They actually run 2^30-way tournament (they derive an equivalent form that doesn't requires 2B operations). You do not need to run the LLM, it only depends on the tokenizer.
Too bad they only formulate the detection positive rate empirically. I am curious what the exact probability would be mathematically.