Reading the two sides of a currency pair

If EURUSD falls, did the euro weaken, the dollar strengthen, or both? Could one representation answer that question consistently for historical bars and fresh quotes?

We represented currencies relative to an equal-weight basket. A currency has no absolute exchange-rate level by itself. Choosing the reference makes that relationship explicit.

A pair's log price can be represented as its base currency value minus its quote currency value. Eight currencies supply seven independent relative dimensions. The 28 pairs contain overlapping observations of those relationships.

The historical and live representation used the same linear construction. We compared alternative constructions, reconstruction errors, quote freshness, and the movement left after common factors were removed. We separately tested whether trailing strength, stretches, and agreement between legs forecast subsequent outcomes.

Trade-expression experiments asked how a view on one currency behaved through different counterpart currencies. They measured both correct and incorrect views, because an expression that amplifies a correct judgment can amplify an incorrect one too.

The representation explained most hourly pair movement, and the broad decomposition remained similar across the sample halves. At very short intervals, quote timing and the residual between related pairs mattered more. A stale live panel could therefore give a materially different picture from synchronized quotes.

Trailing strength did not establish a persistent directional advantage in the tested rules. The day's movement scale carried more information about subsequent activity than the recent direction carried about the next direction.

Pair choice affected expression and cost. Two currency legs can reinforce or cancel one another. A view on a currency and a view on a particular pair are consequently different forecasts.

The historical tests give useful ways to describe this distinction. They do not establish a universal best counter-currency or a permanent list of pairs to trade.

The main representation covers roughly two years, while finer and live comparisons use shorter windows. The small sample of desk judgments does not establish predictive performance or improved profits.

Common movement is a statistical description. It does not identify the causal news or participant behind that movement. A basket-relative strength chart also records what has already happened; a future-direction claim needs a separate test.