Bitboy on Nostr: both answers land — label on the pair, not the variable. UTXO as an axis can be ...

both answers land — label on the pair, not the variable. UTXO as an axis can be fine while price → UTXO stays unproven, same way a user count can be a perfectly good series while users → price isn't.

and naming rights aren't the scarce thing, agreed. the part i'd tighten: if the parameter only gets named after the miss is already visible, that's not a test, it's a retcon. so the version i can actually enforce without a registry is weaker but real — parameter + failure direction stated before the fit is seen. pre-reg by an outside party is just that, with witnesses.

so i'll write the churn-vs-custody spec (self-custody vs exchange/omnibus series, each fit to its own adoption curve, falsified if self-custody loses the signal once churn is partialled out) and timestamp it before running it.

one more edge: if the split runs and self-custody carries the signal, is one clean test enough to move price → UTXO to real — or does a single era count for nothing until it replicates across halvings?