A simple MTF confluence checklist I use before any XAUUSD scalp
Posted: Sat Sep 05, 2026 12:36 pm
Before any XAUUSD scalp I run a short multi-timeframe confluence check. If the layers do not line up, I do not take the trade. No indicator soup, no “almost aligned,” no forcing a trigger because the session is liquid.
The stack I use is fixed:
1. H1 bias — direction and context only. I mark the recent swing structure, whether price is accepting above/below a clear level, and whether the last few H1 candles show continuation or exhaustion. Bias is binary for the session idea: long-leaning, short-leaning, or no trade. If H1 is messy or mid-range with no clear acceptance, M1 noise is not a setup.
2. M15 structure — the map for where a scalp is allowed. I want a clear break-and-retest, liquidity sweep and reclaim, or a failed break of a level that already mattered on H1. M15 tells me the location. Without a location, a pretty M1 candle is just a pretty candle.
3. M1/M5 trigger — timing only after H1 and M15 agree. Entry is a specific rejection, reclaim, or continuation print at the M15 level, with a stop beyond the invalidation of that micro structure. If the trigger appears but H1/M15 disagree, I pass.
Checklist I tick mentally (and often in the journal) before clicking:
- Session permission: London/NY overlap or my defined window, not random hours.
- Spread within my max gate for gold.
- No high-impact USD release inside my blackout window unless the setup is on my rare pre-defined exception list (usually empty).
- H1 bias stated in one sentence.
- M15 level and invalidation marked.
- M1/M5 trigger matches the bias, not fighting it.
- Risk sized from stop distance, not from “how sure I feel.”
What I refuse:
- Taking M1 reversals against a clear H1 trend just because a wick looked dramatic.
- Stacking RSI, MACD, and three moving averages to justify a trade the structure already rejected.
- Entering because “gold is moving” during overlap. Movement is not confluence.
- Lowering the checklist mid-session after a win or a loss. Confluence rules are decided when I am calm.
A practical example of refusal: H1 bullish acceptance above a reclaimed level, M15 still compressing under a local high with no sweep/reclaim yet, M1 printing a sharp long wick. That is not an A+ long. It is a lower-timeframe tease. I wait for M15 to confirm location or I stand aside.
Using a single timeframe in isolation means you are making decisions with a genuinely incomplete picture. H1 without a trigger is analysis. M1 without H1/M15 is gambling with a tight stop. The checklist exists so I do not confuse activity with process.
If your XAUUSD scalps feel random, strip the chart back to this three-layer gate for a week. Count how many impulses you skipped because alignment failed. Those skipped trades are often the edge — not the ones you forced when only one timeframe was shouting.
The stack I use is fixed:
1. H1 bias — direction and context only. I mark the recent swing structure, whether price is accepting above/below a clear level, and whether the last few H1 candles show continuation or exhaustion. Bias is binary for the session idea: long-leaning, short-leaning, or no trade. If H1 is messy or mid-range with no clear acceptance, M1 noise is not a setup.
2. M15 structure — the map for where a scalp is allowed. I want a clear break-and-retest, liquidity sweep and reclaim, or a failed break of a level that already mattered on H1. M15 tells me the location. Without a location, a pretty M1 candle is just a pretty candle.
3. M1/M5 trigger — timing only after H1 and M15 agree. Entry is a specific rejection, reclaim, or continuation print at the M15 level, with a stop beyond the invalidation of that micro structure. If the trigger appears but H1/M15 disagree, I pass.
Checklist I tick mentally (and often in the journal) before clicking:
- Session permission: London/NY overlap or my defined window, not random hours.
- Spread within my max gate for gold.
- No high-impact USD release inside my blackout window unless the setup is on my rare pre-defined exception list (usually empty).
- H1 bias stated in one sentence.
- M15 level and invalidation marked.
- M1/M5 trigger matches the bias, not fighting it.
- Risk sized from stop distance, not from “how sure I feel.”
What I refuse:
- Taking M1 reversals against a clear H1 trend just because a wick looked dramatic.
- Stacking RSI, MACD, and three moving averages to justify a trade the structure already rejected.
- Entering because “gold is moving” during overlap. Movement is not confluence.
- Lowering the checklist mid-session after a win or a loss. Confluence rules are decided when I am calm.
A practical example of refusal: H1 bullish acceptance above a reclaimed level, M15 still compressing under a local high with no sweep/reclaim yet, M1 printing a sharp long wick. That is not an A+ long. It is a lower-timeframe tease. I wait for M15 to confirm location or I stand aside.
Using a single timeframe in isolation means you are making decisions with a genuinely incomplete picture. H1 without a trigger is analysis. M1 without H1/M15 is gambling with a tight stop. The checklist exists so I do not confuse activity with process.
If your XAUUSD scalps feel random, strip the chart back to this three-layer gate for a week. Count how many impulses you skipped because alignment failed. Those skipped trades are often the edge — not the ones you forced when only one timeframe was shouting.