Players & Roles
Who a player is, how winning teams carve up the six categories, and what to actually reach for on draft night — the archetypes, the role-division, and the levers behind the wins.
A Science Bowl roster isn't five interchangeable brains — it's a set of roles. Some players buzz fast and gamble; some sit on one subject and never miss; some quietly cover the gaps nobody else can. The question this page answers is which of those roles actually wins, how the best teams stitch them together across the six categories, and — because it's a snake draft — what that means when it's your pick.
The honest complication up front: what wins changes with the format. Open-tournament play and the NSBA draft league reward almost opposite things, so we'll separate them the whole way through. Everything here is built from the same buzz-and-game logs as the rest of the Lab — read the NSBA-only figures (n=14 teams) as directional.
01What wins
the levers, and why the answer flips between formats
Start with the outcome and work backwards: which measurable team traits actually correlate with winning? Five levers, three columns — two open tournaments and the draft league — because the same trait can be a weapon in one format and dead weight in another.
| Lever (vs win%) | SSB open | dasoni open | NSBA draft |
|---|---|---|---|
| Buzz speed (celerity)* | +0.78 | +0.46 | ~0 |
| Bonus conversion | +0.69 | +0.63 | n/a |
| Breadth (broad roster) | +0.57 | +0.50 | ~0 |
| Accuracy (conversion %) | −0.10 | −0.11 | +0.40 |
| Star concentration** | −0.31 | −0.13 | −0.10 |
Open play rewards speed and breadth. Winners buzz early and cover every subject — and raw accuracy comes out slightly negative, because the teams that win are the ones buzzing aggressively, eating a few extra negs in exchange for getting there first. The draft league rewards clean accuracy instead. Once the snake draft has equalized speed and breadth across every roster, the thing left to separate teams is converting what you buzz and not negging it away — accuracy jumps to +0.40, the strongest signal in the NSBA column.
What's robust across all three formats: leaning on one star never helps. Star concentration is negative everywhere — mildly in the draft, sharply in open play. There is no format in this dataset where stacking production onto a single scorer is how you win.
Read the caveats before you over-fit. Celerity is a different metric per format, so the +0.78 is open-only and says nothing about NSBA. Star-share is partly definitional — a winning team's bigger point denominator mechanically shrinks any one player's slice. And the NSBA column is n=14: a hint, not a law.
02Player archetypes
the five kinds of buzzer, and which one reliably wins
Score every player on five axes — speed (how early they buzz), accuracy (conversion rate), aggression (their neg rate), breadth (how many subjects they touch), and volume (raw buzz count) — and the field clusters into recognizable types. Here's how those types map to team success in SSB:
| Archetype | % of players | Team win% | Net points |
|---|---|---|---|
| Anchor (broad + high volume) | 19% | .630 | 155 |
| Specialist (narrow + deep) | 34% | .555 | 67 |
| Role player (low-volume) | 34% | .399 | 37 |
| Gunner (fast + aggressive) | 4% | .686 | — |
| Sniper (slow + accurate) | 2% | .645 | — |
The Gunner and Sniper columns are eye-catching — .686 and .645 win rates — but ignore them. Eight players and four players, respectively; those numbers ride on a few lucky seasons and evaporate entirely in NSBA. The reliable finding is the top three rows: the Anchor — broad and high-volume — wins about +0.23 in win% over a pure role player, and out-nets everyone by a wide margin.
Generalists beat specialists. Players active in 4+ subjects win .577; single-subject specialists win just .378. Covering more of the board is worth more than owning one corner of it. (Energy is almost never anyone's main — only ~5 players in the entire pool main it.)
Anchor status is as much symptom as cause
Buzz volume — half of what defines an Anchor — is partly an outcome of being on a good team, not just a driver of it. A team that's blowing opponents out plays more tossups, so its best player racks up more reps and looks more "Anchor-like" precisely because they're winning. So "Anchors win" is a correlation you should hold loosely: the type and the outcome feed each other. The breadth half of the story is cleaner — 4+ subjects vs one is a real, roster-construction choice — but treat the raw volume signal as partly a scoreboard artifact.
03How top teams divide up the categories
not clean single-ownership — redundant depth
The intuitive picture of a great team is a clean division of labor: one owner per subject, six specialists, no overlap. The data says that's not what winning looks like. Split teams into top and bottom quartiles by win% and compare how they cover the board:
| Metric | TOP (win .77) | BOTTOM (win .09) |
|---|---|---|
| Distinct answerers per subject | 2.44 | 0.91 |
| Qualifying contributors per team | 3.5 | 1.5 |
| Subjects with a clear owner | ~3.8 | ~3.5 |
Look at where the gap is and isn't. The number of subjects with a clear owner barely differs — ~3.8 vs ~3.5. What separates a .77 team from a .09 team is distinct answerers per subject: 2.44 versus 0.91. Winning teams don't just have an owner for each category — they have a primary owner plus a capable backup in essentially every subject. The losing shape isn't "bad division of labor," it's thinness: not enough people who can answer at all.
The recurring winning composition. ≈ one broad Anchor + ~1.5 Specialists + role support, with about 3.5 real contributors total and a capable second body behind every subject. The edge is redundant depth, not clean single-owner division.
04The NSBA role split
thin rosters push winners toward outright ownership
Everything in Section 03 comes from open play, where teams field enough bodies to build backups. NSBA is different: rosters are thin — only ~5 players take the floor a night — so redundant depth is a luxury most teams can't afford. The consequence is that NSBA winners are more anchor-concentrated, not less. They own more subjects outright — about 4.1 of 6 — with fewer backups behind them. When you can't staff two people per subject, the path to winning is one person owning it cleanly.
dan.k.memes (.83): a hybrid two-anchor split
The top NSBA4 team doesn't run one Anchor or six specialists — it runs a hybrid. One player owns chemistry; another owns earth & space + math; another owns CS + physics; another leads biology. That's two mini-anchors each carrying ~2 subjects, plus specialists filling the rest — the thin-roster version of the "Anchor + depth" shape, compressed onto five people.
One detail worth flagging: CS behaves like any other ownable subject — it has a single carrier, just like chemistry or biology. It is not a shared round that the whole team chips in on. Whoever you draft to cover computer science owns it the same way your chem player owns chem.
05Team-building playbook
order of operations for a snake draft
The levers, the archetypes, and the role-division all point to the same handful of draft-night moves. In a snake draft the rules are different from open play — you don't get to hand-pick a balanced squad, so some "obvious" advice (cover everything!) turns out to be the floor, not a lever. Here's what actually earns picks:
- Coverage is the floor, not a lever. In open play a roster hole is fatal — a team that can't answer a whole subject wins ~0.05 of its games versus ~0.51 with full coverage, a cliff. But in a snake draft, coverage is automatic — everyone ends up spanning all six — so don't spend a high pick "securing" a subject you'd get anyway.
- Weight the high-volume categories. Not all subjects are equal in usage. Earth & Space is ~23% of tossups and Chemistry is ~18% — and chemistry is the one category with a genuine independent edge on winning. That's where the points live; don't leave either thin.
- Prize availability over a small talent upgrade. The best player sits in 25% of games. A slightly-better player who misses a third of the time is worth less than a reliable one who always shows. Where you have any signal on who actually turns up, weight it — attendance is the hidden roster tax.
- Depth in a high-usage subject is insurance, not waste. Given that 25% star-out rate, a second capable body in your best category isn't redundancy for its own sake — it's the hedge for the weeks your owner is gone. This is the one place doubling up earns its pick.
- Don't over-concentrate on one scorer. Star concentration is flat-to-negative for winning in every league measured. Balanced production wins; stacking a single carrier doesn't — and a snake draft mostly prevents the stack anyway.
·Methods & honest caveats
Every figure was computed from the raw buzz and game logs across SSB, dasoni-2, and the NSBA season database. This is exploratory analysis, not a controlled study — with one modest dataset and many correlations, treat the small-cell figures as hypothesis-generating. Specific caveats:
- Format-specific metrics. Celerity is defined differently per format, so the open-play +0.78 buzz-speed correlation does not transfer to NSBA — the ~0 in the draft column is a genuine equalization, not a measurement of the same thing.
- Star share is partly definitional. A winning team scores more total points, which mechanically shrinks any one player's share — so the negative star-concentration signal is partly an artifact of the denominator.
- Anchor status is correlational. Buzz volume is partly an outcome of winning (more reps in blowouts), so "Anchors win" is as much symptom as cause; the breadth half (4+ subjects) is the cleaner, roster-construction signal.
- Tiny cells. Gunner (n=8) and Sniper (n=4) archetype edges ride on a few players and vanish in NSBA; the NSBA draft column is n=14 teams — directional throughout.
- Availability effects are observational. The 25% star-out and 0.05-vs-0.51 coverage figures come from small, confounded cells (a team missing a subject is usually just weaker) — read them as associations and directional magnitudes.
NSBA Lab · built from scibowl.live tournament exports + the NSBA season database · July 2026.