Data Dictionary — raw
Data Dictionary — RAW sources
Processed-table schemas live in DATA_DICTIONARY_PROCESSED.md (written by parser agents).
Everything in data/raw/ is read-only; never modify.
Season mapping
| Season | Year | Combine file | Game scoresheets folder | Question source |
|---|---|---|---|---|
| NSBA1 | 2022 | combine/nsba1_2022_combine.xlsx |
games/nsba1_2022/ (Rounds 1–11 + Play-In/Playoffs/Finals) |
alumni-written |
| NSBA2 | 2023 | combine/nsba2_2023_combine.xlsx |
games/nsba2_2023/ (RR1–RR11) |
alumni-written |
| NSBA3 | 2025 | combine/nsba3_2025_combine.xlsx |
games/nsba3_2025/ (Rounds 1–7 + Playoffs) |
packet submissions |
| NSBA4 | 2026 | combine/nsba4_2026_combine_INPROGRESS.xlsx |
— (drafting now) | packet submissions |
Categories (6)
Biology (b/bio), Chemistry (ch/chem), Computer Science (cs), Earth/Space Science (ess),
Math (m/math), Physics (p/phys). NSBA replaced Energy with Computer Science.
Combine workbooks
- Layouts differ by season:
- NSBA1: one tab per player (handles/nicknames), small per-tab grid.
- NSBA2: organized by category sheets (
Ovr,Bio,Chem,ESS,Math,Physics,CS). - NSBA3 / NSBA4:
Overallsheet + one tab per player. NSBA4 also hasPUBLIC Overall. Overallsheet columns (NSBA4): Rank, Name, Paid?, Discord Tag, Overall, Bio, Chem, ESS, Math, Physics, CS. Red rows = ineligible (no payment or no combine).- Player tab format (e.g.
Barth R): per category, rowsBio1..Bio4etc. with a Points column and a Penalty column, then a category Total; six categories in two column-blocks. - Combine scoring: +4 superpower / +3 power / +2 correct (earlier buzz = more); −1 only for a wrong multiple-choice (WXYZ). Question number 1–4 ≈ difficulty within a category (pyramidal). Players may skip. Gameable — some players tank / "bike."
- Per-question availability (confirmed 2026-05-30): NSBA1 ✅, NSBA3 ✅, NSBA4 ✅ have per-question (Bio1..CS4) data. NSBA2 is AGGREGATE-ONLY (category totals + Overall, CS included) — no per-question exists in any provided file (verified: 0
BioN-style labels). NSBA2 feeds category/overall analyses but not the per-question IRT. - File notes:
nsba1_2022_combine.xlsxandnsba1_2022_combine_full_with_schedule.xlsxare the SAME NSBA1 combine (identical players/totals; the latter is a richer copy with a Signup/schedule sheet = possible availability signal, + video links).nsba2_2023_combine.xlsxis the canonical 53-player NSBA2 copy;*_51p.OLD.xlsxis the earlier 51-player copy. NSBA1 = 5 categories, Energy not CS.
Game scoresheets
- Each workbook: a
templatesheet + one sheet per matchup named"teamA vs teamB". - In a game sheet: row 1 team names; row 2 final scores; up to ~5 player-name columns per side; a central block with TOSSUP #, SUBJECT (cycles the 6), BONUS; per player a paired (T = points, B = boolean) column.
- Game scoring (standard Science Bowl, NOT pyramidal): toss-up correct = +4; wrong interrupt / wrong toss-up = −4; bonus (follows a correct toss-up) = 10. Exact mechanics are validated by the parser against each sheet's displayed final score.
Discord exports
data/raw/discord/— 20 plaintext channel exports (NSBA4 - <channel> [id].txt).- Richest:
general(~450KB),off-topic,questions,request-combine,welcome,Combine - combine-discuss. Mined by agent A5 →data/processed/discord_player_signals.csv.
Other
data/raw/combine_pdf/NSBA4_Combine.pdf— the NSBA4 combine question set/packet.- scibowl.live SSB-2026 data +
~/code/scibowl-org— external real-tournament results; used mainly for the natural (school) vs. drafted roster study, NOT heavily for player rating.