Rationale behind predicting
Apart from maybe being fun and surprising (if enough people take part in such a prediction round), it also has value in creating better, more reliable and consistent questionnaires.

If I were to ask you "have you ever felt so agitated that you couldn't sit still or was overactive" or "have you ever taken risky chances you later regret", you might grab the DSM and look at the description of bipolar disorder and see "psychomotor agitation" and "excessive involvement in activities with the potential of painful consequences" or see these behavioural characteristics in friends or colleagues with BP, and having already made up your mind, you have should have little problem manipulating the test to get the "correct" results.

That is how many questionnaires are made: just questions that ask you about symptoms that are part of the syndrome—as described in e.g. the DSM. To increase possibilities for measuring if someone is just guessing or manipulating, some questions are rephrased and asked again, but that's it—at the very best.

Since aetiologies of psychotic disorders (and mental disorders generally) are unknown, the disorders being just syndromes, there is no way to test for validity: when someone (subconsciously or deliberately) manipulates the test, there is no way of telling.

Therefore it is important to come up with questions that are what may be called "extra-syndromal" or "trans-syndromal": true for people having the syndrome, but not known to anyone else or at least not strictly associated with the syndrome.

The best way to make sure that is the case is to ask for characteristics that (pretty much) everyone with the syndrome shares with each other—but unbeknownst to them—and to ask for things that aren't apparent or strictly behavioural.

The prediction round is to assess which questions are not strictly associated with the disorder, or each other, by those that have it (making it likely that others are even less likely to make those associations).

The answer round is meant to see which answers answers are validity indicators.

Finding an optimal set of questions is done by taking both validity, independence and reliability into account.

Finally, inverse (possibly/likely multivariate) correlations with answers given by people with other disorders, determines which questions can be used for differential diagnosis.

There are probably far better questions than these 10, so this game/process is cyclic: removing useless questions and adding new ones (like a card game).

Given the small number of predictions so far, the prediction round will be extended further and will end when we have 5 predictions (again, giving correlation predictions isn't strictly necessary).