The play. Every BTC 5-minute candle, wait until trigger seconds before the close, then buy
whichever side is ahead at whatever the book asks — no fair-value gate, no edge gate. Who's "ahead" comes
from our own resolution feed when it's healthy, else the book's favorite; each shot records which one
decided (basis). The shot is skipped when the favorite prices outside min–max price — too
dear to be worth it, or too cheap to really be the favorite. With RANDOM on, each candle draws
its own trigger from the range. Stake is what a shot spends — it buys stake ÷ entry price in whole
shares, and each winning share pays $1, so a cheaper entry pays more. Paper and live keep separate books,
so neither borrows the other's P&L.
Smart mode. A shot only fires when our feed (never the book fallback) has had the side
ahead by min lead for hold seconds straight, the model is at least min sure, and the
market agrees by pricing it inside min–max price; otherwise the candle is passed. In 12.8k past
candles a 5bp lead held from 0:60 to 0:30 never lost — but no filter is truly 100%, and at 0.99 one loss
erases about 100 wins. Pacing: the session stops at take profit and rests, any loss rests
longer, and after a loss the next prove wins shots go to paper and must all win before live resumes.