Support and resistance

Why levels might matter

Picture a stock that has turned down at $100 three times. As it nears $100 again, there is reason to think something happens there. Someone who bought at the top may want to sell and get out without a loss. Others have placed sell orders where they saw the price turn before. Round numbers attract too, and more people see the same level.

Support works the same way in the other direction: a level where buyers have shown up before.

When a level is passed, people often say it changes roles, so that old resistance becomes support.

Two real cases

A one-year high is the easiest level to find, because everyone sees the same number. The dashed line is the company's earlier highest close. Here is the latest time a stock among the 20 largest came back under such a ceiling without getting over, and the latest time one got over.

Examples like these can be found in both directions, so they prove nothing. So we counted all of them.

How often did the level hold?

We took the 20 largest US companies since 2000. Each time the price came back within 2% of a one-year high that was at least a month old, we counted whether any close in the following month was above it. One-year lows were counted the same way in the other direction. A company counts at most once a month.

The Ordinary day column answers what is normal: how often the same company, from any day at all, moved that far within a month.

The one-year highs were passed in 79% of cases, so they held in 21%. On an ordinary day the price got that far in 79%. The difference can't be told apart from chance, so in these numbers an old high didn't work as any kind of ceiling.

The one-year lows were passed in 73% of cases, against 68% on an ordinary day. That difference can't be told apart from chance either.

Does a new high mean anything?

We also counted what happened three months after a stock closed at a new one-year high or low, against the same company on an ordinary day. The method is the same as in the study of technical signals.

The table shows how often the stock rose. The difference in the average rise can't be told apart from chance either.

What it means

Levels are places where many people look, not walls. The test has limits: it covers today's largest companies, only one-year highs and lows, and the choices of 2% and one month are ours. Other levels may behave differently. Patterns built on levels, such as the double top and triangles, have lessons of their own.