forex-scalping.com

Key additions to note:

⁠request.security()⁠: This line (⁠htfEma = request.security(...)⁠) calculates the 50 EMA on the 4-hour chart and maps that exact value down to your 15-minute or 5-minute candles.
The Bypass Toggle (⁠useFilter⁠): The variables ⁠htfBullish⁠ and ⁠htfBearish⁠ are designed so that if you uncheck "Enable HTF Filter" in the script settings, the variables default to ⁠true⁠. This bypasses the logic without requiring you to delete code, allowing you to seamlessly test the strategy with and without the trend filter.

Execution Gating: The execution signals at the very bottom now require both the structural reclaim (⁠isHighReclaim⁠ / ⁠isLowReclaim⁠) and the trend alignment (⁠htfBearish⁠ / ⁠htfBullish⁠) to evaluate as true before painting an arrow.

PTScalper wrote:Price action at a key level is a question; the time spent beyond that level is the answer.

That’s the useful filter — not “detector said sweep.”

I treat a spike through Asia low into a marked demand block as a candidate only if HTF is already aligned and the reclaim is fast. If price parks beyond the level, prints acceptance, and time-of-day is a genuine session drive (London continuation, NY news), I stop calling it a sweep and treat it as a break. Speed of reclaim beats the label.

Desk rule: fade only with invalidation beyond the sweep extreme and a time stop. A clean HTF-with-trend sweep at the London open is one trade. The same wick in the middle of a dead book is just a thin print. Indicator alerts don’t get a vote if those three — structure, clock, reclaim — disagree.

PTScalper wrote:To code this, you need a state machine... a true sweep is a sequence of events tied to a strict time limit.

Sweep detectors help when they encode sequence and time-in-zone, not when they spray arrows. Box Asia, watch for liquidity take, require reclaim inside a max bar count, optionally filter with HTF. That architecture matches discretionary rules I already run by hand on London open.

I still want human veto: spread gate, news blackout, correlation cap. Code that cannot see firm rules will happily signal through a challenge fail. Pine is fine for visualisation; live size still waits on execution reality.

Frankfurt desk: indicators highlight permission; they do not get the click. If reclaim is late beyond your time limit, it is a different trade — usually a pass.

Alerts on my desk are watchlist-only until a long sample of alerts-plus-manual veto looks clean. Auto-attached size through a firm news window is how automation donates evaluations.

Do you run sweep alerts as watchlist only, or auto-attached with micro size after a long demo sample?