I think your sentence kinda contradicts itself though, or did I misunderstand ?
They are using probabilities of prefix occurrences to predict where other prefixes are.
no, you're the one who's wrong. According to the latest mcd script, what I see is that they're scanning everything, leaving the least unlikely places for later. This is mathematically true. In 256, it's unlikely to find 2 "aa"s, so in a 256 block, when they find "aa", they skip the rest of the block and leave it for later. This makes a lot of sense theoretically. However, in practice, both methods work the same, so they're equally efficient. Neither breaks any laws, neither is better.
Yep, 100% agree. Neither is better.
Isn’t “skipping for later” predicting that it’s statistically not there ? I mean they don’t do all the range stop and start for pleasure, they do it because they think it makes you go faster. This is where I disagree.
And also, it does not make sense “theoretically” - theoretically every event is independent and having 2 “aa” has zero impact on the odds of finding a 3rd one in the next key.