USD/INR Market Outlook: Aug-31 – Sep-4

What the market did — five facts

1. The climb stopped, and we can date it. The dollar closed at 95.7450 on 24 August, its highest of the month. The next day it fell 33 paise. In the four sessions since it has not closed above 95.55, and August’s peak of 95.7675 has not been approached. The rate is now sitting slightly below its one-month average, having spent all of August comfortably above it.

2. Oil fell, and that is what moved the rupee. Brent dropped 4.74% over the week and remains around $90 a barrel. India buys more than 80% of its oil from abroad, so cheaper oil means fewer dollars leaving the country to pay for it. Of everything on this page, this is the input with the most direct and fastest transmission into the rate. Watch it daily.

3. Foreign investors sold Indian shares and the rupee still gained. On Friday 28 August foreign institutions were net sellers of ₹5,040 crore, while Indian institutions bought ₹5,184 crore. The rupee strengthened 16 paise that day anyway. That combination is unusual. It either means the oil effect was strong enough to overwhelm the outflow, or that the central bank was smoothing the market. We cannot separate the two without reserves data we do not have.

4. Global markets were calm and buying risk. Gold fell 3.8% over the week and silver 4.6% — investors were unwinding safety trades, not buying them. The main US fear gauge fell 4.6%. Nothing in global sentiment currently argues for a weaker rupee.

5. Indian shares stayed soft. The Nifty was down 0.31% on the week and remains 7.5% lower for the year. This is the one domestic signal still working against the rupee, and it has been doing so for months.

What the price action says, without the jargon

What the outside world is doing

Putting the three together

Price behaviour: the climb has ended and downward pressure has increased, but four sessions do not make a new direction, and the market is moving too little to have conviction either way.

Outside drivers: genuinely mixed for the first time in a month. Oil, global calm and Indian interest rates now help the rupee. Foreign selling, weak Indian shares, US rates and the weekend’s Gulf news work against it.

Forward market: neutral. The cost of booking ahead is flat, so waiting is neither getting cheaper nor more expensive.

Together: neutral, with wide two-way risk. Confidence on direction is low. Confidence on the range is medium — lower than we would normally claim, because two large events sit inside the week and because the calm of the last fortnight is the sort that usually ends abruptly.

Just beyond this week

Adjusting for when your money moves

Desk forward calculator, standard month-wise rates, value date 2 September 2026. “Change on the week” compares the extra cost only, against the 27 August value-date curve — not the rate itself, which fell with the market.

What this means for each side

Importers. Two things went your way at once, which is unusual. The market fell 32 paise and the cost of deferring stopped climbing. A payment deferred from a week ago is cheaper on both counts today. That said, one-month cover at 95.6783 still sits inside the exporter’s selling zone — you would be paying forward for a level the market has not reached in eight sessions. Prefer short-dated cover, where the annual cost is around 2.3–2.6% against roughly 2.9–3.0% further out.

Exporters. The whole curve came down with the market, so booked rates are lower across every date. But the extra you earn by booking ahead is intact: six-month cover at 96.8141 is 137 paise above today’s market, and a full year is 275 paise above. For anything settling beyond three months, the forward gives you a better rate than the spot market plausibly will.

Where the forward curve crosses our markers

Read the last row carefully. An importer covering three months out is already paying above the level whose break we would treat as proof of a weaker rupee. Beyond three months, the argument this note is having about direction has already been settled by the forward market — and settled against the rupee. One point in your favour: that crossing sits at three months rather than two, which buys about four extra weeks before the maths turns against deferral.

Desk summary

This week, 31 August to 4 September. Neutral, with wide two-way risk. Friday’s close expected between 95.20 and 95.70, with extremes between 95.10 and 95.85. The climb that ran through August ended on 25 August; nothing yet has replaced it with a downward trend. Today’s growth data and Friday’s US jobs report bookend the window.

One month out. The question is whether the market can get back above 95.77. The US rate decision on 15–16 September is the likely decider. Note that one-month cover already sits inside the exporter’s zone.

Beyond. The forward market charges 96.11 at three months and 98.22 at a year — both above every marker in this note. Any longer view must be measured against those numbers, not against today’s rate.

Technical detail retained so a reviewer can trace every conclusion above to its source. This section deliberately uses full analytical vocabulary; sections 01–06 do not.

A · Technical evidence

Market structure. 28 Aug (last confirmed): O 95.5100 · H 95.5775 · L 95.3150 · C 95.3775. Range 26.3 paise. 31 Aug 09:06 capture: O 95.4900 · H 95.5000 · L 95.4600 · C 95.4650, range 4.0 paise — see the data-quality note in D. Position in 50-day range 52.4%; in 52-week range 85.1%. 50-day high 96.6725, low 94.1350. Changes by window: 1d +0.092% · 5d −0.240% · 15d +0.270% · 30d −0.846% · 45d +1.134% · 90d +2.102% · 120d +4.935% · 180d +6.608% · 360d +9.265%.

Trend. ADX14 22.19, labelled WEAK_TREND, 10-day slope −4.28 (was 18.07, slope −8.28). DI+ 16.82 versus DI− 30.71 — the gap widened from −7.0 to −13.9 in one week. Moving-average alignment MIXED (was STRONG_BULL). Price above SMA20 (95.4582, +0.007%), SMA100 (95.1356), SMA200 (92.8997, +2.761%); below SMA50 (95.5037, −0.041%), EMA8 (95.5189), EMA21 (95.5489). EMA55 95.3898, EMA100 94.7707. Exponential stack no longer bullish. Supertrend DOWN at 96.0524 (X). Parabolic SAR UP at 95.2466. Ichimoku: price BELOW_CLOUD (was ABOVE_CLOUD) — conversion 95.5412, base 95.7963, cloud edges 95.5475 / 95.7169.

Momentum. RSI14 47.62 · RSI7 45.19 · NEUTRAL · no divergence. MACD −0.01761, below signal −0.00669, histogram −0.01092 and not rising. Last week MACD was +0.01757 above signal — the cross reversed within five sessions. Stochastic K 40.98 / D 45.88. Williams %R −59.02 · CCI20 +12.13 · TSI −1.82 (was +1.42). ROC12 +0.142% against ROC30 −0.846%.

Volatility. ATR14 0.2159 (21.6 paise) = 59.5% of its 50-day average. ATR20 0.2480. HV20 2.22% · HV60 4.88% · HV120 6.29% — HV20 is 35% of HV120. Parkinson 1.72% · Garman-Klass 1.58% · Rogers-Satchell 1.50%. Volatility percentile of history 34.1. Bollinger bandwidth 0.00788, percentile 32.0, squeeze ACTIVE, position in band 0.509. 1-day VaR 95: −0.548%. Expected shortfall: −0.869%.

Support and resistance zones (clustering tolerance 0.1146 = 0.35 × average daily range):

Table scrolls sideways on narrow screens. 17 zones identified in total.

B · Historical and quantitative evidence

Statistical ranges, unconditional, anchored on 95.3775. We now publish these split by measurement basis, which is the methodology change flagged in Section 01:

Derived from 878 observations of this pair’s own history, excluding the incomplete 31 August bar. The published closing band of 95.20–95.70 sits marginally wider than the 5-day 50% closing band and materially inside the 68%, which is where we have chosen to sit given the event calendar. The published intraday reach of 95.10–95.85 approximates the 75th-percentile excursion band.

Conditional scenarios and probabilities. Engine output: bullish 30% / base 40% / bearish 30%, regime WEAK_TREND, regime prior 44%, historical tilt +0.20, structural tilt −0.25. Published weights: 28 / 45 / 27. Two adjustments made, both stated:

• −2 from bullish. The engine’s positive historical tilt rests materially on the sudden_volatility_contraction stream (73% higher at five sessions), which is triggered by the 4.0-paise reading on the incomplete 31 August bar. Strip it and the remaining seven streams run 45–58% — indistinguishable from a coin flip. This is the same contamination that inflated last week’s compression reading.
• +5 to base, −3 net redistribution. Offset against the two tier-1 calendar events, which argue against a narrow week even as the compression argues for one.

Weighted historical follow-through — eight streams, 793 observations. Market-condition analogues 45% higher (20 obs) · unusually narrow range day 54% (250) · inside range day 57% (128) · unusually small daily range 47% (106) · sudden volatility contraction 73% (86) — flagged, see above · after volatility squeeze active 58.3% (187) · support test 20D Fib 38.2% 22% (9, weight 0.10) · resistance test 20D Fib 23.6% 57% (7, weight 0.08). Weighted result: 55% higher at five sessions, against 58% last week. Lean classified NO_CLEAR_LEAN at 1, 3 and 5 days; HIGHER at 10 days (59%).

Level-touch probabilities from 20 comparable historical situations: nearest resistance zone centre 95.5068 touched 75% within five sessions; nearest support zone centre 95.2614 touched 35%; 20-day high 95.7675 touched 15%; 20-day low 94.92 touched 0%.

Treasury excursion metrics. 5-day: exporter typical −0.200% (≈19p), severe −0.861% (≈82p); importer typical +0.219% (≈21p), severe +0.960% (≈92p). 10-day: exporter −0.281% / −1.165%; importer +0.358% / +1.392%. 20-day: exporter −0.388% / −1.393%; importer +0.605% / +2.066%.

C · Cross-currency and flow context

Correlation. USD/INR to EUR/USD is −0.023 over 30 days and 0.092 to GBP/USD, against 0.443 between USD/INR and EUR/INR. The dollar-index framing remains structurally downweighted for this pair, as corrected two cycles ago. The two legs of EUR/INR are effectively uncorrelated right now, which means cross volatility is close to the quadrature sum of both.

D · Sources, gaps and method

Sources. Desk quantitative and scenario engines v3.0 / v1.0, run 31 August at 09:06 IST. USD/INR series 884 rows, 2 Jan 2023 to 31 Aug 2026; all integrity checks passed, no duplicate dates, no invalid OHLC. Desk forward calculator, standard month-wise rates, spot value date 2 September 2026. FII/DII combined net figures, 28 August. Real-time commodity futures and world market indices feeds. Fundamental drivers note dated 31 August 2026. Prior week’s published outlook of 24 August for all week-on-week comparisons.

Data marked unavailable — flagged, not estimated.

• Now resolved: FII / FPI flows (received this cycle, single day only — a daily series is still needed) · Brent and WTI direction (received as percentage change; an absolute price series is still needed) · Indian equity indices · global risk indices.
• RBI FX reserves and intervention record — still absent. This is the gap that prevents us separating the oil effect from official smoothing in Friday’s move, and it is now the single most valuable missing series.
• India and US government bond yields — absent. No rate-differential attribution has been performed on the forward curve and none should be inferred.
• USD/INR implied volatility and risk reversals — absent. Would establish whether the options market prices this compression as directionless or skewed.
• Confirmed economic calendar — our macro source lists central bank meetings only, with no data releases. Friday’s US employment report is included on standard-schedule grounds, not on source confirmation. Second consecutive week with a calendar defect.
• Series completeness — 26 August is absent from the price file, and 72 business days (7.53%) are missing across the full series.
• India monthly trade balance — the oil-to-currency mechanism remains directionally certain and unquantified.

Open signal conflicts — three, down from five.

Evidence quality scores HIGH on data quality, historical sample, indicator agreement and historical agreement; overall engine confidence MEDIUM. We publish directional confidence Low and range confidence Medium. Range confidence is lower than last week’s High despite better data, because the calendar carries two tier-1 events where last week carried none — a wider range is more honest than a narrower one that happens to look precise.

Method changes effective this cycle. Four, all arising from last week’s miss:

1. Spot is published as the last confirmed close, with the live morning price shown separately.
2. Ranges are published on two bases — closing and intraday — because readers grade the note on closes and the previous single band was built from intraday excursions.
3. Any latest bar below the 5th percentile of range, captured within two hours of the open, is excluded from volatility, candle-pattern and analogue computation.
4. When ADX is below 25 with a negative 10-day slope, DI+/DI− carries more weight than MACD crossovers and moving-average alignment. Last week that reading was correct and was outvoted; it would have flipped the directional call.

Standing method notes. Levels are drawn from Fibonacci retracements, moving averages, volatility bands, channels, pivots, Ichimoku, swing points and round numbers, then clustered at 0.1146 tolerance. Historical analogues use a weighted robust-z nearest-neighbour search over backward-looking features only, across 664 candidate days; best similarity 82.3, average 71.5; candidates must be at least 20 bars old so outcomes are complete, with a ±5-day exclusion so one episode cannot fill the list. The Ichimoku lagging line is excluded to avoid look-ahead bias. The engine is volume-free by design. Seasonality carries only 3–4 years per month, is rated VERY_LOW_SAMPLE, and has not been used. No rate, level, probability or event in this report has been invented. This is a market-timing framework, not personalised treasury advice, and no hedge percentages are prescribed.