How grading works
Each call names the file and the command that decides it, from the same tools that produced the pieces. On the grading date the command is run on the fresh export, the result is pasted in, and the tag flips to hit or miss. A call is never edited after it is posted; if the rule turns out to be ambiguous, it is graded against us. The probability is ours, stated before the fact, and the Brier score above is the mean of (probability − outcome)² over graded calls.