Chart Pattern Track Record

By pattern type

Share of settled signals in each bucket that beat SPY over the window. Above 50% means the pattern added something; below means holding the index would have done better. Sorted by sample size, because a bucket of 400 says more than a bucket of 6.

Recent signals

The last 120 signals at the 20-day window, newest first — including the ones that have not settled and the ones that lost.

By month

Every month on record, kept permanently. Nothing is removed from these pages after the fact.

• 2026-10
• 2026-09
• 2026-08
• 2026-07
• 2026-06
• 2026-05
• 2026-04
• 2026-03
• 2026-02

Common questions

Do chart patterns actually work?

On this data, mostly not in the direction people expect. Bullish trendline breaks have consistently underperformed SPY over 60 days, while bearish signals have held up better. The two bullish exceptions worth noting are the divergence buckets. The table above is the current state of that, updated daily — it is the answer, not a summary of one.

Why do some buckets show a win rate above 50% but a negative mean alpha?

Because the two measure different things. A bucket can win slightly more often than it loses while its losses are larger, which is exactly what a high win rate with negative mean alpha describes. Read them together; either alone is misleading.

Is anything removed from this page?

No. Every signal that was ever scored stays, including the ones that went against us, and the monthly archives are not edited after the fact. Signals still inside their window are shown as open rather than omitted.

Can I trade from this?

It is not built for that. A bucket win rate is a property of hundreds of past signals, not a probability for the next one, and nothing here accounts for costs or risk. Its purpose is to let you judge how much weight the rest of this site's signals deserve.