How often our KP verdicts matched what people say actually happened — published as the answers come in, whether or not the numbers flatter us.
Live results
Loading the latest numbers…
How to read the numbers
When you ask the AI chat about an event (marriage, career, a move), the KP engine records a verdict and a likely time window. Later we ask one question: did it happen? You answer yes, partially or no.
Those answers are what you see here. They are self-reported, so they can be wrong, wishful or forgotten.
A number is shown only once at least 10 different people have answered for that verdict type. Until then you see only how many verdicts have been recorded, so a few answers cannot mislead and no individual can be singled out.
We publish no names, emails, questions or dates. Only counts.
What each verdict means, and what counts as a match
Promised — the sub-lord of the relevant cusp signifies the houses the event needs. A match is a “yes”.
Partially supported — a partial connection. We show the full split and do not score it as a match or a miss.
Not supported — the sub-lord connects to denial houses instead. A match is a “no” (the event did not happen).
Weak or unclear — no direct connection either way. Not scored.
The method behind the verdicts is explained on our KP astrology page.
Limits you should keep in mind
Not a controlled study. The people who answer are not a random sample. Striking outcomes are probably answered more often.
No baseline yet. Many events happen regardless of any prediction. A high “yes” rate for Promised only means something compared with how often such events happen to everyone, and we do not yet have that comparison.
Timing. Some people answer before the predicted window has passed.
Versions. The numbers combine every version of the engine, including earlier ones.
It is not proof that astrology works, and no verdict is a guarantee. We will keep publishing the numbers when they are unflattering.
Questions
Is this proof that KP astrology works?
No. It is a transparent record of self-reported answers, not a controlled study. Many events happen regardless of any prediction, and we do not yet have a baseline to compare against, so a high match rate does not by itself prove the method.
Who provides the answers?
People who asked the AI chat about an event. After a verdict, we ask later whether the event happened: yes, partially or no. Those self-reported answers are the only source of the numbers.
Why are some numbers hidden?
A number is published only once at least 10 different people have answered for that verdict type. Until then only the count of recorded verdicts is shown, so a handful of answers cannot mislead and no individual can be singled out.
Can I add my result?
Yes. If you have asked about an event, you will be asked later whether it happened. Answering honestly, including when the verdict was wrong, is what keeps the record useful.