What the swarm learned, dated and falsifiable.

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.

Trigger

Brent is between $100–$115/bbl at 08:00 IST on forecast day AND policy and technical both list a Brent price threshold as their primary load-bearing assumption

Applies to

Policy Reaction Modeler · -2.00% weight

Confidence

High

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.

Trigger

≥3 agents are tagged OVERCONFIDENT on the same forecast cycle AND the actual close lands inside the 80% band AND the absolute miss is ≤20 paise

Applies to

Swarm-level · no weight change

Confidence

High

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.

Trigger

Contrarian d1 call diverges from swarm median by ≥25 paise AND Brent is between $100–$115/bbl at time of forecast issuance

Applies to

Contrarian / Red Team · +1.50% weight

Confidence

Medium

All seven agents converged on ₹96.10 (identical d1 calls for macro, bop, technical, geo, history, contrarian) despite carrying materially different load-bearing assumptions — this consensus collapse on the point estimate masks differentiation that only surfaces in the band tails; when ≥5 agents produce d1 calls within a 10-paise corridor, force the contrarian agent's band to be widened by 40 paise asymmetrically toward the contrarian thesis direction before publishing the swarm band.

Trigger

Five or more agents produce d1 calls within a 10-paise corridor of each other AND swarm conviction is tagged Low

Applies to

Contrarian / Red Team · +1.50% weight

Confidence

Medium

When swarm conviction is Low and the actual close lands within 21 paise of the point estimate (in-band), the macro agent's OVERCONFIDENT flag is a calibration artifact rather than a structural problem — do not trim macro weight solely on the basis of an OVERCONFIDENT tag when the absolute miss is ≤25 paise and the band contains the outcome.

Trigger

Macro agent is flagged OVERCONFIDENT AND the actual close is within the 80% band AND the absolute miss is ≤25 paise on a 1d horizon

Applies to

Macro Fundamentalist · no weight change

Confidence

Low

The technical agent carries the highest weight (35%) yet its load-bearing assumption (trend-following regime persists until spot closes below 20-day SMA for 3 consecutive sessions) is structurally correlated with the policy agent's Brent-<$115 threshold — on days when Brent is between $100–$115 AND no SMA crossover has fired, technical's 35% weight overstates independent information; trim technical by 200 bps and reallocate to bop, which carries a distinct BoP-arithmetic framework, to reduce effective single-factor (oil) concentration.

Trigger

Brent is between $100–$115 AND USD/INR spot has not closed below the 20-day SMA for 3 consecutive sessions AND technical agent weight is ≥33%

Applies to

Technical / Quant · -2.00% weight

Confidence

Medium

When swarm conviction is Low and all agents cluster within 50 paise of each other on a 1d horizon, the forecast band is well-calibrated — reinforce the current band-width logic rather than widening it, as the 13-paise in-band hit confirms the band earned its keep.

Trigger

Swarm conviction tagged Low AND all individual agent d1 calls within a 50-paise range AND 80% band width ≥ 110 paise

Applies to

Swarm-level · no weight change

Confidence

Low

Contrarian was tagged UNDERCONFIDENT despite producing the closest d1 call (₹95.72 vs actual ₹95.97, 25-paise miss vs macro/technical at 23 paise) and being the only agent with a structurally differentiated thesis (crowded-short squeeze + ceasefire optionality) — when contrarian's load-bearing condition (Brent >$100 sustained, crowded short positioning) is active and its d1 call diverges from swarm consensus by ≥30 paise, raise contrarian weight by 150 bps for that forecast cycle.

Trigger

Brent > $100/bbl AND contrarian d1 call diverges from swarm-consensus d1 call by ≥ 30 paise AND positioning signals indicate crowded short-INR (e.g., RBI forward book net short > $90bn)

Applies to

Contrarian / Red Team · +1.50% weight

Confidence

Medium

Policy and Technical agents were each weighted at 22% yet both carried identical load-bearing assumptions (Brent <$115) and produced the same directional bias as macro and bop — this structural corr­elation inflates effective weight on a single oil-price scenario, so when Brent is between $100–$115 and all four agents share the same load-bearing threshold, trim policy and technical each by 200 bps and redistribute to contrarian, which held the lone differentiated view.

Trigger

Brent trading in the $100–$115 range AND policy + technical + macro + bop all cite the same $115 Brent threshold as their primary load-bearing assumption in the same forecast cycle

Applies to

Policy Reaction Modeler · -2.00% weight

Confidence

Medium