Skip to content
PatternGrade
How it worksPatternsFeaturesComparePricing
Sign inJoin waitlist
How it works

Methodology

PatternGrade finds shapes in price data and says how closely each one matches its textbook definition. This page explains exactly how that is done — and what the product deliberately does not tell you.

What a scanner does

Each of the 75 scanners encodes one pattern as a set of geometric rules — proportions, durations, the direction a trendline has to run, how volume has to behave. After every close, each scanner walks every liquid US stock and tests those rules against the bars.

The rules are pass or fail. A window either satisfies all of them or it is not that pattern, and nothing is returned. There is no low-confidence consolation match. This is why a scan produces a short list rather than a ranked version of the whole market.

What a grade measures

A grade answers one question: how closely does this price action match the textbook definition of the pattern? Nothing more. A+ is the cleanest structural match, A is well formed with a measurement or two near a tolerance, and B is recognisably the pattern with proportions that stretch the definition.

The score behind the grade is arithmetic on the geometry. For a cup and handle it is how well the cup fits a parabola, multiplied by how shallow the handle is, multiplied by how level the two rims are; the result is cut at 0.80 for A+ and 0.70 for A. Every scanner works this way, with the measurements that suit its own shape.

Two consequences worth being explicit about. The grading is deterministic — the same chart always produces the same grade, with no model, no training and no randomness, so a result you disagree with is a rule you can go and read. And the grading is descriptive — it ranks how well a drawing matches a definition, not how likely anything is to happen next.

What the levels on the chart mean

  • Breakout level. The price the pattern has to cross for the formation to complete — the rim, the neckline, the pivot, depending on the pattern. A setup is marked forming before that level is crossed and confirmed after.
  • Measured move.Where the pattern's own geometry projects to: the height of the formation extended from the breakout level. This is the classical measure rule, a property of the drawing. It is not a price forecast, a target we expect to be reached, or a recommendation to hold until it is.
  • Invalidation. The level at which the price action stops matching the pattern — below the handle low, outside the channel, through the neckline. Past it, the shape on the chart is no longer the shape the scanner found.

What we do not publish, and why

PatternGrade publishes no hit rates, no win rates, no success rates and no probabilities. Not per pattern, not per grade, not per stock. This is deliberate, and it is worth explaining rather than leaving you to notice the absence.

The grade does not predict outcomes, and we can show that. We measure how our own signals resolve. Across several thousand resolved signals, A+ setups have not performed measurably better than A setups — the gap is well inside what chance produces. A grade that sorts charts by structural quality is useful; the same grade sold as a ranking of which trades will work would be a claim our own data does not support.

A percentage next to a stock is read as a forecast. "This grade has historically reached its objective 34% of the time" is a statement about a sample. Printed beside a named company on the day it fires, it is read as that company's odds — and no caption underneath changes the reading. The safest way not to make that claim is not to have the number on the page.

Backtested rates are weaker evidence than they look. They are gross of commission, slippage and the spread. They cover one market era. Companies delisted before our price history begins cannot be recovered. Any of those alone is enough to make a confident-looking percentage misleading.

We still run the measurements. They are how we decide which scanners are worth keeping and which need reworking. They are not a product claim, so they are not on the site.

The limits of the tool

  • End of day only. Everything is computed on settled daily bars after the close. Nothing here is real time.
  • Patterns are interpretations. Technical analysis is a way of describing price history, not a law governing it. Two careful analysts can look at the same base and disagree, which is part of why every rule we apply is written down.
  • No fundamentals, no news, no context. A scanner sees bars. It does not know about an earnings release, a restructuring, a lawsuit or a halt.
  • Identical for everyone. Every user sees exactly the same output for the same stock. Nothing is tailored to your account, holdings, goals or circumstances — this is a reference tool, not a service that considers your situation.

Not investment advice

PatternGrade identifies chart patterns. It does not recommend securities, predict prices, tell you what to buy or sell, or take your circumstances into account. Nothing on this site is investment, financial, legal or tax advice, and nothing on it is an offer or solicitation to trade anything. See the full disclaimer and terms.

Questions

If a rule on this page does not match what you see in the product, tell us: support@patterngrade.com. We would rather correct it than defend it.

PatternGrade

PatternGrade is a technical pattern scanner for ~4,500 US stocks. The information on this site is for educational and analytical purposes only and is not investment advice. Always trade at your own risk.

Product
How it worksFeaturesPattern libraryComparePricingSign in
Company
AboutCareersContact
Legal
MethodologyTermsPrivacyDisclaimer
© 2026 PatternGrade. All rights reserved.
Terms, privacy and risk notice in:
EnglishFrançaisEspañol