At this point, I invite everyone to test it with as many simulations as they wish and to publish it, for a neutral environment.
OK. They did, and the results indicate your prefix method wins in 55.6% of cases, sequential in 40%, and ties for the rest.
Now do you understand where your error is? Or do you consider all of these identical results as a valid proof of your theory?
I'll give you a hint: you're running all tests on the same distribution, so obviously you'll get the same results.
Try again?
