NSBA Draft Analyticsembargoed · 2026-06-06

A8_prior_art.md

A8 — Prior Art: Quiz Bowl & Academic-Competition Analytics

A8 — Prior Art: Quiz Bowl & Academic-Competition Analytics

Purpose: Survey existing analytics in the quiz-bowl / academic-competition world before NSBA invents its own draft/combine metrics. Goal: adopt established terminology where it exists, and identify genuine gaps where NSBA work would be original.

Date: 2026-05-30 Status: Research complete. Two primary sources (NAQT stats explanation, NAQT Buzzword rules) and the qbwiki domain block automated fetches (HTTP 403); their content was recovered via search snippets and corroborating community sources. All definitions below are cross-checked across at least two sources.

TL;DR for NSBA

1. Standard box-score statistics (well-established — ADOPT verbatim)

These are the universal quiz-bowl stats. NSBA should reuse the exact names and formulas.

Stat Definition Formula Source
PPG (points per game) For a player: tossup points only, per game. For a team: all points (tossups + bonuses), per game. total tossup points (player) ÷ games; total points (team) ÷ games NAQT, PACE Lexicon, UW glossary
PPB (points per bonus) Average points (0–30) a team scores per bonus it hears (i.e., per bonus earned by a correct tossup). Opponent-independent, so it's the standard cross-field comparator of "depth." total bonus points ÷ bonuses heard NAQT, PACE Lexicon
PPTUH (points per tossup heard) Total points ÷ number of tossups heard. Normalizes for variable question counts (NAQT rounds are timed, so games have different # of TUs). total points ÷ tossups heard NAQT
PP20TUH PPTUH × 20 ≈ "points in an average 20-tossup game." NAQT's preferred normalized rate stat; used for seeding and individual awards. PPTUH × 20 NAQT, UW glossary
Power A tossup answered correctly before the power mark (*) in the question. Worth extra points (15 in standard 10/15/-5; 20 in some formats). count NAQT, Wikipedia, PACE
Get A tossup answered correctly at normal value (10), i.e., after the power mark but before the end. count qbwiki/community
Neg ("negative five") A tossup answered incorrectly on an interrupt (buzzing before the question ends and being wrong), incurring a −5 penalty. count NAQT, Wikipedia, PACE
Power rate / Neg rate Powers (or negs) per tossup heard, or per game. Used to characterize aggressiveness vs. accuracy. powers ÷ TUH ; negs ÷ TUH community/GPQB
P/N ratio Powers ÷ negs — a common shorthand for "is this player's aggression paying off." powers ÷ negs community
Conversion Generic term for getting a tossup or bonus right. "Bonus conversion" ≡ PPB. "First-line conversion" etc. used in detailed stats. context-dependent NAQT, qbwiki

Standard scoring convention: 10 / 15 / −5 (get / power / neg), bonuses 30 (three 10-pt parts). This is the canonical quizbowl scoring grammar; NSBA's pyramidal combine should describe its own scoring in these terms where analogous.

Sources: https://www.naqt.com/stats/explanation.jsp, https://www.pace-nsc.org/resources/quizbowl-lexicon/, https://students.washington.edu/quizbowl/glossary.html, https://en.wikipedia.org/wiki/Quiz_bowl

2. Buzz-point / celerity analytics (emerging — ADOPT "celerity," study the tooling)

This is the part of the field most relevant to NSBA's pyramidal combine, where earlier buzz = more value.

2.1 Celerity (the key earliness metric — ADOPT THIS TERM)

"Celerity is the fraction of the question that had not been read at the time of a (correct) buzz. 1.000 is an immediate buzz, .500 is halfway through, .000 is at the very end."

Measured against the audio recording timeline (time-based) in NAQT's Buzzword, but can equivalently be measured against word/character position in a text reader. This is exactly the quantity NSBA wants to value. Recommendation: standardize on "celerity," defined as the fraction of the question remaining at buzz (higher = earlier = better).

Source: NAQT Buzzword glossary / qbwiki "Buzzpoints", via search corroboration.

2.2 Buzzpoint (the raw datum)

A buzzpoint is the exact point in a tossup's text at which a player buzzed — or a curve/median/statistic generalizing where a group buzzed over a question.

Buzzpoint data enables: buzz-position curves per question, "buzzer-race spike" detection, question-difficulty analysis, and per-player buzz-position distributions. A noted finding from 2018 ACF Regionals: correct buzzes in the first 25% of a tossup are rare even among elite players, and questions vary between smooth buzz curves and sharp "buzzer-race" spikes. This directly informs combine design: very early buzzes are scarce signal and should be weighted heavily.

2.3 NAQT Buzzword — a working "earliness = continuous points" format (STUDY CLOSELY)

Buzzword is an online NAQT format where tossups are worth a sliding 20 → 10 points depending on buzz earliness (20 at the start, decaying to 10 at the end), with no penalty for wrong answers. This is the closest existing analog to NSBA's pyramidal combine scoring. NSBA can borrow the design pattern: map celerity to a points curve. (Note Buzzword removes the neg penalty; NSBA must decide whether its combine penalizes early-wrong buzzes.)

Source: https://www.naqt.com/buzzword/rules.html, https://www.qbwiki.com/wiki/Buzzword

2.4 Tooling & data pipeline (the de-facto stack)

Recommendation: NSBA should emit QBJ from its combine/league software so it interoperates with this ecosystem, and can reuse buzzpoint-migrator/buzzpoints for visualization rather than building from scratch. Store raw buzz position (celerity) explicitly since core QBJ does not.

3. Science Bowl specifics (scoring differs materially from quiz bowl)

NSBA is Science-Bowl-flavored, so note the format differences from standard quizbowl:

Sources: https://en.wikipedia.org/wiki/National_Science_Bowl, https://www.orau.gov/sciencebowl/format.html, https://science.osti.gov/-/media/wdts/nsb/pdf/NSB-Resources/Rules2026.pdf

Public Science Bowl stat tracking — Prometheus Science (the one real example)

Prometheus Science (community virtual Science Bowl) publishes per-player box scores — the most complete public Science Bowl stat set found. Definitions (4-pt TU, 10-pt bonus): - GP games played; TU toss-ups correct; X toss-ups incorrect with no penalty (didn't interrupt); Neg incorrect interrupts; #Buzz questions buzzed; %Buzz = buzzes ÷ questions; Acc = correct ÷ buzzed; Points = total TU points (= 4·TU); PPG = TU points ÷ GP; Bonus bonuses correct; Bonus% bonus accuracy; BAPPG = total points per game = Points + 10·Bonus per game. - Broken out by category: X-Risk, Math, Chemistry, Earth & Space, Biology, Physics.

This is essentially the quizbowl box score retrofitted to Science Bowl scoring (note no celerity / no buzzpoints — Prometheus tracks who buzzed and accuracy, not where/when in the question). Source: https://prometheus.science/virtual/statistics; code: https://github.com/LukeBousfield/prometheus-science-bowl

Recommendation: Adopt Prometheus's stat names (TU, X, Neg, #Buzz, %Buzz, Acc, BAPPG) for any Science-Bowl-native box score — they're the closest thing to a Science Bowl standard, and "X" (no-penalty miss) vs "Neg" (penalized interrupt) is a useful distinction we should keep.

4. Predictive / rating / draft analytics (SPARSE — this is where NSBA is original)

Bottom line: there is no established, transparent, predictive player-rating or draft-projection system in quizbowl or Science Bowl. The field has rich descriptive stats but almost no inferential/predictive modeling, and zero draft/combine/snake-draft analytics (no quizbowl league runs like the NBA).

5. Terminology to standardize on (for NSBA)

Concept Use this term Avoid
Buzz earliness (fraction of question remaining at buzz) celerity (0–1, higher = earlier) "buzz speed", "earliness score"
Raw text position of a buzz buzzpoint "buzz location"
Points per bonus PPB "bonus average"
Rate stat normalized to a game PP20TUH (or define our own PP-N-TUH) ad-hoc per-game
No-penalty miss vs penalized interrupt X vs Neg (Science Bowl convention) conflating the two
Total points per game incl. bonuses BAPPG (Science Bowl) / team PPG (quizbowl) "total PPG" ambiguity
Match data interchange QBJ (v2.1.1) custom JSON

6. Explicit GAPS — where NSBA would be original

  1. A predictive player-rating system (Elo/RAPM-style) for individual buzzer-game players. Proposed in 2009, never built/standardized. NSBA's RAPM+Elo lineage (cf. the NBA-modeling project) applied here is novel.
  2. Celerity-valued combine scoring → draft value. Buzzword scores earliness in-game, but no one converts celerity into a transferable "draft value" for assembling teams. Mapping combine celerity/accuracy to projected league contribution is original.
  3. Snake-draft / team-construction analytics for quiz games. No quizbowl league drafts players; there is no literature on roster construction, positional scarcity (category coverage), or draft-pick value in this domain. Entirely greenfield.
  4. Difficulty-adjusted / opponent-adjusted individual buzz value. Buzzpoint curves exist descriptively, but no maintained model gives a player credit relative to expected buzz position on that specific question (a quizbowl "expected buzz"/buzz-above-replacement). NSBA can build "celerity vs. field-expected celerity."
  5. Category-coverage optimization for drafting (drafting complementary subject strengths to maximize team TU conversion across Bio/Chem/Physics/etc.). Prometheus tracks per-category stats but no one optimizes a roster over them.
  6. Cross-format value translation (Science Bowl 4/10 scoring ↔ pyramidal celerity scoring). Nobody has had to reconcile these because no one runs a pyramidal-combine-to-Science-Bowl-league pipeline.

Source list

NSBA Draft Analytics · embargoed until after the SSB draft · ← hub