Computed, or not said
The one rule every Scorewit page lives by.
Sports trivia has a quiet quality problem. The internet is full of confidently wrong answers: quiz sites that say a team won a tournament it didn't, listicles that misdate finals, and lately, AI chatbots that will invent a plausible scoreline rather than admit they don't know. Trivia is supposed to be the one genre where the facts are the whole product. Mostly, nobody checks.
Scorewit runs on one rule, and it shapes everything we ship:
Every fact is computed from official match data, or it isn't said at all.
Not "researched." Not "written carefully." Computed. Every question, every answer, every reference page on this site is derived programmatically from public, official match records, and then checked again by a separate program whose only job is to disagree.
The GenEva system
The pipeline has two halves that are deliberately kept apart. We call it a GenEva system: a generator and an evaluator.
The generator reads the frozen dataset of match results and derives facts from it. Who won, how many times, how far a team got, head-to-head records, biggest wins. Every sentence on a Scorewit page traces back to rows of match data.
The evaluator is a completely separate checker. It re-derives every published claim through its own independent logic and compares. It shares no reasoning with the generator. If the two halves ever disagree about a single fact on a single page, the site does not deploy. Nothing overrides that. There is no "publish anyway" button.
This sounds like overkill until you watch it work.
The rule bites us, and that's the point
The most useful thing about a hard rule is what it forbids us from doing.
Quick trivia question: who won the 1950 World Cup final?
Trick question. There was no 1950 World Cup final.
Nineteen fifty is the only World Cup that ended in a round-robin pool instead of a knockout final. The match everyone remembers, Uruguay stunning Brazil before a record crowd at the Maracanã, was simply the last match of that pool. It decided the title, so the world remembers it as "the final." It wasn't one.
We found this out the honest way: our generator assumes nothing, and when it went looking for the 1950 final to compute the champion from, the data had no final to give it. The casual version of history, the one repeated on half the quiz sites on the internet, does not survive contact with the match records. To say "Uruguay, champions" at all, the generator had to derive the final pool's standings from its results, and the evaluator had to re-derive the same table through its own logic before the page could publish.
That is what the rule buys. A human writer would have typed "Uruguay won the final" and been wrong in a way no reader would ever catch. The computation could not be casual, so our 1950 page isn't. It says, exactly: "Uruguay won the 1950 World Cup, topping the final round-robin group ahead of Brazil," and its FAQ answers the question head-on: "Was there a final at the 1950 World Cup? No. The 1950 title was decided by a final round-robin group. Uruguay beat Brazil 2-1 in the deciding match." The data refused our assumption, and the assumption was the thing that was wrong.
The rule cuts the other way too. For a while, our East Germany page refused to say how far they got in the 1974 World Cup, because that tournament's odd format blurred two different stages together in the records, and the computed answer would have understated a fine run. When we later found a reliable way to tell the stages apart in the data, the answer came back, and it came back more accurate than what we had published before the rule made us stop.
And when a tournament ends, nobody here types the winner into a page. The data updates, the generator re-derives, the evaluator agrees, and the site knows. This summer, the World Cup champion appeared on the site within hours of the final whistle: past tense, correct, untouched by human hands.
Why bother
Because the alternative is the internet we already have. Generating plausible sports content has never been cheaper, and plausible-but-wrong is the failure mode of the era. A trivia game that's wrong isn't a game. It's misinformation with a score counter.
Every page on Scorewit carries the same line in its footer: computed from official match data, independently machine-verified, updated nightly. It's not a slogan. It's the build failing if it stops being true.
The facts come from public, openly licensed sports data projects, the quiet heroes of this whole site, credited on every page they power. The games are free, need no login, set no cookies, and give you six verified questions a day.
That's the deal: we don't say anything we can't prove, and what we do say, the generator and the evaluator had to agree on first.
Jay
Scorewit is seven daily sports trivia games on one site: scorewit.com.