How to check a prediction was made before the game
Anyone can say they called it once the final is in. Here is how to tell a real pregame prediction from one written afterwards, with a real night of Clutch's sealed numbers as the example.
BY UĞUR ALTUNBAŞ · OCT 23, 2026
Look for a fingerprint, a clock nobody can set, and the misses.
A real pregame prediction leaves three marks: a fingerprint of the numbers published before tip, a timestamp the forecaster cannot set, and a record that keeps every miss. Screenshots and posts made after the final prove nothing, and a dated blog post can be edited the next morning.
Why it matters
A track record is easy to fake without a single lie you can point to: post only the calls that came in, delete the ones that didn't, nudge a number after the injury news, or write the prediction up once the game is over. Each one makes a forecaster look better than it is. The only defence is a record that was fixed in public before the games and can't be quietly trimmed afterwards.
1. A fingerprint before tip
Clutch writes every game's number into a file and publishes the file's SHA-256 fingerprint: 64 characters computed from its contents. Change one digit of one number and the fingerprint comes out completely different, so a published fingerprint pins the file down exactly. On Oct 24, the night's first file held 8 games and its fingerprint, shortened, is 92d63a11…5deed0.
2. A clock nobody can set
The date on a file or a commit is typed by whoever makes it, so on its own it proves nothing. What counts is when the file reached a public server that keeps its own time. Clutch's files go to a public GitHub repository, and GitHub records when each one arrived. That Oct 24 file arrived at 10:01 AM ET, 7 h 58 min before the first tip. When news moved a number later that day, the file was sealed again, 6 more times; the last seal before each game counts. The file → · GitHub's arrival times →
3. The misses stay
Each file names the fingerprint of the one before it, so an old file can't be swapped out or dropped without breaking the chain. After the final, every call is graded against the sealed number, right or wrong. That night, Clutch's calls held in 6 of 8, and the misses are on the Ledger beside the hits. Calls that can't be shown to predate tip aren't graded at all, and the Ledger prints how many: 31 so far this season. Every graded call →
Check one yourself
Download any night's file, compute its fingerprint with a tool your computer already has (shasum -a 256 on a Mac, certutil -hashfile on Windows), and compare it with the one on the Proof page; then check its arrival time against the tip. It takes about two minutes. Step by step, on the Proof page →
What a seal doesn't prove
A seal proves when a number was made, not that it was any good. For that you need many graded games: how often the favourites won, and how each band of numbers held up. How accurate Clutch is → · Calibration and the Brier score →
The example above is rebuilt from Clutch's public files every day.