The dollar is at its 24th percentile. The rupee is at its 97th.

The broad dollar index closed at 119.06 on 7 August — the 24th percentile of its own past five years, and below every daily close of 2024 and 2025. Yet the rupee, at ₹95.43 to the dollar on 14 August, sits in the weakest tenth of its five-year range, 1.4% from the weakest close in its history, and the yen at ¥159.01 sits in the same decile. The dollar is drifting lower and two of Asia’s major currencies are falling against it anyway. The Stress Index reads 39.3, Settled.

Weak by its own standard

The Federal Reserve’s broad dollar index — daily, Jan 2006 = 100, published with a lag of about a week — peaked at a series record of 130.04 on 13 January 2025. At Friday’s 119.06 it has given back 8.4% from that high, and it now sits below the lowest close of either 2024 (119.25) or 2025 (119.13). Ranked against the past five years of daily readings, the level lands at the 24th percentile: the band the dollar tracker labels Low.

This is not the year’s extreme. The index touched 117.44 on 29 January before reversing within weeks. The difference now is the shape: the monthly close has fallen from 120.92 in June to 119.70 in July, and the first week of August ended lower still, at 119.06. A dip got bought; this is a drift.

Two currencies missed it

The rupee closed at ₹95.43 on 14 August, by the ECB’s reference cross; the Fed’s own series ran at ₹95.21 on the 7th. Its weakest close ever is ₹96.82, set on 20 May 2026, which puts the present level 1.4% from the record. The streak is the more telling figure: through 7 August the rupee had spent 550 consecutive trading sessions — every session since late May 2024 — in the weakest tenth of its own rolling five-year range. It does not spike; it makes new ground and holds it.

The yen took the other route to the same neighbourhood. In the last week of July it closed above ¥163.6 for four consecutive sessions, reaching ¥163.86 on 29 July — its weakest close since December 1986 — and then recovered almost 4% within a week to the mid-¥157s. At ¥159.01 on 14 August it stands at the 93rd percentile of five years: off the extreme, still deep in the weak end of the range.

Over the year to 7 August the dollar bought 8.6% more rupees and 6.9% more yen than it had twelve months earlier — against a trade-weighted index that fell 1.2% over the same period.

Where the weak dollar actually went

An average can only fall while most of its large components rise against it if something heavy is moving the other way. The something is specific. Over the year to 7 August, one dollar came to buy 6.0% fewer yuan (6.75 against 7.18 a year earlier), 8.3% fewer Mexican pesos, 6.7% fewer Brazilian reais and 0.5% fewer pounds. On the other side it bought more of nearly every major Asian currency outside China: 8.6% more rupees, 8.4% more Taiwan dollars, 6.9% more yen, 1.7% more won — plus 1.3% more Canadian dollars and 0.7% more euros.

The two currencies doing most of the lifting belong to the two countries the US import ledger now runs through. China’s share of US goods imports has fallen to its June 2000 level while Mexico has become the largest supplier — and over the past year the currencies of both have strengthened against the dollar, while the currencies of Asia’s other big exporters, India and Taiwan above all, have cheapened. Whatever mix of policy, rates and flows drives each leg, the direction of the pricing effect is not ambiguous. The strong yuan has already shown up at the US border — the dollar price of Chinese goods has been rising for exactly this reason — while a rupee and Taiwan dollar this weak pull the dollar price of what those economies ship the other way.

Fifteen days like this in fifteen years, thirteen of them in 2026

Rank each series against its own trailing 1,260 trading days — the window every tracker on this site uses. Since 2011, there have been exactly 15 trading days on which the broad dollar closed in the bottom quarter of that range while the rupee closed in the weakest tenth of its own: one day in October 2011, one in April 2021, and 13 days in 2026 — five in mid-April, seven in early May, and 7 August. Require the yen to be in its weakest tenth as well and the count falls to ten days, every one of them this spring. As a configuration, a soft dollar with hard-weak Asian crosses is essentially a 2026 invention.

What the percentile can and cannot say

A rolling percentile rewards persistence, not speed. The rupee’s 8.6% year is not an extreme pace; its 550-session streak says something narrower — that in over two years it has never retraced enough to leave the weakest tenth of its range. The yen, which round-trips violently, shows what the alternative looks like. And none of these series says why: rate differentials, portfolio flows and official intervention are invisible in a price. Whether the yen’s snap-back from ¥163.86 was bought is not knowable from the exchange rate alone — Japan’s Ministry of Finance publishes monthly intervention totals, and August’s is the place that question gets settled.

The index view is calm for a reason worth being precise about. The Stress Index reads 39.3, Settled, as of 13 August, and its dollar-pressure component reads 41.4 — because that component scores the pace of the dollar, not its level. Over the latest 63 trading days the broad index moved +0.01% — effectively unchanged. A spiking dollar tightens trade finance everywhere at once; a drifting one does not. Both facts stand together: the level at its 24th percentile, the pace at almost exactly zero.

What to watch

₹96.82. The 20 May close is the weakest in the rupee’s history and sits 1.4% away. A daily close above it would make the record current again, with the Fed’s H.10 release confirming about a week behind.

The dollar-pressure component crossing 50. On the Stress Index board, that is the sign the 63-day change has turned meaningfully positive and the dollar has stopped drifting and started climbing.

Japan’s August intervention total, published at month-end. A large yen purchase on the MoF ledger would date the late-July reversal as policy; a zero would leave it to the market.

Correction, 13 October 2026: An earlier version gave 119.13 as 2024’s floor. That was 2025’s (16 September 2025); 2024’s was 119.25 (8 January 2024). The sentence’s point, that 119.06 is below both, stands. A heading said thirteen days in fifteen years; the section counts fifteen, thirteen of them in 2026.