough current-season sample for a reliable Value Edge yet. Wait for more minutes before judging.

On the player card: last-3 arithmetic mean of per-90 process rates. This is the figure to quote for goals prevented, saves, VAEP, xAG, SCA, and the other card process rates.

Stats Lab and advanced-metrics tables use a last-3 EMA (exponential moving average). Do not quote that EMA as if it were the card Relevant stats mean.

Season FPL totals shown in the card header — not the last-3 Relevant stats window.

Relative bars for the card’s Relevant stats versus the current catalog cohort. Length shows standing in the pool, not an absolute 0–100 quality score.

Recent shot locations and quality for the player. Complements the rate stats; it is not a Model Score input by itself.

Current FPL price and selected-by percentage on the card header.

Very easy fixture for the subject team.

Easy fixture.

Average fixture.

Difficult fixture.

Very difficult fixture.

Mean difficulty of the next five gameweeks for a team or player card, with an Easy / Hard style outlook. Blank gameweeks cut occupancy and soften the run.

A blank removes a fixture in that gameweek. Double gameweeks add fixtures. Both change next-5 occupancy and planning tilt.

Valuing Actions by Estimating Probabilities (all on-ball actions).

Offensive VAEP — chance-creating and finishing value of actions.

Expected Threat from progressive ball movement.

Shot quality attributed to the player.

Raw PitchAPI expected assisted goals. Sparse weeks use xag_filled (PitchAPI → FBref → Understat → FPL expected_assists) for chance creation.

Per-row xag_filled rate: PitchAPI xAG when present, else FBref xAG, Understat xA, or FPL expected_assists (see xag_source).

Actions leading directly to a shot.

Actions leading directly to a goal.

Team xG of possessions the player was involved in.

xG chain excluding the shot and the key pass / assist.

Passes that advance the ball toward goal by a meaningful distance.

Carries that advance the ball toward goal by a meaningful distance.

Passes that lead directly to a shot.

Touches inside the opponent penalty area.

Saves above expected goals on target faced (GK).

Goalkeeper saves per 90.

Completed passes (GK distribution).

PitchAPI defensive action count (tackles, interceptions, etc.).

Aerial duels won.

Loose-ball recoveries.

Official FPL DefCon / CBIT from GW-live.

Creators naming the player as a transfer in or strong buy for the window.

Creators naming the player as a sale or move away.

Creators flagging the player to monitor without a firm buy or sell.

Creators naming the player for the armband (or as a named captain option).

Creators planning to sit the player for the gameweek.

Sum of buy, sell, watch, captain, and bench mentions for that player.

YouTube caption summary for a gameweek window. Bubble counts are creator handles, not Model Score Insights. Each refresh is stamped VYYYYMMDD and kept in the Creator Calls archive.