A pattern name is not an edge
Markets give memorable names to visual shapes: breakouts, crossovers, reversals, consolidations. The names make patterns easy to discuss and dangerously easy to believe. A chart can always be annotated after the move. The real question is whether a rule defined before the move creates a repeatable difference in future returns after costs, across enough securities and enough market regimes.
I tested common technical patterns across Nifty 200 and Nifty 500 data from 2018 through 2026. The work covered nineteen patterns and seventy-six pattern-horizon combinations. The goal was not to prove technical analysis wrong. It was to force each idea into a definition precise enough that code could disagree with the story.
Research choices create the result
The obvious calculation is rarely the dangerous part. Survivorship bias, look-ahead leakage, missing data, overlapping signals, and flexible thresholds can manufacture an attractive backtest. A pattern definition that changes after seeing results is no longer a test; it is a description of the sample. I used consistent rules, multiple horizons, and explicit evidence thresholds so a single lucky period could not carry the conclusion.
It also matters what the pattern is compared with. Positive average returns do not imply an edge if the broader market produced similar returns over the same horizon. The useful quantity is conditional expectancy relative to a defensible baseline, paired with sample size and stability across time. A chart of the best examples is not evidence.
A negative result is a successful filter
None of the tested patterns survived the evidence threshold. That is not a failed project. It is a saved future cost. The research replaced vague confidence with a concrete reason not to automate those signals or build a product around them. Negative results are especially valuable in markets because there is endless pressure to keep searching until a backtest looks exciting.
The larger lesson applies beyond trading. Treat product claims as hypotheses, define what would change your mind, and build the smallest reliable measurement. The point of research is not to produce a green number. It is to update a decision. Sometimes the most profitable output of a system is a disciplined no.