The claim

Session models teach that earlier sessions set up the next one: London reverses Asia, New York reverses a London sweep, a manipulation is followed by a reversal and an accumulation by a continuation.

How we tested it

We used five years of NQ and gold, about 1,380 days each, and the outcome was whether the target session closed up. Each of 22 rules was scored against the same prediction made on every day, not against a coin, with a Sidak correction for 22 tests (|z| > 3.04). A separate two-year study of 516 NQ days, with sessions on daylight-saving-aware New York time, tested a popular five-profile day model and its 50% level.

What we found

Zero of 22 rules passed the correction, and none reached |z| > 2, where about one would be expected by chance alone. The best was 1.49, and lifts ranged from −4.2 to +2.2 points.

Scoring against 50% first produced a false winner (z = −3.12) that passed the correction. It was the session's own drift: 'always short' already beats a coin on a market whose session closes down more often. Scored against the same call made on every day, it disappeared.

The same rule pointed opposite ways on the two markets. 'London reverses Asia's direction' was right 53.36% of the time on gold and 49.38% on NQ, on opposite sides of a 50% coin. Real structure would point the same way on both.

The five-profile model, labelled at 08:00 New York time, called the day's direction on 50.7% of 351 days, against a coin's 50%. Its 50% level made a −2.0 point difference (z = −0.64) once the window used to judge the level was kept apart from the window used to judge the outcome; letting them overlap had produced a false +57.6 points.

What to do with it

Sessions are useful for knowing when price tends to move and where the day's reference levels sit. In these tests they did not say which way to trade, so treat a directional session story as a hypothesis to check, not a bias.

Limits

The outcome was a simple session close up or down, not a trade with a stop and target. The day-profile study covered two years of NQ CFD prices only.

Education only, not financial advice. These are measurements of past data, not a forecast or a signal.