Buying near-certain favorites before resolution
We expected 90–98¢ favorites close to resolution to be slightly underpriced. With real money they won less often than their prices implied — and each loss cost about fifteen wins.
The hypothesisHolders of near-certain positions sell early to free capital, and gamblers prefer the lottery side — so favorites at 90–98¢ with hours to resolution trade a hair below fair. Buy them, collect the last cents, recycle capital fast.
The test
A family of live strategies bought high-probability favorites near expiry across sports, esports, and general markets, at small real size with fill confirmation, every entry recorded at its executable price. Variants tested tighter windows, whale confirmation, volume floors, and sustained-price filters. Only orders that actually filled count: of the 2,483 orders the main lane sent, 1,766 never filled. This is the same mechanism behind every "steel nerves" post ever written — see our claim-check on one — tested with our own money instead of a screenshot.
The result
When this entry was first published we reported a thin premium: early win rates ran slightly above what the prices implied. The full real-money record does not support that. Across three lanes, 780 real trades settled between May and July 2026. They won 704 times — 90.3% — while the prices we paid (93.8¢ on average) implied about 732 wins. The shortfall is about four standard errors, well outside luck. Net result: −$151.99 on $4,485 staked (−3.4%).
Every lane lost. The main lane: −$113.28 over 717 trades (−2.7%). The tightest-window variant, buying at 96¢ and up with minutes left: −$19.67 over 26 (−25.2%; 19 wins where the prices implied 25). The esports lane opened 24 wins and 3 losses and finished 31–6 at −$19.04 (−10.6%), after a single upset took four positions at once — match winner ×2, map winner ×2, one event wearing four market costumes.
The payoff shape does the rest. The average win paid $0.38 and the average loss cost $5.54, so one loss erased about fifteen wins. The worst afternoon came early: a favorite that held its price for hours was re-bought every scan cycle until 13 real positions sat on a single match, which lost alongside two other games for −$162 and tripped our daily loss breaker.
Conclusion
Downgraded from "edge confirmed" to negative. What looked like a premium in the first few hundred trades was not there in the full record: at the prices we could actually buy, these favorites lost slightly more often than their prices said they would, and the payoff shape — tiny wins, large losses — turns a small shortfall into a steady loss. Per-market caps and a daily loss breaker made the losses smaller; they did not make the strategy profitable. Buy what's nearly certain and collect the last cents is the pitch of every "steel nerves" post. On our money, it lost.