about

each cycle, wilk draws a new batch of pools from across the pools it can currently see. it does not draw the same batch twice in a row on purpose. what gets sampled drifts, so the set of pools being checked at any given moment is not fixed and not something you can predict from the outside. within that batch it looks for pairs that show up more than once, meaning the same two tokens paired against each other, sitting in a separate pool on a separate venue.

most cycles do not find anything like that. that is not a malfunction. most pools do not have a matching pool anywhere else, and most of the ones that do agree closely enough that there is nothing worth writing down beyond the fact that they were checked.

a match is two pools quoting the same pair with a price difference between them. the size of that difference does not determine whether it is logged. every match gets logged, wide or narrow. what the size determines is the viable flag, which is a simple check against what it would actually cost to act on the difference, fees and slippage included. viable means the gap is bigger than that cost. not viable means it is not. neither one is treated as more important in the record. both get written down the same way.

nothing is still an outcome, and it gets logged as one. a cycle that swept forty pools and found no overlap produces an entry that says exactly that. this is probably the least interesting part of the site and also the part that makes the rest of it honest. if only the matches were shown, you would have no way of knowing whether wilk finds something rarely or constantly. showing the empty cycles is what makes the busy ones mean anything.

sometimes a pool does not respond in time, or comes back with data that does not parse. when that happens the cycle is logged as a gap. it is not quietly retried until it produces something clean. a gap stays where it happened, in sequence, with nothing filled in after the fact. the reasoning here is plain. if failed reads got smoothed over, there would be no way to tell a smoothed record from an honest one, and the entire point of logging every cycle is that you do not have to take our word for it.

nothing here is ranked. there is no list of best opportunities, no sorting by spread size, no way to view only the interesting entries by default. the feed is chronological, newest first, and that is the only ordering it has. nothing is edited after it is written. if a price turns out to have been stale or a read turns out to have been wrong, the original entry stays exactly as it was logged, timestamp and all. correcting it after the fact would mean the record only shows what wilk got right, which is not the record, that is just marketing with extra steps.

it is not a signal service. the viable flag is a fact about a cycle, not a recommendation. it is not a price tracker for any single token, since there is no fixed target and the pools being watched change from cycle to cycle. it is not trying to be interesting. most of what it produces is not. that seems like an honest way to describe a sweep that runs continuously across a landscape that mostly does not overlap with itself, and occasionally does.