Seven Trading Ideas, One Protocol: Six Without an Edge, One Without Enough Evidence

Seven popular intraday trading ideas, each rejected earlier in separate tests, were run through one protocol on CME gold futures from June 2010 to September 2026: the same data, costs, stop resolution on one-minute bars, a placebo with the same entry time on random days, a check of the mirror-image version of each pattern, and a year-by-year count. Six stay rejected. Fair value gaps, liquidity sweeps, fading streaks of candles, buying small dips for a quick profit, switching between trend and reversion by the Hurst exponent, and the Hikkake pattern all lose money after costs or do not beat the placebo; buying small dips wins 87% of the time and still loses before costs. The seventh is rejected for a different reason. A straddle around US data releases at 08:30 New York time, first discarded because of a clock error, earns +0.20 R per trade after costs at the correct time (t = 5.1, 14 of 17 years), nothing on weekdays without a release, and +0.51 R per trade on the broker's own ticks since June 2025. But it earned nothing for five consecutive years, 2016–2020, for reasons the data do not explain, and half of its profit comes from payroll days alone. An effect that can disappear for five years without explanation cannot yet be told apart from one that belongs to a regime, so it is rejected for lack of evidence and kept under observation rather than accepted. AI disclosure. The author used Claude (Anthropic), an AI assistant, to write the data-analysis code, todraft and edit the text, to assist with wording and structure, and as a consultative tool for discussingmethodological and technical questions. The research questions and the decisions on what to test and whatto accept are the author’s. The author reviewed, verified, and edited all AI-assisted content and takes fullresponsibility for the content of the paper, including its analysis, results, and conclusions.