Apply one rule identically, at scale
Check far more instruments, far faster, than a person scanning charts by eye.
The Boundaries of Automated Pattern Recognition
A rule-based scanner only ever tells you what it was configured to look for. Learn where automated shape-matching genuinely helps, where curve-fitting risk creeps in, and where human contextual judgment still has to do the rest of the work.
- Explain what a rule-based pattern scanner can and cannot do.
- Recognize where context a scanner can't see still matters.
- Understand curve-fitting risk in a pattern-recognition rule.
- Avoid treating an automated match as validated truth.
- Build a habit of auditing an automated flag before acting on it.
A Scanner Applies a Fixed Rule Consistently — Nothing More, Nothing Less
An automated pattern scanner encodes a specific geometric rule — for example, how symmetrical two shoulders need to be, or how flat an upper boundary must stay — and checks every instrument against it the same way, every time. Consistency and speed are its real strengths.
Flag exactly what it was told to flag
A match confirms the configured criteria were met — nothing broader than that.
Judge context outside its rule
News, liquidity regime, or cross-asset conditions aren't part of a shape-matching rule.
Head & shouldersNeckline ~96.0
The scanner encodes this geometry into rules and repeats it across thousands of symbols. It sees the shape — not why the price moved.
The Same Data, a Different Number of Matches
Adjust a scanner's tolerance setting on a fixed set of candidate shapes and watch the match count change — without a single underlying price ever moving.
Rule Scanner Sensitivity LabThe same fixed set of candidate shapes — observe how scanner tolerance changes match counts across asymmetry tiers
Matches: 2 / 6
Gap 1%Matched
Gap 4%Matched
Gap 7%Ignored
Gap 10%Ignored
Gap 13%Ignored
Gap 16%Ignored
2 of 6 shapes match at this tolerance — the underlying price data never changed; only the rule strictness did.
An Automated Flag Confirms the Rule Fired — Not That the Setup Is Sound
Whether a match came from a scanner or from your own eye, it carries the same evidentiary weight: it's a candidate. Automation changes how fast and how broadly you can find candidates — it doesn't change how much scrutiny each one still needs before you'd act on it.
A Rule Repeatedly Tuned to Fit One Stretch of History May Only Describe That Stretch
If a scanning rule gets adjusted over and over until it produces an impressive result on last year's data, that process risks fitting the rule to that specific data's noise rather than to something likely to recur. This is the same caution that applies to any backtest, applied here to pattern-recognition parameters specifically.
Overfit ruleHugs every dip
The overfit boundary explains every historical wiggle — that is curve-fitting, not validation. The straight rule is cruder but far more likely to generalize to unseen data.
A Shape-Matching Rule Has No Way to Know the World Around It Changed
A textbook shape flagged moments after a surprise headline is in a different context than the same shape on a quiet day.
The same rule's match means something different in a thin, illiquid moment than in a normal session.
A pattern in one instrument can be shaped by moves in a related market a single-instrument scanner never looks at.
What Does an Automated Match Actually Tell You in Each Case?
Pick a case and judge what an automated flag does and doesn't confirm — including situations a rule-based scanner has no way to account for.
Automation Trust AuditPick a case and judge what an automated match actually tells you
Four Checks Before You Act on an Automated Flag
01
Rule
What exact criteria triggered this match, and how strict are they?
02
Sensitivity
Would a slightly different setting on the same data still flag it?
03
Context
Is there a news event, liquidity condition, or related market the rule can't see?
04
Manual check
Have you applied the same evidence-and-invalidation process you'd use for a pattern found by eye?
- 1Freeze the ruleWrite the input fields, thresholds and known exclusions before reviewing results.
- 2Inspect the missesCheck both matches and non-matches for changed data, ambiguity or a rule that is too narrow.
- 3Version the revisionRecord why a rule changed and test it separately rather than silently replacing the old rule.
Automate Repetition Without Automating Judgment
Consistency and speed are its strengths — judgment beyond its rule is not.
Automated or manual, a flagged pattern still needs the same scrutiny before it means anything.
News, liquidity, and related markets can all change what a matched shape actually means.
Put Your Understanding to the Test
Submit your answers to see detailed explanations.
Question 1 of 3
A rule-based scanner flags a chart pattern. What does that flag, by itself, actually confirm?
Question 2 of 3
A scanning rule was adjusted repeatedly until it produced a strong result on one year of historical data. What risk does this raise?
Question 3 of 3
A scanner flags a textbook continuation shape two minutes after a surprise headline hits that instrument. Three readings get offered. Which one holds up?
Stuck? Ask Mira to Break It Down
Describe the rule, the match, and any context around it, and Mira can help you check for sensitivity to parameters and missing context — it won't tell you a flagged match is validated.
Checking sign-in status...