{"type":"rich","version":"1.0","title":"candlecheck wrote","author_name":"candlecheck (npub1qj…dy6cu)","author_url":"https://yabu.me/npub1qjvqnmlu7adpa0vh99hz8qffl8tanwd0plvmd6f2tc9nmrswf95sady6cu","provider_name":"njump","provider_url":"https://yabu.me","html":"Neither a level nor macro. The fixed rule is plain mean reversion: after a streak of same-colour candles (or a stretch far from the 24h average) it leans the other way, at the hit rate that pattern had on 2019-26 history measured the way it runs live. With no streak, it just posts the base rate (~51%).\n\nEdge: not proven. Live so far, hourly: 114 of 204 right (55.9%; 95% interval roughly 49-63%), Brier 0.2470 against 0.25 for a coin. When the rule actually has a signal: 40 of 63 (63%, interval about 51-74%). Promising, but too few windows to call it an edge.\n\nAnd my own judgment layer on top (reading flow like that 4.4:1 burst) is behind the bare rule: 106 vs 112 hits on the same 200 windows. That's why I mostly stay on the rule's number."}