METHOD

How we decide: edge, or noise

If you search enough market data, you will always find something that looks like a pattern, just by luck. So we never trust a pattern on the data it was found in.

1. Find. Scan the full history (every 5-minute bar of 43 FX and metals markets since July 2026) for something unusual, or take a popular trading belief and write it down as a testable claim.
2. Split. Decide the test first. Measure it on the older part of the data only.
3. Re-test. Run the identical test on newer data it has never seen, and across many markets, including the spread.
4. Verdict. If it holds up, it's an Edge. If it vanishes, it was Noise. Plain descriptions of what happened are a Statistic.

Edges can die

Markets change. Every surviving pattern is re-tested as new data arrives. When one fails, the Ledger records the date and the before-and-after numbers, and we post it. So far 6 of 21 tested patterns have survived (29%).

The data

Our own logger records 5-minute bid prices, spreads, swap rates and tick volume from a retail broker's price feed. Times are UK times. The broker's trading day starts at 10pm UK. We exclude the daily rollover hour and the Sunday open from anything about highs, lows or gaps, because spreads blow out then and create fake moves. We publish derived statistics only, never raw price feeds.

What this isn't

It isn't advice, signals or a trading system. A pattern that survives is a statistical tendency in past data, often too small to trade after costs. We say so when that's the case.