Data
Open data
Every register publishes its raw entries. Not a summary of them, and not a chart we drew — the rows, so you can run the analysis yourself and disagree with ours.
File
| Register | Status | File |
|---|---|---|
| The Blind Horoscope Test | Protocol published · collection not yet open | blind-horoscope-test.csv ↓ |
| The Forer Machine | Protocol published · collection not yet open | forer-machine.csv ↓ |
| The Psychic Self-Test | Protocol published · collection not yet open | psychic-self-test.csv ↓ |
| The Tarot Frequency Lab | Protocol published · collection not yet open | tarot-frequency-lab.csv ↓ |
A register that has not opened has no rows to publish yet. The endpoint exists and returns a header, so a script written today keeps working the day it opens.
Why the raw file
A published percentage asks you to trust the arithmetic behind it. A published file does not. Everything on this site that claims to be measured can be recomputed from these downloads with the code beside them, which is the difference between showing a result and asserting one.
What is held back, and why
Two things. Same-day entries, until 00:00 UTC — a participant who could see today's distribution before choosing would be anchored by it, so the seal is enforced in the database rather than the interface. And anything that could identify a person: no IP hashes, no account details, and a stable pseudonym in place of the participant identifier so repeat participation stays analysable without being traceable.
Reading the files
CSV, UTF-8, comma-separated, with a comment header explaining that register's columns. participant_key is a pseudonym, consistent within a file, so you can group by participant without knowing who anyone is.
Verifying the card game
The Zener export is the one worth poking at. Each row carries the commitment shown to the participant before they guessed, plus the card and the nonce revealed afterwards. Hash the card, a colon, and the nonce; the result must equal the commitment. If it does, that card could not have been chosen after the guess — and you have checked that yourself rather than believing us.
The analysis code
The preregistered analysis is in the repository and runs against these same files. It imports the site's own statistics library, so a number on a page and a number from the script cannot diverge — which is also its weakness, and the reason to swap those imports for your own.
Licence
CC BY 4.0. Reuse it, including commercially, with attribution to occultresearch.org. If you publish something built on it we would like to read it, but you do not need our permission.