How Artemis Pro Thinks

Filters, Not Textbook Matching

Naive pattern matching does not survive contact with chart timeframes from 1 to 5 minutes. With loose definitions, dozens of bars per session qualify as "engulfing" or "star," and almost all of them are noise. Artemis Pro therefore treats the textbook shape as necessary but not sufficient. Depending on the pattern, a candidate must also pass:

• Body dominance. The candles that carry the pattern need real bodies. An engulfing leg must be at least 50 percent body, and each candle in a Three White Soldiers or Three Black Crows sequence must be at least 50 percent body.
• ATR scale. The move must be large relative to current volatility. The engulfing candle in a Three Outside pattern must span at least 0.5 ATR and be 1.5 times the prior body; a soldiers or crows sequence must cover at least 1.0 ATR in total.
• Wick structure. Star patterns require a pronounced rejection wick on the middle candle — a small body alone is not evidence of rejection. Soldiers and crows require small counter-wicks (at most 25 percent) on all three candles.
• Symmetry. Strict and relaxed star detections reject lopsided formations: the first and third candles must be of comparable size (default maximum asymmetry 0.30), unless the third candle is outright stronger than the first — a power move earns an exemption. The classic small-bodied-star detection is exempt from this filter.
• Trend context. Three White Soldiers only count when the trend before the sequence was bearish, and Three Black Crows only when it was bullish. Three green candles inside an existing uptrend are continuation, not a reversal, so they do not fire.

The result is that Artemis Pro prints fewer signals than a textbook scanner would — deliberately. What survives the filters is a shortlist worth your attention, not a catalog of everything that vaguely resembles a pattern.

One Bar, One Verdict: The Priority Chain

Several patterns can be true on the same bar; a Three-Line Strike, for example, often satisfies engulfing conditions as well. Artemis Pro resolves every bar to exactly one color and at most one label through a fixed priority chain:

The ranking prefers the rarer, more structurally demanding pattern: a Three-Line Strike says more than the engulfing conditions it may also satisfy. Only the highest-priority pattern per bar prints, so you never have to arbitrate between two labels on one candle, and the chart stays readable at speed. If no pattern fires, the candle simply wears its trend color. The full color and label glossary is in the Signal Reference.

Prominence Encodes Context

Pattern labels come in two sizes, and the size is information. A label that prints within about 3 bars of a liquidity grab or level break, or on a bar that confirms a setup, is drawn enlarged; a standalone pattern prints small.

The reasoning: a reversal pattern into freshly swept liquidity is a different event from the same pattern in the middle of nowhere. Stops were just taken, the sweep failed, and now a filtered reversal signal confirms in the same window — that confluence is what the large label flags, and it is the only condition the script escalates into a Setup Long/Short. Everything else remains a standalone signal. See Liquidity Grabs for the mechanics.

What Repaints And What Does Not

Honesty about timing matters more than any feature, so here is exactly where the lag and the movement are.

• Patterns evaluate on bar close. Once a bar has closed, its color and label never change. While a bar is still forming, its state is provisional — treat any signal as final only after the close.
• Swing lines lag by design. A pivot needs 3 bars of confirmation (the Bars Right default) before its line appears, so lines are born 3 bars after the actual swing. Grabs are consequently only detected on levels that already existed.
• Some HTF levels develop live. PWH and PWL come from the previous closed week, and the current day open is fixed when the day begins; none of them moves intraday. CWH and CWL track the developing weekly extremes: they update whenever a new extreme forms, and on historical bars they are drawn at the week's final extreme from its start. Treat them as context, not as backtestable levels.
• HTF mini-candles redraw on the last bar. The redraw is cosmetic, but the newest mini-candle in each group represents a higher-timeframe candle that is still open — its color and signal are provisional until that candle closes on its own timeframe.