Evidence, not promises
Every day the engine ranks every coin in the scan from 0 to 100 against the rest of the market. Below is what those ranks went on to do over the following seven days — first replayed over four years of stored history, then graded live, call by call, as the product runs. The table updates itself every night. Blank cells mean fewer than thirty samples; nothing here is tuned to look good.
Each UTC day the engine scores trend alignment, momentum and overextension for every liquid coin, ranks the scores as percentiles, and logs the board. Rank 90+ is the only bullish claim; everything else is tracked for its return but claims nothing.
Seven closed daily candles after the call: the close (the 7-day return), the highest high (best point) and the lowest low (worst point), from stored exchange candles — never the live ticker. The replay sees only bars before each call; no look-ahead.
The edge, where there is one, lives in the tail: the median top-decile call loses a little, the average gains, and the pullback verdict is the strongest slice. Regime matters — bear tapes flatten everything. Read it as a base rate for a kind of setup, not a forecast for a coin.
Research and analytics only. Not financial advice. Past behaviour of a scoring rule says nothing certain about its next call; crypto assets are volatile and you can lose money.