Loading…
How Blindside Alpha builds projections, tiers, and scores – and where the numbers come from.
Nearly every number traces back to a play, a snap, or a public box score – and where a value is a derived index (like VORP), we say so. No third-party rankings — our scores are computed in-house. Raw charting and tracking inputs that power the models are credited on the data credits page. No mystery scores with undefined inputs.
This page documents exactly what data we pull, how each model works, and – critically – where the models fall short. Understanding limitations makes you a better analyst.
All data originates from official league feeds or authoritative public sources. We do not rely on third-party scrapers for quantitative inputs.
Data Pipeline
Each stage runs in isolation – failures are isolated and retried without blocking downstream consumers.
Every snap of the current NFL season, including down, distance, field position, personnel, formation, route labels, and outcome.
Snap participation by game and role. Drives share-of-snaps and opportunity rate metrics.
Time-to-throw, receiver separation, completion probability over expected (CPOE), and rushing yards over expected (RYOE) – sourced from official tracking data.
Per-route targets, yards, and snap context. Powers YPRR with a play-by-play proxy fallback where full route data is unavailable.
Practice designations, injury classification, and status updates pulled from the official league feed.
Implied team totals and point spreads from the consensus market. Used for game-script adjustments in the projection model – not surfaced as wagering data.
Aggregated beat-reporter and transaction news via RSS. Displayed on the News Desk; not used in quantitative models.
Refresh Schedule
Projection Model – Inputs to Outputs
Confidence / Error Band
We show the full P10–P90 range – not a single point estimate dressed as a prediction.
Small-Sample Warning Zone
Efficiency metrics (EPA, YPRR) are flagged below 50 snaps and weighted accordingly.
Opportunity-based expected output: what the plays were worth, not what the box score says
xFP assigns an expected fantasy point value to each target, carry, or snap based on field position, down-and-distance, and personnel context — then sums these per-play values to produce a player's expected output for the week or season. The gap between actual fantasy points and xFP isolates conversion luck from opportunity volume.
Inputs
Outputs
Update cadence
Updated weekly after all games resolve
Confidence req.
Minimum 4 games for stable gap signal; 8+ games for reliable pos_week_rate
Used in
Limitations
Position-specific expected touchdowns from red zone opportunities, weighted by zone proximity
The Red Zone expected-TD model converts a player's red zone opportunity profile into an expected touchdown count using position-specific league TD-per-opportunity rates, weighted by zone proximity. Inside-5 opportunities (the goal line) carry a dramatically higher TD rate than the broader 6–20 yard zone, so the model weights them separately. The result is expected_td — and td_diff = rz_td − expected_td determines whether a player is over- or under-converting (the BUY-LOW/SELL-HIGH split).
Inputs
Outputs
Update cadence
Updated weekly via supabase/migrations/20260601150000 migration
Confidence req.
Meaningful td_diff requires 8+ red zone opportunities; I5 rate requires 4+ inside-5 looks
Used in
Limitations
Per-game PPR scoring, normalized for opportunity
Combines raw fantasy points per game with snap-share and target-share context to produce an opportunity-adjusted weekly output score.
Inputs
Outputs
Update cadence
Updated after each game
Confidence req.
Minimum 3 games played
Used in
Limitations
Model probability vs. implied market probability
Compares the model's win/outcome probability to the implied probability in the market line. The difference is reported as 'edge' – a signed percentage.
Inputs
Outputs
Update cadence
Refreshes as lines move
Confidence req.
Confidence score > 50 for actionable edges
Used in
Limitations
Value Over Replacement Player – positional scarcity premium
Calculates how many fantasy points a player scores above the replacement-level player at the same position (typically the last starter in a standard 12-team league).
Inputs
Outputs
Update cadence
Recalculated weekly
Confidence req.
5+ games for stable VORP
Used in
Limitations
Excess return vs. market consensus expectation
Measures how much a player consistently outperforms their consensus ADP and pre-game projection. Players with sustained positive alpha are systematically undervalued by the market.
Inputs
Outputs
Update cadence
Updated weekly
Confidence req.
8+ games for reliable alpha
Used in
Limitations
Monte Carlo simulation of your fantasy matchup
Runs 2,000 Monte Carlo simulations of a fantasy-league matchup – your projected roster vs. your opponent's – drawing each starter's score from a normal distribution around their projection. Produces weekly win probability, season playoff odds, and first-round bye probability.
Inputs
Outputs
Update cadence
Recalculated as projections update
Confidence req.
Requires live roster and opponent data from connected league
Used in
Limitations
Discount future weeks, weight playoff weeks, account for positional scarcity
Calculates the net present value of a player's rest-of-season contribution using discounted weekly projections, schedule difficulty adjustments, and positional scarcity penalties. The current Trade Desk uses a simpler PPG-surplus proxy until this ships.
Inputs
Outputs
Update cadence
Recalculated weekly
Confidence req.
High confidence only for Weeks 1–4 of the projection horizon
Used in
Limitations
Draft capital + ADP + rookie touches → readiness index
Composite 0–100 score built from three heuristic components: draft pick capital (50%), fantasy ADP as a market-consensus signal (30%), and rookie-season targets or carries per game (20%). No vendor data required – all inputs are derived from public draft and fantasy data.
Inputs
Outputs
Update cadence
Draft season; updated as in-season rookie touches accumulate
Confidence req.
Score is most reliable after 8+ games when production component stabilizes
Used in
Limitations
Contract-year motivation signal
Identifies players in the final year of their contract and applies a historical production premium based on observed walk-year performance patterns across similar age/position profiles.
Inputs
Outputs
Update cadence
Static per season; recalculated at off-season
Confidence req.
Applies only to confirmed contract-year players
Used in
Limitations
Formation and alignment signal for the upcoming week
Analyzes a player's pre-snap alignment tendencies and their opponent's historical defensive responses to those alignments. Produces a verdict card for the upcoming matchup.
Inputs
Outputs
Update cadence
Weekly
Confidence req.
Minimum 4 games of formation data for both player and opponent
Used in
Limitations
Defender success rate in coverage against similar receiver profiles
Evaluates how a cornerback or safety has performed in coverage against receivers matching the upcoming opponent's profile (alignment, route depth, speed tier).
Inputs
Outputs
Update cadence
Weekly
Confidence req.
30+ coverage snaps against similar receiver type
Used in
Limitations
Route-level efficiency across the full tree
Maps a receiver's YPRR, target rate, and catch rate by route type across the current season. Surfaces which routes in their tree are generating real value vs. which are volume fillers.
Inputs
Outputs
Update cadence
Updated after each game
Confidence req.
50+ routes run for stable per-route splits
Used in
Limitations
Weighted composite of five high-value situational usage zones, percentile-ranked within position
Leverage Score quantifies how much of a player's usage comes from the plays that actually produce fantasy points. Five situational zones are measured — Goal-Line (GL), Red Zone (RZ), Close-and-Late (C/L), Two-Minute drill (2M), and Late-Down (LD) — each capturing a player's share of team touches in that situation. The weighted composite is then percentile-ranked within position to produce a 0–100 score. A player who dominates goal-line and two-minute usage has genuinely outsized scoring relevance relative to their snap share.
Inputs
Outputs
Update cadence
Updated weekly after all games resolve
Confidence req.
Meaningful signal requires 6+ games; GL and 2M components need 4+ opportunities in each zone
Used in
Limitations
Measures how much of a player's fantasy output is driven by touchdowns rather than volume
TD Dependency computes the fraction of a player's total fantasy points derived from touchdowns (td_pts ÷ total_fpts), then percentile-ranks this ratio within position to surface players whose value is unusually TD-dependent. A player with a high TD Dependency percentage is likely overperforming their volume-based outlook and carries regression risk; a player with low TD Dependency is scoring primarily through yardage and receptions — more sustainable production.
Inputs
Outputs
Update cadence
Updated weekly after all games resolve
Confidence req.
Minimum 6 games for a stable td_dep_pct; players with fewer than 4 touchdowns may show volatile readings
Used in
Limitations
Risk-adjusted fantasy output: consistency vs. ceiling, modeled as a financial Sharpe ratio
Fantasy Sharpe adapts the financial Sharpe ratio to weekly fantasy scoring: S = mean_fpts ÷ std_dev_fpts. A high Sharpe player scores reliably near their average (cash-game safe); a low Sharpe player has a wide boom/bust distribution (tournament upside). Boom% is the fraction of weeks above a position-specific floor threshold; Bust% is the fraction below a separate floor. The scatter hero plots players on a mean/std-dev plane with iso-Sharpe contour lines.
Inputs
Outputs
Update cadence
Updated weekly after all games resolve
Confidence req.
Minimum 6 games for stable mean and std-dev; 10+ games for reliable Sharpe
Used in
Limitations
Air yards share + target share, weighted 0.7/1.5 to balance volume and downfield usage
WOPR (Weighted Opportunity Rating for Receivers) is a single receiver-efficiency metric that combines target share and air yards share with weights chosen to balance sheer volume against downfield opportunity. WOPR = 1.5 × target_share + 0.7 × air_yards_share. A player with high WOPR commands both frequency of looks and depth of routes — the combination most predictive of sustained fantasy production. WOPR above 1.0 typically signals a true WR1 role in the passing game.
Inputs
Outputs
Update cadence
Updated weekly after all games resolve
Confidence req.
Minimum 4 games; most informative after 8+ games
Used in
Limitations
Opportunity-based expected output: what the plays were worth, not what the box score says
xFP assigns an expected fantasy point value to each target, carry, or snap based on field position, down-and-distance, and personnel context — then sums these per-play values to produce a player's expected output for the week or season. The gap between actual fantasy points and xFP isolates conversion luck from opportunity volume.
Inputs
Outputs
Update cadence
Updated weekly after all games resolve
Confidence req.
Minimum 4 games for stable gap signal; 8+ games for reliable pos_week_rate
Used in
Limitations
Position-specific expected touchdowns from red zone opportunities, weighted by zone proximity
The Red Zone expected-TD model converts a player's red zone opportunity profile into an expected touchdown count using position-specific league TD-per-opportunity rates, weighted by zone proximity. Inside-5 opportunities (the goal line) carry a dramatically higher TD rate than the broader 6–20 yard zone, so the model weights them separately. The result is expected_td — and td_diff = rz_td − expected_td determines whether a player is over- or under-converting (the BUY-LOW/SELL-HIGH split).
Inputs
Outputs
Update cadence
Updated weekly via supabase/migrations/20260601150000 migration
Confidence req.
Meaningful td_diff requires 8+ red zone opportunities; I5 rate requires 4+ inside-5 looks
Used in
Limitations
Per-game PPR scoring, normalized for opportunity
Combines raw fantasy points per game with snap-share and target-share context to produce an opportunity-adjusted weekly output score.
Inputs
Outputs
Update cadence
Updated after each game
Confidence req.
Minimum 3 games played
Used in
Limitations
Measures how much of a player's fantasy output is driven by touchdowns rather than volume
TD Dependency computes the fraction of a player's total fantasy points derived from touchdowns (td_pts ÷ total_fpts), then percentile-ranks this ratio within position to surface players whose value is unusually TD-dependent. A player with a high TD Dependency percentage is likely overperforming their volume-based outlook and carries regression risk; a player with low TD Dependency is scoring primarily through yardage and receptions — more sustainable production.
Inputs
Outputs
Update cadence
Updated weekly after all games resolve
Confidence req.
Minimum 6 games for a stable td_dep_pct; players with fewer than 4 touchdowns may show volatile readings
Used in
Limitations
Value Over Replacement Player – positional scarcity premium
Calculates how many fantasy points a player scores above the replacement-level player at the same position (typically the last starter in a standard 12-team league).
Inputs
Outputs
Update cadence
Recalculated weekly
Confidence req.
5+ games for stable VORP
Used in
Limitations
Excess return vs. market consensus expectation
Measures how much a player consistently outperforms their consensus ADP and pre-game projection. Players with sustained positive alpha are systematically undervalued by the market.
Inputs
Outputs
Update cadence
Updated weekly
Confidence req.
8+ games for reliable alpha
Used in
Limitations
Contract-year motivation signal
Identifies players in the final year of their contract and applies a historical production premium based on observed walk-year performance patterns across similar age/position profiles.
Inputs
Outputs
Update cadence
Static per season; recalculated at off-season
Confidence req.
Applies only to confirmed contract-year players
Used in
Limitations
Weighted composite of five high-value situational usage zones, percentile-ranked within position
Leverage Score quantifies how much of a player's usage comes from the plays that actually produce fantasy points. Five situational zones are measured — Goal-Line (GL), Red Zone (RZ), Close-and-Late (C/L), Two-Minute drill (2M), and Late-Down (LD) — each capturing a player's share of team touches in that situation. The weighted composite is then percentile-ranked within position to produce a 0–100 score. A player who dominates goal-line and two-minute usage has genuinely outsized scoring relevance relative to their snap share.
Inputs
Outputs
Update cadence
Updated weekly after all games resolve
Confidence req.
Meaningful signal requires 6+ games; GL and 2M components need 4+ opportunities in each zone
Used in
Limitations
Air yards share + target share, weighted 0.7/1.5 to balance volume and downfield usage
WOPR (Weighted Opportunity Rating for Receivers) is a single receiver-efficiency metric that combines target share and air yards share with weights chosen to balance sheer volume against downfield opportunity. WOPR = 1.5 × target_share + 0.7 × air_yards_share. A player with high WOPR commands both frequency of looks and depth of routes — the combination most predictive of sustained fantasy production. WOPR above 1.0 typically signals a true WR1 role in the passing game.
Inputs
Outputs
Update cadence
Updated weekly after all games resolve
Confidence req.
Minimum 4 games; most informative after 8+ games
Used in
Limitations
Formation and alignment signal for the upcoming week
Analyzes a player's pre-snap alignment tendencies and their opponent's historical defensive responses to those alignments. Produces a verdict card for the upcoming matchup.
Inputs
Outputs
Update cadence
Weekly
Confidence req.
Minimum 4 games of formation data for both player and opponent
Used in
Limitations
Defender success rate in coverage against similar receiver profiles
Evaluates how a cornerback or safety has performed in coverage against receivers matching the upcoming opponent's profile (alignment, route depth, speed tier).
Inputs
Outputs
Update cadence
Weekly
Confidence req.
30+ coverage snaps against similar receiver type
Used in
Limitations
Route-level efficiency across the full tree
Maps a receiver's YPRR, target rate, and catch rate by route type across the current season. Surfaces which routes in their tree are generating real value vs. which are volume fillers.
Inputs
Outputs
Update cadence
Updated after each game
Confidence req.
50+ routes run for stable per-route splits
Used in
Limitations
Draft capital + ADP + rookie touches → readiness index
Composite 0–100 score built from three heuristic components: draft pick capital (50%), fantasy ADP as a market-consensus signal (30%), and rookie-season targets or carries per game (20%). No vendor data required – all inputs are derived from public draft and fantasy data.
Inputs
Outputs
Update cadence
Draft season; updated as in-season rookie touches accumulate
Confidence req.
Score is most reliable after 8+ games when production component stabilizes
Used in
Limitations
Model probability vs. implied market probability
Compares the model's win/outcome probability to the implied probability in the market line. The difference is reported as 'edge' – a signed percentage.
Inputs
Outputs
Update cadence
Refreshes as lines move
Confidence req.
Confidence score > 50 for actionable edges
Used in
Limitations
Monte Carlo simulation of your fantasy matchup
Runs 2,000 Monte Carlo simulations of a fantasy-league matchup – your projected roster vs. your opponent's – drawing each starter's score from a normal distribution around their projection. Produces weekly win probability, season playoff odds, and first-round bye probability.
Inputs
Outputs
Update cadence
Recalculated as projections update
Confidence req.
Requires live roster and opponent data from connected league
Used in
Limitations
Discount future weeks, weight playoff weeks, account for positional scarcity
Calculates the net present value of a player's rest-of-season contribution using discounted weekly projections, schedule difficulty adjustments, and positional scarcity penalties. The current Trade Desk uses a simpler PPG-surplus proxy until this ships.
Inputs
Outputs
Update cadence
Recalculated weekly
Confidence req.
High confidence only for Weeks 1–4 of the projection horizon
Used in
Limitations
Risk-adjusted fantasy output: consistency vs. ceiling, modeled as a financial Sharpe ratio
Fantasy Sharpe adapts the financial Sharpe ratio to weekly fantasy scoring: S = mean_fpts ÷ std_dev_fpts. A high Sharpe player scores reliably near their average (cash-game safe); a low Sharpe player has a wide boom/bust distribution (tournament upside). Boom% is the fraction of weeks above a position-specific floor threshold; Bust% is the fraction below a separate floor. The scatter hero plots players on a mean/std-dev plane with iso-Sharpe contour lines.
Inputs
Outputs
Update cadence
Updated weekly after all games resolve
Confidence req.
Minimum 6 games for stable mean and std-dev; 10+ games for reliable Sharpe
Used in
Limitations
No model is perfect. Being explicit about limitations is part of the product – because understanding the error bars is what makes you a better decision-maker.
Weekly fantasy projections carry an inherent ±4–6 point standard error for skill positions. Floor/ceiling bands reflect this – the midpoint is not a guarantee.
Efficiency metrics like EPA/play and YPRR require 50+ snaps to stabilize. Early-season or injury-depleted samples are labeled accordingly and should be weighted lightly.
We surface injury designations but do not model injury probability. A player's projection does not factor in unannounced health changes – always check the wire before games kick off.
Implied team totals inform game-script adjustments, but garbage time, weather, and coaching decisions can diverge sharply from model assumptions.
We don't publish model accuracy statistics on this page. Track record claims require rigorous out-of-sample testing – that infrastructure is in the roadmap below.
ECR rankings and rank-spread metrics measure analyst sentiment. They are useful for identifying volatile or divisive players – not as a primary valuation input.
We're committed to publishing model accuracy statistics. It takes rigorous out-of-sample testing to do this credibly – here's the phased plan.
Track mean absolute error (MAE) and bias per position week over week. Establish baseline accuracy benchmarks.
Out-of-sample edge model testing vs. realized outcomes. Identify systematic biases by game type and market condition.
Publish weekly calibration results to users. Full accuracy transparency – prior-week projections vs. actual scores.
Rookie Breakout Signal is a signal-detection read of first-season usage, efficiency, and role trajectory — not a projection, ranking, or guarantee of future production. Scores are normalized within the 2025 rookie cohort by position.
Data transparency
| Type | Inputs |
|---|---|
| Faithful | Early snap% & routes (wk1–4), earned-by-wk4, target share / touches / TPRR / weighted opportunity, YPRR / EPA-per-target / EPA-per-rush / EPA-per-dropback / CPOE, red-zone opportunity share / goal-line share / two-minute share / air-yards share, snap% & opportunity early-vs-late deltas |
| Proxied | Usage stability (inverse opportunity variance), small-cohort scoring (cohorts <6 use league positional percentiles), red-zone series (display-only sparkline, not scored) |
| Omitted | First-read target share, missed/broken tackles, player explosive-play rate, big-time-throw rate, pressure-to-sack |
Component weights
Confidence modifier bands
| Band | Condition | Modifier |
|---|---|---|
| Full | Sample ≥ threshold AND games ≥ 12 | ×1.00 |
| Moderate | Sample ≥ 60% AND games ≥ 8 | ×0.90–0.95 |
| Limited | Sample ≥ 30% AND games ≥ 4 | ×0.75–0.85 |
| Tiny | Below the above (early/sparse season) | ×0.60–0.70 |
Threshold = 100 routes (WR/TE), 50 carries (RB), or 100 dropbacks (QB). Modifier interpolates continuously within each band by ratio.
Not wagering advice. All projections, tiers, and metrics on this platform are statistical estimates for research and entertainment purposes only. Blindside Alpha is not a sportsbook, a wagering operator, or a paid-picks service. Past model performance does not guarantee future results. See our Terms of Service for full details.