Where the money comes from
Every currency has a policy rate set by its central bank. A position in a pair is long one and short the other, so you earn the rate on the one you hold and pay the rate on the one you borrowed. The difference between the two is the interest rate differential, and in a retail account it arrives as the swap, credited or debited each night your position stays open.
The current rate of all eight majors, and the gap between every pair of them, is on the rate differentials page. Two warnings that belong with any such table. Your broker takes a spread on the swap, so what lands on your account is worse than the policy difference, sometimes far worse. And on a pair where both rates are close, the swap can be negative on both sides.
Why it works for months
A rate gap does not change every day. Central banks meet roughly every six weeks and move in small steps, so the currency that pays the most tends to keep paying the most for a long stretch. That patience is the whole point: the position earns while you wait, which changes what a slow market costs you.
It also shapes the ranking itself. The currency the world borrows in, the funding currency of the moment, tends to sit near the bottom of every strength list for months. That is not a coincidence, it is the same force showing up twice.
Why it unravels in days
Carry positions are crowded by construction. Everyone can see the same rate gap, so everyone is on the same side, usually with leverage. When the mood turns, all of them want out at once and there is nobody to sell to, so the funding currency jumps far more than the news of that day can explain.
This is the connection between carry and risk on and risk off. A carry trade is a bet that nothing dramatic happens. A risk-off shock is the thing it was betting against, which is why the yen and the franc rise hardest exactly when the highest yielding currencies fall hardest.
The level is not the story, the path is
A currency can pay the most in the market and still fall, because the market prices what it expects the rate to be, not what it is. A rate of four percent that everyone expects to be two percent next year is a worse hold than a rate of two percent that is going up. The gap you actually want is the one between the expected paths, and that is what a statement or a set of projections can move in a single afternoon.
What our own testing says
In our backtests the rate gap is the most stubborn thing on the board. A simple baseline that does nothing but hold the three widest rate gaps, long the higher rate, beat our own mechanical setups over both periods we measured, and it got better the longer the position was held rather than worse.
The other side of the same finding: setups that ran against the rate gap did worse than setups that ran with it, in every year but one. That is not a rule we have put into the model yet. It is the first candidate for the review in October, and the tests are in the open in our repository rather than in a claim on a sales page.
How the rate level counts inside a score today, as part of the fundamentals pillar rather than as a rule of its own, is on the methodology page.