At the moment, you've done very little to test it out so how do you know that's not a false positive? QS or someone else will just bring this up later.
I think I get your point. But to be fair to me, I haven't really set any criteria which count as a "positive" to start with. It's not like I set thresholds and said, people with these metrics are alts. What I did was create a model of the language of quickseller, and tested that model as a predictor of other account's texts. I then released these results. The results are merely descriptive. They put a direct number on how similar a particular set of texts is to the model of quickseller. They don't say "if you're this similar, you must be an alt". That last part is something I purposely left for other people to judge for themselves.
Maybe this will help clarify things: the model is a collection of all of the 1, 2 and 3 word phrases that quickseller has used along with their frequencies across his .5 million words. It's a pretty big model. The perplexity numbers I published are nothing but direct evaluations of the postings of other accounts as they relate to the model of quickseller. To put this in more layman's terms: if I used the same 1 2 and 3 word phrases at the same relative frequencies as quickseller, then my posts would get a very low perplexity score when we use the model of quickseller to predict them. It turns out that I don't do this, and neither does dooglus, and neither does hillariousandco. So far, these are the only corpora I've checked.
I hope this explanation helps. I fell like the way you're calling it my "script" makes it sound like I may have done something mysterious. I'll I've done is publish some concrete descriptive statistics about the words of 6 particular bitcoin talk accounts. What you take away from this is up to you.
This thread hasn't been derailed. It's all about TSP's script, QS, Panthers, their shady or alleged shady activity.
Maybe, given the explanation above, and the fact that I've provided all the code you'll need to replicate my work, you could start calling it "the information that tsp presented" or "tsp's experimental results". I feel like calling it my "script" or "algorithm" is a bit of a misuderstanding.