The Weird Prediction Markets Report

77.1% of everything that ever landed on this board showed up once and never came back. Four hundred and three markets, one appearance each, gone by morning. That's the board in one number: a revolving door dressed up as a leaderboard.

This report runs on our own daily market-board snapshots — 94 days of them, April 25 through August 9, covering 523 unique markets across 938 total board slots. Nobody else has this file. It's not the whole prediction-market universe, just what made our curated top-10 each day.

Over four and a half months we watched 523 distinct markets cycle through a board that only ever shows ten at a time. Median board volume climbed from $8,108 in April to a peak of $25,285 in June before settling back to $20,536 in July. Mean payout size did the same climb — $16.93 in April, $39.65 by June — which tells you the board got both busier and weirder as summer went on.

All five hit the $2,000 payout ceiling, and every single one lasted exactly one day. Below the ceiling, the board kept finding new ways to be strange: Lebanon's Taqaddom Party paid $666.67 to win ("Hezbollah pivoting to competitive pickleball"), Delhi Capitals winning the 2026 IPL paid $400 ("not with a bang but a 0.0025"), and Kim Moon-soo winning a South Korean governorship also paid $400 ("four hundred dollars says hello from the void").

While hundreds of markets flashed and vanished, a handful just stayed. The lunar landing market — "Will NASA land a person on the Moon before 2028?" — logged 48 days on the board between April 26 and July 25, more than double the runner-up, captioned "one small bet for man." Starship-carries-humans-to-Mars held 17 days, the Japan 8.0-magnitude earthquake market held 15 (across two separate stints), and Taylor Swift's papal audience market also ran 15 days, still open as of this report's last snapshot. The Iran nuclear deal market matched it at 15 days before resolving out of the picture by May 31.

Nothing moved harder than "Trump drops more UFO files this month," which swung from a 3% YES on June 8 to 65% YES on May 28 — a 62-point round trip in six days. The California 8.0 earthquake market swung 29 points across its 10 tracked days, a perfect-game market jumped from 14.5% to 39.5%, and South Korea's 3% inflation market swung 23.5 points in just four tracked days. Even the slow-burn Japan earthquake market drifted between 43% and 66% over its 15 days — proof that "long-running" and "stable" aren't the same thing here.

April was a soft launch — 58 board slots, dominated by Science and Technology. May and June each filled out to 300 slots, and that's when "Other" took over as the top tag. Median volume roughly tripled from April's $8,108 to June's $25,285. July cooled slightly (250 slots, median volume $20,536) but kept the same category order. August, so far, is a partial month with Politics making the top five for the first time all year.

This board is a curated daily top-10 (up to 30 in busier stretches), not a scrape of the entire prediction-market world. There's a real gap in the record too: snapshots stopped after July 25 and didn't resume until August 7, a 12-day hole in an otherwise daily cadence. Category labels are the scanner's own automated tags — useful for trend-spotting but imperfect. None of the payout figures should be read as predictions; they're illustrative snapshots of market prices at the time, and markets move.