NDI Monthly Outlook: October 2026

The monthly outlook does not begin with a view on where anything will trade. It begins with a list of questions whose answers would most change what a reasonable allocator should do over the next thirty days.¹ Each question is binary, resolves on a fixed date against a public source, and is forecast independently by several models before being aggregated. The external probability is the best available market-implied or survey-derived estimate of the same event.

Method, briefly: questions are proposed by the decomposition stage, pruned for redundancy, and frozen on the last business day of the prior month. Forecasts are frozen at 06:00 UTC on issue. Nothing in this document is revised after publication; every number below will appear in the archive with its outcome and score.

§ ILast month, scored

September resolved with a mean Brier score of 0.158 against 0.167 for the external probabilities, a skill of roughly 5%. That is modest, and it was achieved almost entirely through restraint: nine of ten September forecasts sat within six points of consensus. The one material gap — systematic positioning above the 75th percentile, NDI 64% versus 52% — resolved yes.

Restraint was the correct response to July, when the ensemble was confidently wrong on liquidity and positioning at the same time. It is not a strategy. A system that only agrees with consensus has nothing to contribute; its job is to disagree when, and only when, the disagreement survives scrutiny. October is the first month since June in which it does so at scale.

§ IIThe questions

§ IIIStructure

Read individually, the questions say: liquidity probably improves, systematic funds probably buy, credit is slightly under-priced for stress, the dollar slightly favours weakness. Read together, they describe a single scenario with one dominant failure mode.

Questions 01, 04 and 08 are positively linked; they are three windows onto the same improvement in financial conditions. Treating them as independent would triple-count one idea. The allocator collapses them into a single factor before sizing, which is why three constructive gaps produce only one unit of constructive exposure.

Questions 07 and 10 are the opposite: they describe what happens if the improvement is already over-owned. The probability that both resolve yes is low — the ensemble estimates 14% — but in that state, correlated losses arrive through every constructive position at once. That joint tail, not any single question, sets the risk budget.

§ IVStrategy state

The regime label is a compression of the probability structure, not a forecast in its own right. Constructive follows from the first cluster. Crowding follows from the second, and from the observation that our own constructive view is shared with an increasing share of systematic capital — a position we hold in common with others is less valuable and more fragile than one we hold alone.

Risk budget falls from 0.64 to 0.58. This may look inconsistent with a more favourable central case. It is not: the budget is driven by the shape of the downside, and the downside got fatter. If question 10 resolves no by mid-month, the budget will be revisited.

§ VWhat would change our mind

• A liquidity print that reverses the Treasury cash drawdown. Question 01 would fall below consensus within a week, and the constructive cluster would lose its anchor.
• Realized volatility rising while positioning still climbs. This is the configuration that preceded July, and the Historian agent assigns it a high weight in every comparable episode.
• Any sustained positive equity–bond correlation. In that state the Risk agent has standing authority to cut exposure without waiting for the next aggregation cycle.

Interim revisions are published on the system page. The November outlook will open by scoring this one.