As a matter of policy, I would never accept such a prediction.

This may be unfair to the predictor but in scientific evaluation the rule is to reject anything where fraud is possible.

Roger