The Value Board — WPA, Buzzer Elo & the MVP Case

Win Probability Added — "win shares for science bowl"

Full leaderboard, top 20. Reconciles across all 30/30 games. SOLID

The clutch board (2nd half, game within 14)

Who shows up when it's tight. Small samples — a handful of clutch plays each — so this is flavor, not a title. TINY n

• Kensuke O +0.53 · Advai S +0.47 · Advik S +0.42 · Jianyu W +0.42 · William J +0.38 lead the clutch splits.
• Cold under pressure: Daniel Y (−0.09), Ishaan K (−0.08), Sanjay S (−0.06).

Points Above Expectation — the corroborating stat

Top overperformers — the MVP case's third witness. SOLID (canonical-grouped; a handful of unmapped buzzes are dropped, so team totals slightly undercount — doesn't affect the ranking.)

Fairness note: at the bottom, Ryan K is −7.1 gets below expectation (1 get on ~8 expected), the league's biggest underperformer. Short season — this moves.

Everyone ranked head-to-head

Validated: across 183 decided head-to-head duels (both players real contenders, ≥5 shared on-floor tossups), the higher-Elo player held the winning buzz record 96% of the time (predicted-vs-observed correlation +0.92). The overall ladder is the trustworthy headline; per-subject splits are thinner (below). Overall ladder top 25 — each on 21–78 contests. SOLID

Naming note for air: Rahib H is the canonical name for the player known in-game as snapninja (id-map confidence only ~0.50 — same person, pick one name). "Abhinav A" correctly merges two GidTheKid2 accounts.

⭐ Stats lie, Elo doesn't

• Patrick L is #14 in raw points-per-game but #53 in Elo (1415) — big counting stats piled up against weak fields. The classic empty-numbers guy. SOLID
  Say "Patrick L looks like a top-15 scorer on paper. Head-to-head against real competition, he's outside the top 50. Volume against weak defenses."
• Caveat in the other direction (keeping ourselves honest): the Elo slightly under-penalizes negs — a correct buzz beats the whole opposing field, but a neg only loses to the one player who cleans it up. So a high-volume gunslinger like Ishaan K (9 gets but 9 negs in 4 games, zero net points) rates higher in Elo (#23) than his real game value — we flag him as a draft bust elsewhere for exactly that reason. Don't over-trust a single rating. CAVEAT

The category specialists COLOR

Fun to name, but every per-subject leader below is provisional by the model's own bar (provisional = under 15 contests) — each sits on ~9–14 contests, and the runners-up are thinner still. Call them "current leaders," not settled crowns.

Who beats whom — the expected-win matrix

From the same Elo fit: the probability the row player wins the buzzer over the column player on a neutral tossup. Top-6 overall. SOLID (as a model)

Read a row left-to-right: Akhil B is a favorite over everyone, but it's razor-thin at the top — he's only a 53–47 pick over Edward C and a 54–46 pick over Kian D. The top three are a virtual tie; the edges only open up once you drop to Suzuko O and Roshan A.

⭐ Marquee subject matchups

Where the per-subject fit produces genuinely lopsided expected duels: COLOR (per-subject, provisional)

• Physics — Lucas W over Kian D, 77%. The single most lopsided marquee physics gap. Lucas W and Harry G, though, are a coin flip (51–49).
• Chemistry — Kian D over Roshan A, 70%, and over Eric L, 68%. Kian D is the clearest per-subject #1 on the board.
• Comp Sci — Owen M over Edward C, 71%. Owen M's CS rating is the highest single-subject Elo in the league (1817).
• Biology — Eric L over Kian D, 63%. Bio is the flattest subject up top: Eric L is barely ahead of Uddip K (51%) and Theenash S (52%).

The pecking order — who's the go-to guy on each team

Correction from an earlier draft: a "rivalry" needs two opponents. But opponents only play once, so any two enemies share just ~4 subject tossups — too few to crown anyone. What the data can show cleanly is how teammates divide the subjects — who's the designated specialist. (The Harry G / Monish S "duel" I first flagged? They're teammates on GidTheKid2 — that's a role split, not a rivalry.) For real head-to-head, the Elo matrix in Part 4 is the honest tool. SOLID

GidTheKid2's split personality

Harry G takes physics (8 gets), Monish S takes math — two teammates each owning a lane. Not a rivalry; a division of labor. SOLID

True head-to-head → see Part 4 COLOR

Genuine opponent duels are all ~4 shared tossups — too small to claim "owns" (e.g. Eric L 4–0 Kian D in bio, Edward C 3–0 Kian D in math, on 4 tossups each; fun to note, nothing more). The validated Buzzer Elo matrix above is how you actually rank who beats whom.