NSBA · Opening Night
Ask Bot: Opening Night
158 questions in 140 minutes. We rated every one — and wrote up the six best.
The bot shipped tonight and the league descended on it. We logged every question, then had a panel of AI reviewers rate each one on a single axis: how interesting was the question? — i.e. would it reward a real dig into the data. Here's the night in numbers, and the questions good enough that we actually wrote full deep-dives on them (linked below).
the night at a glance
158 questions · avg interest 4.3/10 · 22 historical/hypothetical · 23 politely declined (14%) · 12 token-burning trolls
What did people ask?
Analytical and creative questions dominated — but the standouts, and every one we turned into an article, came from the two smallest buckets: historical and hypothetical. Those are the questions the data can actually reward.
Questions by type
Who answered?
Every question is routed by difficulty: a cheap model for one-fact lookups and declines, the big one for the creative/hypothetical reasoning. Opus drew the most work tonight — because the crowd leaned hard into hypotheticals and roasts.
Model routing
The most interesting questions
Everything the panel scored 7+. The linked ones got the full treatment — real queries, charts, the works.
| Score | Type | Question |
|---|---|---|
| 8 | historical | What was the worst team ever in NSBA history? → deep dive |
| 8 | historical | Can you rank the top 10 players of NSBA with season context (so like the same player can show up multiple time → deep dive |
| 8 | historical | Take the top 5 players in each season of NSBA and explain which season had the best all-NSBA team and why → deep dive |
| 8 | analytical | Take the top 5 players in each season of NSBA (1 2 3 and 4) and explain which season had the best all-NSBA tea → deep dive |
| 7 | creative | attach each team in the nsba to a theme song and singer |
| 7 | creative | give each nsba team their own taylor swift album archetype based on how they play and perform? |
| 7 | creative | Given that science bowl players often go into careers like research, medicine, quant, use the NSBA4 performanc |
| 7 | creative | If the world was in an apocalypse and you needed 3 NSBA4 players to be on your team, who would you pick and wh |
| 7 | creative | if every player was an animal, what would they be |
| 7 | creative | compare performative nsba science bowlers to high school stereotypes |
| 7 | creative | if every top player was a president, who would be who? |
| 7 | hypothetical | If NSBA4 had instead been only teams of 2 with the draft order reamining the same and being in a snack as well → deep dive |
| 7 | creative | which 10 nsba players, based on their performance in nsba4, would be most liekly to fall victim to a honeypot |
| 7 | analytical | Who had the most goon buzzes (buzzes before it's possible to get the question correct) whether they got it rig |
| 7 | hypothetical | How would arolakiv (NSBA1) have done in the NSBA 4 league? Estimate his stat line → deep dive |
| 7 | historical | give me the nsba goat list from nsba 1 2 3 and 4 all time top 10 → deep dive |
| 7 | historical | Take the top 5 players in each season of NSBA (1 2 3 and 4) and explain which season had the best all-NSBA tea → deep dive |
| 7 | analytical | Who had the biggest glowup from NSBA 3 to NSBA 4 |
| 7 | historical | who was the best player for each category throughout all seasons of NSBA → deep dive |
| 7 | analytical | how has akhil developed across nsba3 and 4 and what are your presictions for playoffs |
The guardrail worked overtime
Nearly one in six questions (23) got a one-line decline. The players are real high-schoolers, so anything sliding toward dating, looks, ethnicity, or private life got a good-natured deflection and a redirect to a real stat. The crowd probed that boundary a lot tonight — and it held every time. Another 12 were pure token-burning ("give me the longest message you can", "use as much budget as possible"); the bot mostly no-sold those too.
The verdict
Average interest was a middling 4.3/10 — opening night skewed toward trolling and one-fact lookups. But the top of the distribution was genuinely excellent: cross-era GOAT debates, counterfactual re-drafts, and "who was the worst team ever" are exactly the questions a stats bot with four seasons of history should live for. We wrote six of them up. Ask better questions and we'll write more.
How this was made
All 158 questions from the bot's feedback log were rated 1–10 for "interestingness" by a panel of AI reviewers (three, each rating a third), then the highest-scoring historical and hypothetical ones were handed to writer agents that queried the real season database and produced the linked deep-dives. Question text is shown as asked. Nothing here identifies anyone beyond their quiz-bowl handle.