The continuation stack: Ichimoku, MACD, and the momentum filter.

Ichimoku

Structure, with cloud thickness as the strength read.

The Ichimoku component runs the classic configuration — Tenkan-sen at 9, Kijun-sen at 26, Senkou Span B at 52 — and the manual's justification for keeping them is that they're anchored in market cycles rather than curve-fitted: roughly one-and-a-half, three, and six weeks on a daily chart. That balance of short-term reaction at 9 with trend confirmation at 26 and 52 remains appropriate for intraday continuation work. The component's most useful output for this strategy isn't the cross but the cloud: thickness serves as the trend-strength filter, with a wide cloud indicating conditions more favorable to continuation.

MACD

Acceleration, at settings deliberately left standard.

MACD runs 12/26/9 — fast EMA, slow EMA, signal — and the reasoning given is worth quoting in spirit: the standard settings are already optimal for spotting trend acceleration without overfitting. That last clause carries the doctrine. Tuned MACD parameters can always be found that would have performed better on any given historical sample, and they generalize poorly precisely because they were selected to fit that sample. Leaving the canonical settings alone is a deliberate refusal of an optimization that mostly produces confidence rather than edge — the same instinct that keeps the regime engine on fixed thresholds and the coefficient ladder locked.

The momentum filter

RSI as baseline, MFI as conditional confluence.

The third component carries the stack's most instrument-aware logic. RSI is price-only and therefore dependable everywhere. MFI adds volume participation — which strengthens continuation reads meaningfully — but depends on tick volume being a reasonable proxy for real activity. On majors, where global volume is enormous, it is. On crosses and minors with thinner order books, tick volume gets noisy and can underrepresent real participation, producing divergences that reflect a quiet hour or a quirky broker feed rather than genuine fading. The resulting recommendation is graduated: RSI as the baseline filter across all pairs, MFI confluence added on majors, or on crosses only during their high-liquidity sessions.

• Majors: MFI is genuinely additive — the volume filter strengthens continuation setups.
• Crosses and minors: RSI is more dependable; MFI usable only when restricted to liquid hours.
• A session filter or pair-type toggle is the mechanism — this is a configuration decision made once, not per trade.

Reading confluence

Agreement is the signal; disagreement is information too.

The stack's value comes from its components answering independently. All three aligned — structure supportive with a thick cloud, MACD accelerating, momentum confirming with participation — is the high-quality continuation read the strategies are looking for. Partial agreement is genuinely useful information rather than a failed signal: structure and acceleration without participation often marks a move running on thin involvement; participation and momentum without structure often marks a counter-trend push inside a larger corrective. Reading those partial states honestly, rather than treating any two-of-three as a green light, is where the discretionary skill actually lives.

The key idea

Robust settings and independent components beat optimized ones.

Everything about this stack's construction resists the temptation that ruins most indicator work: classic parameters kept because they're cycle-anchored rather than fitted, standard MACD left standard specifically to avoid overfitting, and an instrument-aware filter choice made on liquidity mechanics rather than backtest performance. The result is a tool whose signals mean the same thing across pairs and across years — which is the only property that makes an indicator's history worth anything.