Every PSX course teaches this rule: find the support level, buy when price comes down to it. We measured it.
A "support test" here is defined mechanically, with no judgement: the stock is trading within 2% of its own lowest low of the last 20 sessions. Then we asked our standard question, did it touch +20% within the next 30 sessions?
| sample | touched +20% | vs the market | |
|---|---|---|---|
| At the 20-day low ("support") | 6,333 | 15.1% | 0.86× |
| Same rule, held-out years | 4,456 | 25.2% | 0.87× |
Both numbers are below 1.00×. Buying a stock because it has fallen to support performed worse than picking at random from the same universe, in the years we designed on, and again in the years we had never looked at.
This is not a small effect and it is not noise. It is the single most-taught entry rule in this market, and on 10,789 measured tests it subtracts value.
Why it fails
Support is a line drawn on a chart after the price bounced there before. It has no memory and no obligation. What actually happens on PSX is that a stock making 20-day lows is usually a stock where someone informed is selling, and "support" is simply the level at which they have not yet finished.
The mirror image is the part worth keeping. A stock that has fallen more than 50% from its 52-week high touched +20% in 30.5% of 3,459 cases (1.74×) and 36.2% of 1,896 held-out cases (1.26×). Deep damage pays. Mild damage does not.
The useful distinction is not "has it fallen to support" but "has it fallen far enough that the sellers are finished". 10-25% below the 52-week high measured 0.79× and 0.82×, the worst band on the board. Half-price measured 1.74× and 1.26×.
Do this now, practical activity
- Open the Screener.
- Find any stock currently sitting near its 20-day low.
- Now open SuperCharts for that same stock and look left, find the last three times it touched a similar low.
- For each, note what happened over the following month. Write the three outcomes down.
What you are training: the habit of checking a rule against the chart in front of you instead of accepting it. Three observations will not prove anything, that is the point. Our table above used 10,789.
Measured: stocks at their 20-day low touched +20% within 30 sessions in 15.1% of 6,333 cases (0.86× the base rate) and 25.2% of 4,456 held-out cases (0.87×). Both below random.