The claim

Double tops and bottoms, head and shoulders, triangles, wedges and rectangles are taught as reliable signals, with the break of the neckline as the entry.

How we tested it

We detected 11 patterns by fixed rules from confirmed swing points on seven instruments and three timeframes, two to five years each: 348,287 instances. Entry was the close through the neckline with a 2R target, where the chance line is 1/(1+2) = 33.3%. Each pattern was compared with two controls: a random bar, and an ordinary 20-bar breakout close with no pattern.

What we found

Hit rates at 2R: random bar 32.97% long and 32.13% short; plain breakout 33.57% and 32.92%; pattern breakout 33.77% and 32.62%. The chance line is 33.3%. Moving from a random bar to any breakout added 0.59 points on longs and 0.80 on shorts. Adding the pattern on top added 0.21 points on longs and took away 0.30 on shorts.

Against the breakout control, 1 of 11 patterns survived a multiple-testing correction, and it was negative: the symmetrical triangle, −0.75 points (z = −4.14). The best positive was the double bottom at +0.55 points, which did not survive (q = 0.28).

Scoring against the textbook 33.3% instead of a measured random entry made bullish patterns look 3.25 points better and bearish ones 5.24 points worse. That was the markets' upward drift over the period, not the patterns.

Even the breakout effect is about +0.02R per trade before costs, roughly one CME round trip (0.018R). The result echoes Chang and Osler (1995, Federal Reserve Bank of New York), who found a rule-based head-and-shoulders detector was outperformed by simple filter rules.

What to do with it

If you trade chart patterns, you are trading breakouts. Judge a pattern setup against an ordinary breakout with the same stop, not against the sense that the shape adds confirmation.

Limits

Patterns were defined by fixed rules, so a discretionary read of a shape was not tested. Results are hit rates at a fixed 2R before costs; other targets and exits were not tested.

Education only, not financial advice. These are measurements of past data, not a forecast or a signal.