StockSetupLists

Methodology

What the quality score is, and isn't.

The score is a 1-to-5 ranking of how cleanly a detector saw its target pattern. It is not a forecast. The page below explains, in detail, what that means.

A 95-quality Breakout can fail. A 60-quality Mean Reversion can work. The score is a screening aid, not a forecast.

01 · What it is

The quality score answers one question: how structurally clean is this pattern, relative to its detector's ideal? That's it. A high score means the pattern is well-formed. A low score means the detector saw the shape but the read was weak.

Scores are computed pattern-by-pattern. They do not look across the universe to answer "is this the best Breakout today?" — they answer "is this a clean Breakout?" Two stocks can both score five; the list is ranked by score, then by structural sub-features, then deterministically by ticker.

02 · What it isn't

The quality score is not:

  • —
    A probability of success
    It does not estimate the odds that the pattern will work.
  • —
    A forecast of return
    It does not estimate how far the move will travel.
  • —
    A confidence interval
    It is not a statistical confidence value — it is an ordinal rank.
  • —
    A trade signal
    A score above any threshold is not a "buy" or a "sell."
  • —
    A comparison across types
    A 5-quality Breakout and a 5-quality Mean Reversion are not equivalent.
  • —
    A guarantee of cleanliness
    Patterns evolve. A clean read today can deteriorate tomorrow.

03 · How to read a score

Use it as a screen, not a verdict. Start with the quality floor that matches your tolerance for noise; three is a sensible default. Look at the chart. Form your own view. The score did its job by surfacing the candidate.

Detected
Borderline
Workable
Strong
Textbook
A score is not a target
Two patterns at the same score are not interchangeable. A Breakout and a Mean Reversion with identical scores are two different reads on two different things. Compare scores inside a setup type; don't average them across.

04 · Worked examples

Below are two anonymized specimens with placeholder tickers. The charts are schematic; real lists carry real candles.

XXXX Breakout
XXX Mean Reversion

Left: a textbook Breakout above multi-month resistance with confirming volume. Right: a borderline Mean Reversion — stretched far from the trend but with weaker structural footing. Same product, two very different reads.

05 · What we publish, what we don't

We publish

  • ✓ Setup names and definitions
  • ✓ Quality scores and milestone badges
  • ✓ The macro context block, daily
  • ✓ A generation timestamp on every artifact
  • ✓ This methodology page in full

We don't publish

  • — Detector logic, code, or thresholds
  • — Scoring weights or hyperparameters
  • — The AI digest's system prompt
  • — Backtests as marketing material
  • — "Best" or "worst" daily picks
Versioning
When detectors or scoring change in a way that would alter past output, we increment the methodology version and note the change in the release log. We do not silently re-rate historical setups.