Chart patterns and candlestick setups

Does the shape of a pullback predict the recovery?

Rejected Pre-registered and tested, August 2026

The strongest signal found was weaker than the average strongest signal that pure chance produces.

What the pattern claims

Two stocks can fall the same distance by completely different routes. One drifts down steadily over six weeks. The other holds up and then gaps down in a single session. The intuition — and it is a good one — is that the route says something about who was selling and what happens next.

We measured five things about the path from the peak to the bottom: how efficient the decline was, how much of it happened on its single worst day, whether the damage came early or late, where price closed within each day's range, and how volume tilted across the window.

How we tested it

4,602 usable events on S&P 500 companies. Each feature was split at its median within four separate groups, so that fast and slow declines, and deep and shallow ones, were compared only against their own kind.

Every study here was pre-registered: the exact rule, the pass and fail thresholds and the data window were written down and cryptographically fingerprinted before the test was run. That makes it impossible to move the goalposts after seeing the answer.

Results assume you buy at the next morning's open, not at the closing price that triggered the signal, and they include companies that were later delisted or went bust. Returns are measured against what the rest of the market did on the same days, so a rule that made money only because it fired on days everything rose scores zero here.

What happened

Nothing, and unusually clearly.

Across all ten comparisons, the largest t-statistic we observed was 1.24. To know whether that is impressive we ran 500 randomised versions of the same test, where any structure must be chance. In those, the largest t-statistic averaged 1.48, and one in twenty runs exceeded 2.11.

So the strongest result from the real data was weaker than what chance typically produces. No individual comparison came close. When we repeated the study on a wider set of events, four of the five features disagreed in sign with the first run.

What it means

This is the most valuable kind of null result, and it is worth explaining why.

The recurring problem with this family of studies is that the setups differ from each other in ways that have nothing to do with the pattern — different sizes of fall, different companies, different market conditions. Here the event was constructed so that every case ends in nearly the same place: the typical decline was 11.2%, with most falling between 10% and 18%. The endpoint is held nearly fixed, so the only thing varying is the route.

That design removes the usual excuse. A zero from a badly-designed test tells you little; a zero from a test built specifically to give the idea its best chance tells you a lot. This was the fourteenth consecutive rejection, and it closed price geometry as a research direction.

A rejection is not "this never works"

It means: tested on this universe, over this period, against a bar written down in advance, it did not clear the bar. Where a test lacked the power to decide either way we record it as undecided rather than rejected. The full research ledger has all 192 experiments and what the failures have in common.

Other patterns we tested

All 14 studies published so far