A trend's age doesn't tell you when it will end

557,155 trend episodes on six markets: how long trends lasted was set by the detector, not the market, and an old trend had as much life left as a new one.

The claim

Traders often trim or reverse because a trend has 'run too long', or hold because it is 'still young', as if trends had a natural lifespan.

How we tested it

We ran an exact copy of our three-tier trend detector (short-term, governing and long-term) over NQ, ES, Bitcoin, Ether, EURUSD and gold, 1.4 to 8.7 years each, on eight chart timeframes: 557,155 trend episodes. Every cell was re-run on a surrogate that keeps each market's real bar ranges and volatility but randomises direction, which asks how long a trend would last if direction were a coin flip.

What we found

Median trend life was nearly the same on every market: 12.9 governing candles for the governing tier and 49.5 for the long-term tier. The random-direction surrogate gave 12.0 and 50.0. Duration was a property of the detector and the volatility, not of trending behaviour in the market.

The chance of a trend ending did not rise with age. On Ether 1-minute (82,383 short-term episodes), trends that had reached the 25th, 50th and 75th percentile of age each had a median of 32 more bars to run. From three-quarters of its typical run onwards, the chance of a trend flipping within the next quarter-run was 20.0–23.1% on all three tiers, including trends more than twice the median age, and the surrogate reproduced the whole curve.

The one tier that outlived chance was the shortest: short-term trends lasted 6–23% longer than the surrogate on five of six markets, including 24-hour crypto, which is consistent with short-horizon momentum. That is a duration effect, not a demonstrated profit.

What to do with it

In this test a trend's age said nothing about how much longer it would run. Treat 'this trend has gone on too long' as a feeling to check, not a reason to act.

Limits

Trends were defined by our own detector, so other trend definitions may behave differently. Gold had only 1.4 years of data, and the long-term tier can't be measured on 4-hour charts because five years holds only a handful of its trends.

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