How it worksNOW = what price is doing in the last 15–60 minutes (updates every 30 s). Next 4 hours = a model re-learned every month and tested on the following month it never saw (about 33 unseen months, 3 years); its hit rate is printed next to it, and it only shows LONG/SHORT where that test passed (BTC and DOT: one call per UTC 4-hour block, decided at the block start).
Next move ranges = where price stayed 80% of the time over the next 4 h / 24 h / 7 days, re-tested every night on 3 years of unseen months. LONG / SHORT appears only for a signal that passed 3 years of unseen tests (BTC and DOT 4 hours, DOT next day); everything else says No signal.
Next day leans = the only daily rule that passed 228 tests on 5–9 years of history (trend, sentiment, funding, stablecoins, stocks, dollar, gold, yields, options, CME, Fed liquidity, attention, calendar): after a big 3-day move the next day tended to go the other way. Re-tested every night on days it never saw and shown only while it passes; small edge after fees.
The page re-checks the market every 5 minutes; a new 4-hour trade opens only at the 4-hour decision (opening at any 5-minute check was tested and was less accurate). Before the decision the 4H box shows the model's reading (how close it is to a signal), not a chance: the chance tables did not pass their tests. Blinking is on box borders only: green = moving up / LONG, red = moving down / SHORT, amber = forced trades.
4H / Day / Week tabs on each chart = that timeframe's fixed range and its tested signal only (entry, target, exit if wrong). On 2026-10-02 we tested whether a second reading makes a 4-hour signal more accurate: the coin's own daily or weekly trend, the other coin, the whole market, or a stronger score. None did (a stronger score was right less often), so no extra "confirmed" label is shown.
4-hour target = half the 4-hour 80% range edge in the call's direction, exit if wrong = the opposite range edge; their hit rates come from the 3-year walk-forward test. After fees = every test result minus a standard exchange round trip (the 0.05% taker fee in and out of a regular Binance / OKX account, the median spread of Binance, Bybit, OKX and MEXC, a 0.01% slippage allowance and funding for the time held), re-measured every night; MEXC is used only for public coin data; "profit not proven yet" means the average swings too much from trade to trade to be sure of it, even when the hit rate is proven. Nothing here is guaranteed or financial advice.