The staircase: buying underdogs as they climb
A 27-million-trade backtest said underdogs that climb keep climbing — +15% per cycle. At real order books the same rule lost 13.5%. The difference was exactly the spread.
The hypothesisA 30–40¢ underdog is a bad blind buy — but conditional on climbing to 50¢, then 60¢, its true win probability beats its price at every rung. Add on the way up (front-loaded, since the edge decays near the top) and momentum pays.
The test
The backtest, built on 27 million historical trades, was genuinely strong: at every rung, climbers' subsequent win rates exceeded the rung price. The live probe then executed the exact rule with one honesty upgrade — every rung purchase logged at the live executable ask at the moment the rung was crossed, not the last-trade price the backtest had assumed. 188 ladders settled at resolution.
The result
−13.5% ROI at real fills (−$93 on $687 of paper basis), against the backtest's +15%. The reconciliation is almost poetic: a market crossing 50¢ on the way up doesn't offer you 50¢ — it offers 52–54¢, because the climb itself has pulled the ask, and you are buying into momentum everyone can see. Roughly 2–4¢ of slippage per rung, three rungs per ladder, on an edge of the same magnitude. The entire backtested profit was the spread the backtest didn't pay.
Conclusion
Momentum edges measured at trade prices are systematically inflated by exactly the cost of chasing. This is the cleanest before/after pair we own — same rule, same markets, same period, midpoint fills vs real fills, thirty points of ROI apart — and it is why nothing in this operation is trusted until it has been priced at the ask. When someone publishes a momentum backtest, the first question is not whether the pattern is real; it's who pays for the climb.