Majors versus crosses: why one filter setting can't serve every pair.

The mechanism

Volume proxies degrade with liquidity.

Majors like EUR/USD, GBP/USD, and USD/JPY trade in volumes large enough that almost any broker's tick count is a reasonable proxy for genuine global participation. Crosses and minors — EUR/GBP, AUD/JPY, NZD/CHF — have thinner order books and less continuous liquidity, so tick volume becomes noisy and can materially underrepresent real activity. The practical consequence is specific: an MFI divergence on a cross may indicate buyers or sellers genuinely fading, or it may simply reflect a low-liquidity hour or a quirk of the broker's feed. RSI, being price-only, doesn't inherit the problem at all.

The graduated response

Baseline everywhere, confluence where it's earned.

The recommendation follows the mechanism rather than a preference. RSI serves as the baseline filter across all pairs — dependable, price-derived, immune to volume-feed quality. MFI is added as confluence only where its input is trustworthy: on majors generally, and on crosses only during their high-liquidity windows, where tick volume becomes more reliable — London for EUR/GBP, Asia for AUD/JPY. In those windows MFI adds genuine value by helping avoid price spikes that lack volume follow-through. Outside them, it adds noise wearing the costume of confirmation.

The implementation choice

Auto-adjust or manual toggle — decided once, documented.

The tool can handle this either by auto-adjusting — defaulting to RSI on crosses and MFI on majors — or by exposing a manual toggle for full operator control. Both are defensible, and the choice matters less than the discipline around it: whichever mode is used, the setting is decided before the session rather than adjusted while a setup forms. An auto-adjusting tool has the advantage of removing a decision that's easy to get wrong under pressure; a manual toggle has the advantage of visibility, since the operator knows exactly which filter is running. What neither survives is being switched mid-setup to produce the confirmation the other mode withheld.

• Session filters are the cross-pair mechanism: restrict MFI signals to the pair's own liquid window, not to all of London.
• Instrument context belongs in the journal alongside session — the analytics layer slices by pair and session for exactly this reason.
• A filter that fires differently on the same pair week to week is a configuration problem, not a market read.

The broader lesson

Indicators inherit their inputs' quality, always.

This case is a specific instance of a general truth the whole system keeps re-encountering: every analytical layer is bounded by the honesty of what feeds it. The volatility matrix carries sample-status grades because young baselines can't support percentile claims. The EV lab gates thin lenses as provisional. The SDE keeps immature windows blank. And the momentum filter degrades on thin instruments because tick volume degrades there. In each case the correct response is the same — know the input's quality, and let it govern how much the output is trusted.

The key idea

The tool should know what it's looking at.

A single indicator configuration applied identically to every instrument implicitly assumes every instrument reports data the same way, and forex conspicuously does not. Building pair-type and session awareness into the filter is what keeps its signals comparable across the whole traded universe — so that a confirmation on EUR/GBP at 3am and one on EUR/USD at London open carry something like the same weight, or the tool declines to offer one at all.