Explainer5 min read
A chart pattern is a description, not a promise
In 2025 Tesla drew a textbook double bottom and kept climbing. In 2023 Nvidia drew a textbook head and shoulders, completed it, and a month later closed above the head. On the day each pattern was confirmed, the two looked almost the same.
Two of the best-known shapes in technical analysis had their textbook moment two years apart, and they ended in opposite ways. Tesla's double bottom in the spring of 2025 did what the pattern is meant to announce: the fall stopped and the stock climbed. Nvidia's head and shoulders in the autumn of 2023 did everything the pattern asks for, was confirmed, and failed within a month.
I don't think either case proves anything about the patterns. What they show together is narrower, and I think more useful. A chart pattern is a name for a shape and a rule for when the shape counts. It describes what has happened. It does not carry odds about what happens next, and the textbook that defines these two does not claim it does.
What the two charts did
The rules here are those of StockCharts' ChartSchool. The prices are Yahoo Finance's daily candles.
A double bottom needs a downtrend first. Tesla had one: from a high of $488.54 on December 18, 2024, it fell to $217.02 on March 11, 2025, 55.6% lower. It bounced to $291.85 on March 27 and fell again, to $214.25 on April 7, 1.3% under the first low. ChartSchool treats a second trough within about 3% of the first as valid, and says the pattern is complete when the price breaks the highest point between the two troughs. Tesla closed at $292.03 on April 29, above that $291.85. Four weeks later, on May 27, it closed at $362.89, 24.3% higher. Not in a straight line: on June 5 it fell 14.3% in a single session, from $332.05 to $284.70.
A head and shoulders top is the mirror image. It needs an uptrend, then three peaks with the middle one highest, and a neckline joining the two lows between them. Nvidia's prices here are adjusted for its ten-for-one split of June 2024, so they are a tenth of what the shares traded at in 2023. From $37.36 on June 7, 2023, the stock rose 28.7% to a left shoulder of $48.09 on July 14. The head came at $50.27 on August 24, the right shoulder at $47.61 on October 12, 1% under the left. The neckline ran through the lows of $40.31 on August 14 and $40.98 on September 21. On October 20 Nvidia closed at $41.39, under the line, which stood at about $41.50 that day. By ChartSchool's rule, the pattern was now complete. The stock went as low as $39.23 on October 31. On November 20 it closed at $50.41, above the head and 21.8% above its close on the day of the break.
On the day, they looked alike
What I find more telling than the outcomes is how thin both confirmations were. Tesla's close cleared its line by 18 cents, and the next day it closed back below it. Nvidia's close broke its line by about 11 cents, 0.3%, and the next session it closed back above. Both breaks came on lighter trading than usual: Tesla's volume that day was about 13% under its average of the previous 50 sessions, Nvidia's about 8% under.
ChartSchool asks for more than a close past the line. For the head and shoulders, it wants the break to be convincing, ideally with expanding volume. For the double bottom, it wants more volume or a faster climb at the breakout. Measured that way, neither day was a strong signal. One worked anyway, and one did not.
The patterns' own measuring rule, a sketch of how far a move might run, did no better at telling them apart. For the double bottom, ChartSchool adds the height of the formation to the breakout level: for Tesla, about $369. It came within two dollars of that by May 29. For the head and shoulders, the rule subtracts the height of the head over the neckline from the line: for Nvidia, about $32. It got no lower than $39.23 before it turned. ChartSchool itself says to treat such figures as a rough guide.
The case for patterns
The strongest objection is that one failure does not refute a pattern, just as one success does not prove it. A pattern is supposed to work more often than not, not every time. Qualcomm drew a head and shoulders in 2024: after it closed under its neckline at $186.21 on July 19, it traded as low as $151.39 eleven sessions later, 18.7% down. Coinbase drew one in late 2021: after it broke on January 7, 2022, at $232.33, its low on March 15 was $150.12, 35.4% down.
I accept that. But it is an argument about frequencies, and frequencies are what the textbook does not give. ChartSchool calls the head and shoulders "one of the most common reversal formations" and offers no success rate for it or for the double bottom. On the double bottom, it says that even once support has held, only the possibility of the pattern exists. Possibility is the honest word for a single chart. And the examples people remember are chosen afterwards: a shape that worked becomes the illustration, while a shape that failed tends to be renamed or forgotten.
So this is how I read them. A pattern tells you that a market has done something specific: fallen twice to the same level and held, or failed to make a new high after a peak. That is information about the past. What came next for Tesla and for Nvidia was decided by things the shape did not contain.
What would change my mind
A count done in advance, with fixed rules, of how often confirmed breaks follow through, with the failures counted beside the successes and both set against what the same stocks did when no pattern was there. If confirmed patterns beat that baseline by a wide margin, I would give them more weight as forecasts than I do. Until someone shows that, I read a pattern as a description, not a promise.
Comments