September 2026 · Issue No. 021 · Featured Research
Equal false-alert rates do not make model monitors interchangeable. In 5,000 paths per case, residual CUSUM detects every specified mean shift; a rolling beta monitor detects every coefficient break. A fixed Treasury model shows why the statistic must match the failure: an alert can identify deterioration without identifying a regime.
Format
One research feature each monthVolume II · 2026
Nine issues, January to SeptemberResearch Standard
Results before rhetoricIssues
August 2026 · No. 020 · Markets
Sponsored DVP grew 169 per cent between May 2023 and August 2026, while the cleared share of publicly observed venues peaked and then receded. Only the uncleared quarter-end result survives adjustment for four comparisons.
July 2026 · No. 019 · Markets
Capital spending at five US builders rose from 40 to 82 per cent of operating cash flow in two years, while net borrowing moved from repayment to issuance.
June 2026 · No. 018 · AI
No simulated systemic event was observed in 200,000 periods when controls were independent. Common-mode authority, not model concentration alone, moved the tail.
May 2026 · No. 017 · Quant
Trailing factor correlation sat below its own median the week the 2007 quant unwind began. The co-movement appeared inside the episode, not in the window before it.
April 2026 · No. 016 · Markets
The intended spread study failed and is reported as such. What remains is the parameters and their dates, five of which have moved since the release set them.
March 2026 · No. 015 · Markets
Four Cboe constant-maturity indices trace the term structure every session. It inverts on one session in eight, and the level explains about a third of its variation.
February 2026 · No. 014 · Markets
Price impact per dollar traded roughly doubles in stress. A simulated book shows the cost curve is most convex on the orders that looked small.
January 2026 · No. 013 · Quant
A thousand strategies with no edge at all produce a best in-sample Sharpe of 1.46. Five genuine signals planted among them did not make the top five.
Archive
Volume II · 2026 · Current Volume
Volume I · 2025 · Previous Volume
About
ANANNYE is a monthly research journal on markets, systems, and risk. Each issue centres on one research feature, written to make a single question precise: what is being measured, what mechanism may be at work, what evidence is available, and where the argument stops.
The journal is quantitative in method, but not mechanical in tone. Its focus is market microstructure, liquidity, volatility, execution, leverage, model risk, and the infrastructure through which decisions become prices. Artificial intelligence is treated as part of that infrastructure when it affects trading systems, governance, information flow, or operational risk.
The aim is a durable record of research judgement rather than a stream of commentary. Each issue pairs a clear thesis with technical framing, an original figure, primary references, and stated limitations.