Every trader has heard it: each time price tests a support or resistance level, the level gets weaker. It sounds right. So we checked.
We took 22 years of gold one-hour data, from 2004 to 2026, and found every swing high and low: 17,074 levels. Then we followed price forward and scored 19,302 tests. A test counted as a hold when price came back to the level and was pushed clearly away before it closed through.
What we found
Levels that had already held, held more often.
- First test of a level: 56% reacted.
- A level that had held once: 62.5%.
- Held three times: 66%. Five times: 70%.
Older levels held more often than fresh ones.
- Swings less than a day old: 54%.
- Swings more than four weeks old: 60%.
And the same ordering showed up in every era we checked: 2004-09, 2010-14, 2015-19 and 2020-26.
So on gold, at least, the old rule has it backwards. A level that keeps holding is earning respect, not running out of it.
What it doesn't mean
These are reaction rates, not trading profits. A few percentage points is a real edge in how you read a chart, not a system on its own. And it was measured on gold H1, so treat other markets as something to test. These rates rank levels against each other. A random point at a similar distance also gets "reactions" often, so this is about which levels to watch, not proof that levels beat chance.
The free tool
We turned the study into a free MetaTrader 5 indicator: Thanos AI Level Strength. It draws the unbroken levels on your chart and shades each one by strength, 1 to 5. A label like S4 6d 2 holds tells you at a glance: strength 4, six days old, held twice.
- Works on any symbol and timeframe.
- Uses closed bars only, so nothing repaints.
- No trades, no DLLs, no sign-up.
Free download: https://www.mql5.com/en/code/79577
Trading involves substantial risk of loss. This is an analysis tool, not financial advice. Past price behaviour does not guarantee future results.
