How to turn a future prediction into questions you can actually evaluate

To evaluate a future prediction, define the event, deadline and evidence before you know the outcome. “Change is coming” gives you almost nothing to check. “I will receive a written job offer by December 31” gives you a question with a clear resolution.

Making a prediction testable does not make it true. It lets you distinguish a meaningful match from an interpretation that changed after the event.

What forecasting research contributes

The first year of a forecasting tournament reported by Mellers and colleagues in 2014 began with 2,246 participants. Their original paper describes selected volunteers answering geopolitical questions with defined outcomes and probability estimates. This was not a study of fortune telling or personal destiny. Its useful lesson here is methodological: keep the question and scoring rule clear enough to check later.

You do not need to run a tournament to make a personal claim less vague. A dated note can be enough to reveal what the original statement did and did not say.

Use a simple prediction record

Write these fields before the deadline:

• Original statement: Preserve the wording without improving it afterward.
• Observable event: What exactly would need to happen?
• Deadline: When will you stop waiting for it?
• Evidence: Which record or observation will settle it?
• Alternatives: What outcomes would count as a miss or remain unresolved?
• Your actions: What might you do that changes the outcome?

If you cannot fill in the event and deadline without inventing details, label the original statement too vague to score. Do not give it credit for a more precise prediction it never made.

A worked example

Imagine a fictional reading says, “A career opportunity arrives before the end of the year.” You ask whether an informal message counts, whether the opportunity must be paid, and whether it concerns your current field. If those details are never supplied, record that ambiguity.

For your own planning, you could separately track: “Will I receive at least one written offer for a paid role in my field by December 31?” Use an offer email as evidence. That is a new operational question you created, not necessarily what the reader predicted.

Now record your starting situation: are you already interviewing, or have you sent no applications? An outcome that was already likely provides different information from a surprising one. Avoid judging a reading from the outcome alone.

Count misses and changed behavior

Keep all recorded predictions, including ones you forget about or dislike. If the event does not occur, retain the miss instead of extending the deadline. If the evidence is unavailable, use “unresolved” rather than guessing.

Also note actions taken because of the reading. Applying for more roles may help produce an offer; avoiding applications may prevent one. Neither situation cleanly isolates what the reading knew in advance.

Reflection can stay separate

A symbolic statement can still inspire a useful question even when it cannot be scored as a forecast. You might ask what kind of work you want or what conversation you have postponed. Our guide to what Saju cannot do helps maintain that distinction.

If the prediction came from a chatbot, can ChatGPT tell your fortune? explains why generated confidence is not a substitute for evidence. Keep the reflective value and the accuracy claim in separate parts of your notes.