No one listens to me even though I am nearly always correct on the markets. How many examples would you need.
Would you mind making each future market prediction in a binary form (which can be objectively judged as true or false by a specified date) along with a specific numeric probability estimate (which you can subsequently update any time)?
Probabilistic predictions are okay as long as you keep a comprehensive list of all of them. You can then test how many of the "80%" predictions are satisfied relative to the ideal 80%.
Being explicit about the probabilities inherent in all predictions is important if you want them to be evaluated objectively. To measure the accuracy of a forecaster such as Anonymint, you can use a proper scoring rule, such as the Brier score, applied to his complete history of quantified predictions (including any updates).