Metrics,Defined.

Total pass attempts in a season.

Total passing yards in a season.

Total touchdown passes thrown in a season.

Total interceptions thrown in a season.

Total rushing yards in a season.

Total rushing touchdowns in a season.

Rushing yards per game — how many yards a ball carrier averages each time his team takes the field. Where a season total tells you who ran for the most yards, YPG normalizes for schedule length and missed games, making it a fairer measure of game-to-game production. A player who runs for 1,200 yards in 10 games (120 YPG) is outperforming a player who runs for 1,400 yards in 14 games (100 YPG), even though his season total is lower.

Touchdown rate per carry — how often does a rushing attempt end in a score? This metric reveals how dangerous a ball carrier is on any given handoff, independent of volume or game count. It is especially useful for identifying high-leverage backs who play limited roles: a goal-line specialist who touches the ball 30 times and scores 12 touchdowns has a TD% of 40%, far higher than a workhorse back with 250 carries and 25 touchdowns (10%). High TD% with low total attempts often signals a targeted, situational role rather than full-game production.

Rushing touchdowns per game — how consistently does a back find the end zone each week? Where TD% measures efficiency per carry (how lethal each touch is), TDs/G measures frequency per game (how reliably he scores regardless of how many touches he gets). A workhorse back who gets 25 carries a game and scores once has a modest TD% but a reliable TDs/G. A specialist with a sky-high TD% may only play in two or three series, producing a low TDs/G. Together, TD% and TDs/G answer different questions: one tells you how dangerous the carry is, the other tells you how often the scoreboard changes.

Rushing Explosiveness Score — a composite that combines yards-per-carry efficiency with touchdown conversion rate into a single number. YPC measures how much ground a back gains per touch; TD% measures how often those touches end in a score. Multiplying them rewards backs who do both simultaneously: a player gaining 8.0 yards per carry and scoring on 12% of his attempts produces an R-EXP of 96, while a player at 5.0 YPC and 5% TD rate scores only 25. The compounding nature of the formula is intentional — a back who is merely good at one dimension is not rewarded the same way as one who excels at both. Think of R-EXP as a single answer to the question: how dangerous is this back, per carry, in both space and around the goal line?

Total receiving yards in a season.

Total receiving touchdowns in a season.

Total wins in a season, including non-division and playoff games.

Total losses in a season.

Win percentage calculated as wins divided by total games played.

Conference record — wins and losses within the team's division only.

Points scored across all games in a season.

Points allowed across all games in a season.

Average win percentage of all opponents faced during the season. Higher values indicate a tougher schedule.

Tiered defensive classification based on average points allowed per game. Elite (≤14), Strong (≤21), Average (≤28), Vulnerable (>28).

Average points scored per game during the season.

Average points allowed per game during the season.

Point differential — the margin between points scored and points allowed.

Opponent rank — the opponent's standing among all Section XI teams that season, ordered by win percentage. Rank 1 is the highest win-percentage team. Provides a quick read on opponent quality for each game.

Yards per attempt — average yards gained per pass attempt. Combines accuracy, aggressiveness, and overall passing efficiency.

Yards per completion — average yards gained on completed passes. Captures verticality, downfield aggression, and yards-after-catch impact.

Percentage of pass attempts resulting in touchdowns. Reflects scoring efficiency, red-zone execution, and explosive scoring capability.

Percentage of pass attempts resulting in interceptions. Reflects ball security, decision-making, and risk management.

Touchdown-to-interception ratio — compares touchdown passes to interceptions. Measures decision-making, scoring vs. turnover balance, and offensive discipline.

Yard Value Above Average — how far above or below a quarterback's total passing yards are compared to the average of all passers in the dataset. Designed for benchmarking and contextual production analysis.

Identifies quarterbacks who create explosive plays, scoring swings, and momentum-changing production. GBI emphasizes explosiveness, scoring efficiency, and high-impact decision-making.

Reflects a quarterback's ability to make smart, high-value, efficient decisions that maximize scoring while minimizing risk. Multiplies total touchdowns by yards per attempt, then adjusts for turnover impact.

A custom quarterback rating system designed specifically for high school football. Balances total production, scoring ability, accuracy, decision-making, and workload scaling. Adjusts for volume to reduce inflation from small sample sizes.

A composite dominance metric summarizing overall quarterback impact into a single evaluative score. Designed for rankings, leaderboards, recruiting visibility, and comparative scouting.

A football-adapted Elo rating system modeled after chess Elo. Measures relative team strength dynamically over time, adjusting for opponent strength, expected outcome, actual result, and upset magnitude. Provides dynamic power rankings, matchup prediction, and longitudinal strength tracking.

Average combined SAT score (Math + Evidence-Based Reading and Writing) for the school. A composite indicator of academic performance, used here as one of four inputs to the ACAD index.

Percentage of Advanced Placement exam takers at the school who earned a score of 3 or higher (the standard threshold for college credit). A signal of academic depth in upper-level coursework.

Percentage of the entering freshman cohort that earned a diploma within four years. The standard high school outcome measure used by the New York State Education Department.

Percentage of the graduating class accepted to at least one two- or four-year college. Captures post-secondary intent and access at the school level.

A composite signature stat that blends SAT, AP pass rate, graduation rate, and college acceptance rate into a single 0–100 score. Each input is converted to a z-score on the visible set, the four z-scores are averaged with equal weight, and the result is rescaled to 0–100. Higher is stronger relative to the cohort currently in view; the score is recomputed whenever the division filter changes, so it always answers "how does this school compare to the schools alongside it on this page."

Share of K-12 students at the school reported as Black or African American (NUM_BLACK / K-12).

Share of K-12 students at the school reported as Hispanic or Latino (NUM_HISP / K-12).

Share of K-12 students at the school reported as Asian or Native Hawaiian / Other Pacific Islander (NUM_ASIAN / K-12).

Share of K-12 students at the school reported as Multiracial (NUM_Multi / K-12).

A composite signature stat for racial heterogeneity at the school, using Blau's index — also known in the education and demography literature as the Gini-Simpson index.

The formula is 1 minus the sum of squared population shares across all six race groups (American Indian, Asian, Black, Hispanic, Multiracial, White). It answers a single intuitive question: pick two students at random — what is the probability they belong to different race groups? The raw Blau value is rescaled 0–100 within the currently visible set, matching the ACAD pattern. Higher = a more even distribution across groups; lower = a more concentrated single group. The score is recomputed whenever the division filter changes.

Share of K-12 students at the school flagged as economically disadvantaged — typically eligible for free or reduced-price lunch, public assistance, or comparable supports (NUM_ECDIS / K-12). One of NYSED's primary need indicators.

Share of K-12 students at the school classified as English Language Learners (NUM_ELL / K-12). High values indicate a district serving a substantial multilingual population.

Share of K-12 students with disabilities — students who have an Individualized Education Program (IEP) or a 504 plan on file (NUM_SWD / K-12).

Share of K-12 students at the school identified as homeless under the McKinney-Vento Homeless Assistance Act — includes students in shelters, doubled-up with other families, and unaccompanied youth (NUM_HOMELESS / K-12).

NYSED's district-level classification of need relative to local resource capacity. Tier 3 = High-Need Urban-Suburban district (highest need among Suffolk). Tier 5 = Average need. Tier 6 = Low-need district. The index is the ratio of estimated district poverty to the Combined Wealth Ratio (district wealth per pupil ÷ state average wealth per pupil). Every building inside a district shares the same tier.

Absolute change in grade 9-12 enrollment between the earliest and latest school year on file (currently 2019-20 through 2024-25, a five-year window). Positive = the high school grew; negative = it shrank.

Percentage change in grade 9-12 enrollment between the earliest and latest school year on file. Smaller schools swing more on equal student counts; use alongside the absolute Δ when comparing across the cohort.

All seasons in the database are identified by the fall year of the school year — the year the season begins. DB season 2024 means the 2024-25 school year (games played in fall 2024). This is the standard used by Newsday, Section XI, and MaxPreps. Everywhere in SportsGPT the season is displayed in school-year format: "2024-25" or "24-25". The raw integer (2024) is an internal DB identifier only and is never shown to the user.

The top-25 seasonal leaders published by Newsday are stored as "blessed" rows in the player_season_stats table. These rows are ingested once per season and never deleted or updated — they represent the official Section XI record and carry the highest data authority in the system. When a stat appears in the official leaderboard, that value takes precedence over any other calculation. All other player stats in the table are backfilled from box scores and are clearly distinguished by their source.

Box scores have been ingested for all five seasons in the system: 2021-22 through 2025-26. Coverage is near-complete for Suffolk County — for 2024-25, 238 of 240 games are represented; for 2025-26, all 224 games are present. Nassau County schools have occasional gaps because their results are reported differently by Newsday. The total box score archive covers over 1,100 individual games and links player performance directly to game context.

The player_game_stats table stores one row per player per game per stat type, sourced from Newsday box scores. Each row includes the school_slug and player_slug, enabling logos and profile links on every game stat view without any name-matching logic. Stat types covered are Passing, Rushing, Receiving, Kicking, and Defense. This table powers all game detail stat tabs — Box Score, Passing, Rushing, Receiving, Kicking, and Defense — and is the backbone for per-game leaderboards and single-game records charts.

Players who appear in box scores but did not make the official Newsday top-25 leaderboard are promoted into player_season_stats via an additive backfill process. The rebuild script (rebuild-player-season-stats.mjs) aggregates per-game stats by player and season, then inserts only new rows — it never deletes or overwrites existing data. Two guards protect integrity: a name+school+season+stat_type match and a slug+season+stat_type match. A final ON CONFLICT DO NOTHING clause handles any within-batch slug collisions. The script is safe to re-run at any time.

All 29 charts in SportsGPT are organized into eight metric families: Passing, Rushing, Receiving, Composite, Scoring, Wins, Schedule, and Defense. Each family has a dedicated data module (e.g., src/lib/chart-data/passing.ts) that owns all fetch functions for that family. The family taxonomy is enforced at the config layer — every chart config specifies its metric family, and the chart data registry maps each slug to both its family and its canonical fetch function. This structure makes it possible to add a new chart by writing one function and one registry line.

The chart data registry (src/lib/chart-data/registry.ts) is the single source of truth mapping every chart slug to its metric family and its canonical fetch function. Before this registry existed, the same fetch logic was duplicated across three consumers — the API route, the chart detail page, and the panel page — creating 87 copies of chart SQL spread across the codebase. The registry eliminates that duplication: each of the 29 registered charts has exactly one fetchFn, and all three consumers resolve to the same function via the registry. Slug validation, documentation tooling, and future generic chart loaders all derive from this one file.