Macro Fundamentalist
12-month lens
₹94.50
↑ INR80% band 91.50 → 97.80
REER undervalued 5.6%, real rate diff +1.7%
The cone widens because honest forecasts are uncertain. By May 2027, USD/INR sits between ₹92.74 and ₹100.81 with 80% confidence. The dotted vertical lines are events that can move the curve — RBI MPCs, FOMC, the Warsh transition.
Purple line + shaded band = what the swarm called and its 80% confidence interval. Dashed white = the actual close that landed. Dots inside the band are calibrated; dots outside are honest misses. Hover for the exact numbers per day.
When swarm conviction is Low and the in-band hit is ≤15 paise, the OVERCONFIDENT flag on multiple agents (macro, bop, policy, technical, history all tagged OVERCONFIDENT this cycle) is a systematic mis-calibration of the confidence-labeling logic rather than a true forecasting error — the label should only fire when the miss exceeds 25 paise OR the outcome falls outside the stated band.
The contrarian agent's d1 call of ₹95.75 was the closest to the actual ₹95.95 close (20-paise miss) yet again — reinforcing prior lesson #3 — but this cycle it diverged from swarm consensus by only 35 paise (not ≥30 paise threshold already codified); refine the trigger: when contrarian diverges from swarm median by ≥25 paise (lowering from 30) and Brent is in $100–$115 range, raise contrarian weight by 150 bps for that cycle.
Policy and technical remain at 22% each (total 44%) despite prior lesson #2 prescribing a 200 bps trim each when Brent is $100–$115 and all four agents share the same load-bearing threshold — this lesson is being under-applied; implement it as a hard pre-publication rule: if Brent is in $100–$115 on forecast morning AND policy's and technical's load-bearing assumptions both reference the same Brent threshold, the weight trim to each must execute automatically before swarm aggregation, not as a post-trade reflection.
Called ₹96.44 ± 0.35; the rupee closed at ₹96.57 — 13p weaker, well inside the 80% band.
Geopolitical / Structural earned its weight: the daily lens held against the tape.
Magnitude was tight, but no whip. Nothing structural to revise.
Weights stay. No regime shift in the data; tomorrow's call is built on the same lens hierarchy.
Each agent runs independently against the same data. We aggregate by horizon-weighted mean. Outliers are flagged, not hidden — disagreement is information.
REER undervalued 5.6%, real rate diff +1.7%
Oil normalisation, FPI debt inflows expected
RBI forward book overhang, intervention fatigue
Mean reversion from +7.2% above 200WMA
Iran ceasefire, Warsh dovish pivot, IPO inflows
2020 COVID recovery analog (63/100 fit)
Crowded short-INR, Warsh more dovish than priced
Disagreement is information. The fact that Policy stands apart from the other six tells you the swarm is split on whether RBI lets INR appreciate toward REER fair value or absorbs the gains to preserve export competitiveness.
Five indicators we monitor daily. Any one breaking would shift the forecast — and we'd say so, immediately.
Tomorrow's call is ₹96.31 ± 0.34. That's marginally stronger than today's spot of ₹94.47, and it sits inside a 12-month consensus that points to ₹96.61. On paper that's a calm, appreciation-biased forecast. The reality underneath is that six of our seven agents see appreciation and one — Policy — sees structural depreciation to ₹99.80. The interesting story isn't the consensus. It's the disagreement.
The macro agent's argument is the cleanest. India's 40-currency REER sits at 92.72, roughly 5.6% below the 12-year mean. Real rates favour India by 1.7 percentage points. Growth differential is +4.5 points. Mechanically, the rupee should be trading nearer ₹94.50. Mechanically, it isn't — and that's where the policy lens earns its weight.
The Reserve Bank of India holds a forward book of roughly $104 billion in net short positions. Every dollar of that is a future dollar-buying obligation, and it functions as a structural cap on rupee appreciation. The RBI has burned $31.4 billion of reserves in the last six weeks defending against depreciation; it has no equivalent appetite to defend the other side. The policy agent is therefore not predicting a strong dollar — it is predicting RBI's revealed preference for orderly depreciation in the 2.5-3.5% per annum range. That is why it stands 6.75 rupees apart from the rest of the swarm at the 12-month horizon.
Yesterday made the case in miniature. We called ₹96.44; the rupee closed at ₹96.57. Direction was right; magnitude was 14 paise off because PSU banks intervened in the afternoon session on the RBI's behalf. That is exactly the mechanism the policy agent has been flagging, and it's why we are bumping its weight from 22% to 28% over the next five sessions and trimming the macro agent's weight in turn.
The contrarian view, ₹94.50 at 12 months, deserves a hearing precisely because it's the one most likely to be wrong. The argument is that Warsh arrives at the Fed in June and cuts harder than markets expect, the dollar breaks, the crowded short-INR trade unwinds violently. Each step is plausible. The compound probability isn't. Our CIO critique flagged this as overconfident, and we've weighted it accordingly.
Three observable conditions would invalidate the appreciation bias entirely: Brent sustained above $115 for six weeks (forces a current-account flip); FPI outflows above $5 billion per month for three consecutive months (forces a capital-account flip); or US Core PCE staying above 2.7% through Q3 2026 (removes the Fed easing assumption Warsh's nomination has priced in). Any one of those, and the cone shifts up 3-5 rupees within a week. We monitor all three daily and will publish the moment one breaks.
The June 4 RBI MPC and June 16 Warsh confirmation hearing are the next two dates that meaningfully widen the cone. Between now and then, expect range- bound trade, ₹93.80 to ₹94.80, with a downward bias if oil holds below $100 and an upward bias if FPI debt flows turn negative. If neither happens, the rupee drifts into May's MPC meeting near ₹94.20 — a position our weighted consensus is comfortable defending.