Iteration 101 — ETE + Copula Entropy Double Kill KILLS #44/#45 · CERTIFICATE CLASS: 8 MEMBERS

In plain words: the transfer-entropy meter came back for its second chance — this time with the resampling-based noise floor its literature recommends instead of the textbook formula that killed it three iterations ago. The verdict: the resampling floor is better (13% false alarms instead of 16%) but still broken, because shuffling data destroys the very memory structure that de-aligned market series keep. Its companion — copula entropy, a rank-based information meter — failed the same test far worse (39% false alarms) while measuring something spectacularly real (27 to 43 times its noise floor) that no other instrument reads (it barely correlates even with its exam partner). The broken-certificate family now counts eight members across four exam campaigns, and one mechanism has earned a unique distinction: the intensity→duration information flow is real, twice-confirmed, and now twice-orphaned — both noise floors its literature offers are invalid on this substrate. Its only way forward is the matched-pair proposal already on the operator's desk: pair the real readings with the one betting-based noise floor this campaign has certified. Two candidates remain; the batch closes next.