Research

Measurements from my own trading, including the ones that failed. Method first, numbers second, and the criteria written down before the result exists.

Method

• Five gates I set before looking at a backtest result Cost fixed in advance, break-even win rate from measured R:R, same-bar rate, a minimum sample, and pre-registration — with the numbers from a setup of mine that failed anyway.
• How much history is enough, and why more tuning makes it worse Overfitting scales with the size of your search, not your effort. Walk-forward, plateaus versus cliffs, and the minimum span below which a run is not a result.
• What excursion distributions tell you that P&L does not MAE and MFE in R, logged per trade. How they separate an entry problem from an exit problem, and why aggregate P&L cannot.

Reading the market

• What a low-volume node actually claims Where the auction-theory reading comes from, what it asserts, and the gap between that and a trade — plus the two design choices that follow.
• The hours I turn the tool off Scheduled releases break the comparison order-flow reading depends on. Why the honest response is to stop measuring rather than widen a threshold.

Why this exists

Almost everything written about retail trading is a winning strategy. Negative results are rarer and more useful, so what gets published here includes the runs that did not work and the hypotheses I abandoned.

Everything is one instrument on one feed unless it says otherwise. The transferable part is the method, not the decimals.