Home / Blog / Guides

Chart formations: sixteen patterns tested on 34 million days of data

GuideNumbers as of 2026-10-14

Every occurrence of sixteen classic chart patterns across 8,067 US stocks and up to 32 years of history — 12,149,503 events — with the outcome measured against the S&P 500 over the same days, and everything split into before-2018 and since. Five patterns repeat. Three flipped sign. The rest are indistinguishable from noise.

The method

Outcomes are market-adjusted — the stock minus SPY over the same days — and shown against the average stock for the era, because the average stock itself trailed the index badly after 2018. Days are clustered, so 3,000 stocks crashing on the same day count once, and outliers are capped so one 10× stock cannot own an average. A pattern is only called real if it holds with statistical significance in both halves of the sample.

The results

The full table of all sixteen is free to read. Members get the interactive page — every pattern's average path and full distribution of outcomes — and the ThinkorSwim study that plots the five survivors on any chart.

Every Friday the blog takes one formation and walks through its numbers.

Included in the $29/month membership

All seven tools, the full two-year feed, the JSON API and email alerts, refreshed every morning. A free quarterly sample of the dataset needs no account.

See plansFree sample

Questions this post answers

Do chart patterns work?

Of sixteen tested on 12 million events, five repeat in both eras — four negative (pops, vertical runs, upper band breaks, RSI above 70) and one positive (a pullback in an uptrend). Three flipped after 2018, one faded, seven are noise.

Does the golden cross work?

No reliable signal. Across 47,325 golden crosses the 21-day market-adjusted return was about −0.3% before 2018 and −0.2% since — neither significant. The death cross is the same story.

What does 'market-adjusted' mean here?

The stock's return minus SPY's return over the same trading days, so a pattern is not credited for a rising market.

Why split at 2018?

So every pattern is tested twice on non-overlapping data. A result that appears in one era and not the other is what noise looks like; only agreement in both counts.

Can I see the arrows on my own charts?

Yes. A free ThinkorSwim indicator plots one surviving formation; members get the full five-signal study with the measured outcome printed beside each arrow.

Related

Built from SEC EDGAR ownership filings. Only open-market purchases (transaction code P) and sales of $1M or more are counted; grants, option exercises and tax withholding are excluded because they run on schedules set years earlier. Historical description, not a prediction, and not investment advice.