Why do prices so often stall twice at the same level?

A double top is two rally peaks at roughly the same price with a dip between them — an "M". A double bottom is the mirror — a "W". They're the simplest reversal shapes, and the mechanism behind them is almost entirely about memory and resting orders.

Why the second visit matters

The first time a price is rejected at a level, participants learn something concrete: supply lives here. Three groups then act on that memory:

• Sellers who were too slow the first time place orders at that level, so they don't miss again.
• Buyers who bought near the first peak and then suffered decide to exit at breakeven when the price returns. This is loss aversion made visible in the order book — the same behaviour you met in Markets Foundations.
• Buyers generally get more cautious approaching a level that visibly rejected the price before.

So the second approach can face more resting supply than it otherwise would, purely because the first one happened. A double bottom is the same story upside down: buyers who missed the first bounce park bids at the level, and the second visit meets a thicker bid.

Note what this explanation does and doesn't do. It gives a genuine reason why levels have some memory. It gives no reason to expect the second rejection to hold — a third visit often clears the level precisely because two visits have already worn the resting orders down.

A worked example

Rounded and illustrative: a stock rallies to $78.00, falls to $70.20, rallies again to $77.60, then declines back through $70. Two peaks about 0.5% apart, a trough between them about 10% down.

Notice that the peaks are not identical — they almost never are. Real double tops usually differ by 1–3%. Traders who insist on exact matches simply find fewer patterns, not better ones, and traders who allow 5% find them everywhere. That dial — how much slack you permit — silently controls how many patterns exist in the world. You set it.

The uncomfortable fact about M and W

Of every shape in this course, the M and the W are the ones that random data produces most abundantly. Any wandering series that goes up, comes down and goes up again by a similar amount has drawn a W. In a few hundred sessions of noise there will be dozens.

This doesn't prove double tops carry no information. It proves that finding one tells you almost nothing by itself, because you'd have found one either way. The only way past that is counting — which is Unit 4's job.

One practical honesty note: whether two peaks look "equal" depends on your chart's vertical scale. Zoom out and a 3% difference vanishes into a flat ceiling; zoom in and the same two peaks look like clearly lower highs. Fix your scale first.

Try it now

1. Read the two marked peaks off the schematic above and calculate the % difference between them. Then decide what slack you would allow — and note that your answer, not the market's, is what decides whether this shape exists.
2. Now apply your own number where you cannot see the outcome first. In the five years below, find a stretch where price twice reached almost the same high within a few months. Drop a Level at the first peak and see whether the second one falls inside your slack.

3. Then check participation. The same five years of turnover are below: was the second approach busier than the first, or quieter? A retest on falling volume is the version the textbooks describe, and it is not the version you will always find.

4. Failure hunt: find a case where price stalled twice at a level and then, on the third approach, went straight through it. How would you have distinguished that setup from the one in step 2 while it was happening?