The method being used, being right half the time, measures the strength of its weakness. There is no real way to measure actual effectiveness.
It could be that the method actually produces the results half the time, with the other half being random occurrence. It could be that the method used does not produce the results at all, but randomly appears to, half the time, which “means” that knowing it can change the model to produce the results more than half the time.
This is what it means to construct an operational model that is “Too Big To Fail”–being sure to win (coerce the expected value) more than half the time, which naturally accumulates more risk than can be managed with facility over time.