TIA
TRX
XRP
GRAM
DOT
BCH
APT
FIL
HYPEHow the momentum score works
The universe is today's 100 most traded Binance USDT pairs with at least $5M of 24-hour volume. Stablecoins, wrapped tokens, leveraged tokens and Binance's tokenized stocks are left out; memecoins stay in. A coin can only be picked when its 15-minute close is above its 20-period average, that average is above the 50-period one, its 1-hour close is above the 1-hour 20-period average, it is up over the last 4 hours and the buyers lead the sellers on the directional index (+DI above −DI).
The score adds up to 100: 25 points for the 4-hour gain against Bitcoin, 15 for the 24-hour gain against Bitcoin (both ranked against every other coin), 20 for a volume surge (the last hour against the median hour of the day), 15 for a new 24-hour high, 15 for trend strength (ADX) and 10 for an RSI between 55 and 75. A coin more than 4 ATR above its average loses 15 points and one that jumped more than 8% in the last hour loses 10, because chasing a spike that already happened is the classic momentum mistake. Scores of 75 and more are rated A, 60 to 75 B, the rest C; nothing under 50 is picked.
The trade and how it is graded
Each pick is a spot long entered at the close of the 15-minute candle at the round time. The stop is 2 times the 15-minute ATR below the entry, kept between 1.5% and 6%, and the target is 2 times that risk above it. The trade runs until the stop or the target is hit, with no time limit, and nothing else closes it early. While it runs, the same coin is not bought again; a new round takes the next best coin instead. A candle that touches both levels counts as stopped, and every result pays 0.1% fees on each side. The same trade is placed on every coin of the round (followed for up to 7 days), which gives the "average coin" that the picks must beat; picking at random would land there.
How the system learns every day
After each UTC day the scanner reviews every closed candidate trade of the last 30 days (the trade on every coin that passed the uptrend filter, not only its own picks, so the sample is large enough to say something). For each number it watches (the score, the gain against Bitcoin, the volume surge, RSI, ADX, the distance to the 24-hour high, the stretch above the average, the stop distance, Bitcoin's own move and the hour of the round) it splits the trades into thirds and compares how each third did. Thirds that earned more add points to similar coins, thirds that lost take points away, at most 15 either way, and small samples count for less. Nothing is learned before 30 trades have closed. A fixed-rules copy keeps picking by the plain score, so the record shows whether the learning helps. The daily journal lists every trade of every day, which won, which lost and why.
Why the AI picks separately
The score is a rule; the AI reads the same numbers for the 10 best candidates it does not already hold, the last 12 hours of each and the latest lessons, and chooses the 3 it thinks will reach the target before the stop. It never sees prices after the round time and it cannot change the levels. Keeping its record apart shows whether the judgement adds anything to the rule. The AI analyst makes slower 4-hour calls of its own.
Use it responsibly
Momentum on 15-minute candles is the fastest and noisiest kind of trading on DanioX: most of the move is often over before anyone reads a list, and stops get hit within minutes. The first rules (M1, 22 October 2026) bought a coin again while it was still held and closed every trade after 24 hours; since 23 October (M2) each coin is held once until its stop or target and the only tuning is the daily review, out in the open with its own control. The M1 trades stay in the archive of the track record. The slower, backtested setups are on the signals board. Nothing here is financial advice; read the disclaimer.