Methodology

The Predict Six Index is published every Sunday at 1 PM ET. Each week's Index covers the Sunday 12:00 AM ET → Saturday 11:59 PM ET week that just closed, giving a full week of buffer for slow-resolving questions to land before they're scored.

What counts toward the headline number

Every Predict Six question that resolved during the Sun-Sat window. We publish 6 yes/no questions every day across six categories — Sports, Markets, Politics, Culture, Discovery, and World — so a typical week's Index covers around 42 questions.

What's excluded

Two things. Voided questions — a question is voided when the underlying event becomes unanswerable (e.g., a sports event is cancelled, a vote is postponed indefinitely); with no resolved outcome to score against, they're omitted everywhere. And, since August 2, 2026, thinly-answered questions: a question needs a minimum number of answers to count anywhere in the Index — the callouts, the category breakdowns, and the headline number. The bar is 75% of that week's average answers-per-question (never below 3), recomputed each week so it scales with the community rather than being a magic constant. Two people are not a crowd, and a two-vote question lands on 0% or 100% by arithmetic rather than by story — precisely the extremes the "biggest miss" and "biggest hit" slots select for. Each week's bar is stored with that week's data.

How "correct" is defined

For each resolved question, we compare the community majority (more than 50% of voters agreed on yes or no) against the actual outcome. If the majority matched the outcome, that question counts toward the headline as "correct." If not, it counts as a miss. We use raw majority — not weighted by individual user confidence — because the methodology has to be defensible and easily explained.

Per-category breakdowns

Same math, restricted to questions in that category. A category with 0 resolutions in the week is omitted (not zeroed out — that would be misleading).

Sample size and limitations

Predict Six users self-select. The Index is not a probability sample of the general population. Treat it as crowd performance among users who chose to answer. The "uniqueVoterCount" reported on each week's page is the sum of all per-question vote counts that week — a single user voting on multiple questions is counted once per question.

On weeks with fewer than 10 resolved questions, the headline is statistically noisy and we'll soft-pedal the framing on the X post. The webpage shows the number as-is without a small-sample disclaimer; readers should note the count.

Update cadence

Sundays at 1 PM ET, automatically. The X thread and the public webpage update simultaneously. Each week's page is permalinked at /the-index/week/{Sunday-date}/ and remains stable forever.

Methodology changes

Once the Index is cited externally, methodology changes create a credibility shock. We commit to: (a) any methodology change is announced as a footnote on the next week's Index post and added to a public changelog at the bottom of this page; (b) historical weeks are not retroactively recomputed under new methodology — they remain as published.

Changelog

Contact

Press, citations, and corrections: info@predictsixapp.com