Currency strength meter

Eight currencies, ranked from all 28 pairs over the last 24 H1 bars. One pair cannot tell you whether the euro is strong or the dollar weak; this takes all twenty-eight and gives you eight numbers instead of twenty-eight charts. Last closed bar 2026-09-16 06:00 UTC, snapshot built 20:07 UTC (14 h later).

The bars share one scale, marked at its end, so a quiet day reads as a quiet day rather than filling the width. The line beside each currency is the last 168 readings of the same number: strong and still climbing is a different thing from strong and giving it back.

Where the spread is widest

Widest spread: AUD/CAD, +0.44 — AUD strongest at +0.202, CAD weakest at −0.240. That is the arithmetic maximum, not a recommendation: it names the pair that has moved most over the window, which is not the same as the pair that will move next.

Neither side saying much: USD/JPY, NZD/JPY, NZD/USD. Both currencies sit within 0.06 of zero — the quietest quarter of this window’s own history. These are the pairs the meter has nothing to say about.

Every pair, by how far its two currencies have parted

Spread is the difference between the two currencies’ strengths. Move is what the pair itself did over the window: the same fact, scaled by 8/7 — see how it is computed. Pairs we publish a page for link to it; the fifteen crosses we do not trade have no page, and no link.

The window, currency by currency

The eight lines sum to zero at every point — that is arithmetic, not a market fact, and it is why this instrument can never show “every currency weakening”. What it can show is the shape: a currency that has been at the top all week, or one that arrived there this morning.

Which pair, currency against currency

Each cell is the pair’s own move over the window, scaled by 8/7. The matrix is a convenience, not a second measurement: the only thing this meter adds to a table of percentage changes is the decomposition into eight currencies above. A bold cell is in the widest 5 % of what this window usually gives (spread above 0.85); the shades below it are the 80th and 50th percentile (0.45 and 0.22), measured on this window’s own history rather than chosen.

How it is computed

For every pair, the log return over the window: r = log(close / close 24 bars ago), times 100 so it reads like a percentage. A currency takes +r from each pair where it is the base and −r where it is the quote, and its strength is the mean of those seven numbers. For the dollar that is

USD = 100/7 · ( −EURUSD −GBPUSD +USDJPY +USDCHF −AUDUSD −NZDUSD +USDCAD ), each term the log return of that pair over the window.

What “+0.202” means. It is an average over seven pairs, not a price change: AUD gained about 0.20 % against the other seven on average over this window. No single pair moved by that number, and the meter never claims one did.

• Bars, not hours. The window is 24 bars. Forex trades 24 hours for five days, so 24 H1 bars on a Monday morning reach back into Friday’s session rather than over the weekend.
• One grid. A bar counts only if all 28 pairs have it. A currency whose pair was silent would otherwise get a strength built from a stale price, and the eight would stop summing to zero.
• No normalisation. Raw numbers, comparable between currencies inside the same window and nowhere else.
• Closed bars only. The forming bar is not in it, which is why the ranking does not reshuffle while you read it.

Self-test: the eight strengths sum to −8.3e-17 (28 of 28 pairs loaded). It is arithmetic, not a market fact — if it is not zero with all 28 loaded, a symbol failed and this page is wrong.

What we measured

The arithmetic above is exact. Whether it is useful is a different question, and the honest answer today is that we have not found it to be. One pilot, run on three years of hourly bars: every 24 bars, take the pair with the widest spread and hold it for 24 bars, in the direction the meter points. Costs excluded.

Picking by the meter was indistinguishable from picking at random, and already negative before costs — a typical spread on these pairs is 0.02–0.05 % per round turn, which is the size of the whole effect. That is a pilot, not a study: one window, one holding time, no stop, no target, no filter by hour. It does not prove there is nothing there. It does show there is nothing easy there, and it is the bar any later claim has to clear.

Questions people ask

Next to this: how these markets move together, what each hour of the day does, and how far a day travels.