Winning 96.6% of the time, and still losing money.

The lesson, stated plainly. A 96.63% win rate still loses money when the average price paid is 0.98 — the price eats the edge. At an entry of 0.97, one full loss costs the profit of about 32 wins. This record is the proof, kept in public: 86 wins banked +$88.75 in total, and three losses of $50.00 each — a Charlotte FC exact-score, a Mavericks spread and a CS Cristal exact-score — took it all and more. Every win here is thin (about $0.50–$1.30 on a $50 position). Paper only, not financial advice.

The style belongs to the Polymarket wallet known as Sharky6999: buy outcomes the market prices as near-certain (0.95–0.99), in size, and hold to settlement. This sim copies his strategy, not his trades — no wallet of his is followed, and no signal here comes from him. Worth knowing before admiring the style: our reference check on Oct 26, 2026 put his all-time profit at +$947,070 on $254.4M of volume — about a 0.37% return on volume, an edge so thin it needs enormous size to matter — and his most recent month at that check was negative: -$32,118 on $7.03M of volume. Certainty-buying is a volume business with a razor edge, and lately the razor has cut the other way.

Certainty Reality mirrors the parent sim's entries with every fill priced about 15 seconds late. It is also down — -$7.45 (-0.75%) — though less than the parent, and the reason is boring: it started two days later, so it holds a different set of trades and happened to dodge two of the parent's three $50 losses (it did take the CS Cristal one). Settlements don't care about lag — the winner is the winner — so here the delay only moves entry prices, and on near-certainties a 15-second-later price is usually a hair worse: $15.50 of lag cost across its 54 closed trades, plus $0.44 still riding on its open five. The twin doesn't rescue the style; it just shows the style loses a little more slowly when it starts later.