Three more August runners trailed out, four fresh entries were cut inside a week, and the cumulative record moved from below zero to above it. Same rule, same stops, no decisions made.
The month is still a thrust; the week was a pullback. All 27 markets we track are above where they closed 30 days ago, and only 2 of 27 are up on the week. Bitcoin gave back 4.0% over seven days after a 22.4% month, and sits 38.9% below its one-year high. That is the shape of a breakout wave taking a breath, not one ending, though the rule does not care which: it only sees channels and stops.
Volatility is split down the middle. 13 of the 27 markets are in their low-volatility regime on the measure our staged filter uses (this week's 8H range against each coin's own year). No market is within one ATR of its upper channel. Three sit within one ATR of their lower channel, so if the next signals come, they lean short.
Loud on bitcoin, quiet on everything else. 234 bitcoin items this week, 58 labelled positive against 20 negative, while the price fell. The next-busiest coin had 31 items. The tape was arguing about an asset that was quietly going down, which is the usual arrangement.
Two positions are open. As always, the count is published and the names are not.
Three of the four winners were entries from the third week of August that had been trailing their stops higher for three weeks. All three exited the same way, touched by the ratchet after the pullback. Winners were held 18.7 days on average; losers 3.3 days. That asymmetry is the whole system: the losers were cut before they could matter, the winners were left alone until the rule said otherwise.
Two further positions closed after Friday's cutoff. They join the record on the site when their 48 hours pass, and next week's issue. Two signals were skipped because one minimum lot would have risked more than the sleeve allows; we count those instead of hiding them.
The record stands at +10.97R after 75 closed trades (57 days live). Against the probability envelope fixed on 19 July 2026, before the first trade closed, that is inside the band, between the 50th and 75th percentiles. Three weeks ago it sat below the 5th. Nothing was changed in between, which is the point. The full chart is on the Record page.
This is the item we would rather not write, so it goes first. One of the three rules staged for the October review is a volatility filter: skip an entry when the coin is louder than its own typical year. Since 12 August its shadow ledger has been recording, for every real entry, whether the filter would have skipped it. On 9 September we found the ledger had been using a daily proxy for the measure instead of the 8H rule the filter was actually validated with. Zero skips had been booked in four weeks; the validated measure would have booked five.
The fix went live the same day: the ledger now computes the validated measure from the same bars the research used, and the live evidence for the October verdict counts only from that moment. The four weeks of proxy data are kept, labelled as what they are, and count for nothing. Its first real skip landed on 10 September, a short in a loud market that stopped out the next day: 1.01R saved on paper. One event is an anecdote, which is why the verdict waits for the review.
The other two staged rules kept their own ledgers busy. The concurrency cap would have refused two of the week's fresh entries, both of which stopped out within days. The early-exit rule would have closed a going-nowhere position a week sooner than the channel did. All three ledgers are running; all three verdicts are due 1 October. The full list of what has been tried and rejected is in Tested & Rejected.
Not investment advice. Past performance does not predict future results. We publish outcomes, not recommendations. Closed trades appear after a 48-hour delay; figures are R-multiples on the same basis as the public record.