Higher-timeframe agreement didn't improve our entries

On 150,008 setups, entries with all three trend reads agreeing reached 2R 32.87% of the time, and entries with none agreeing 32.60%. The chance line is 33.3%.

The claim

Traders are told to take trades only when the higher timeframe agrees: governing trend, internal trend and the current leg all pointing the same way. Full agreement is meant to be the high-probability version.

How we tested it

We scored 150,008 setups from our own continuation entry over two years on NQ (1, 3 and 5-minute), gold (3 and 5-minute) and Bitcoin (3 and 5-minute). Each setup was tagged with how many of three trend reads agreed with its direction: the governing trend, the internal trend and the direction of the governing leg. A win was price reaching 2R before the stop; the chance line at 2R is 33.3%, and the base rate across all setups was 32.53%.

What we found

All three agreeing: 32.87% (95% interval 32.3–33.5, n = 25,038). Two of three: 32.37% (n = 57,246). One of three: 32.52% (n = 50,986). None agreeing: 32.60% (n = 16,601). Every interval straddles the base rate, so full agreement was indistinguishable from full disagreement.

On Bitcoin it reversed. On the 5-minute chart, full agreement reached 2R 30.75% of the time and full disagreement 34.01%, against a chance line of 33.3%. Gold 5-minute was the one cell that looked positive, 35.11% with all three agreeing (n = 1,501), one cell in seven with a lower bound just above the base rate, which is what noise looks like.

A separate volatility-matched study of 133,056 non-overlapping price ranges pointed the same way. When all three trend tiers were bullish, the range resolved upward 49.89% of the time against a 51.89% base rate for all ranges (z = −3.59). That is a directional lean with no stop, target or costs, not a trade, but it does not support agreement as confirmation.

What to do with it

Agreement across timeframes describes the chart, but in this test it was not a reason to expect a trade to work. If you use it as a filter, compare your own results with and without it before relying on it.

Limits

This scored entry outcomes at 2R for one entry model; it did not test whether higher-timeframe structure predicts the direction of the next leg, which is a different question. The third read used the governing leg's direction as a stand-in for a signal that was not stored in the data.

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