Predictive Information and Trading Decisions: Residual Risk, Costs and Sharp Attainable Bounds
Year: 2026
Type: preprint
Abstract: Predictive information does not determine a unique trading value. We begin from the classical binary entropy–MMSE envelope and conditional-information portfolio geometry, with their attribution explicit. The unconstrained erasure equality experiment permits arbitrage at every positive information level. Imposing a uniform conditional residual-risk floor yields a bounded, fully attainable information–Sharpe region without that arbitrage. Adding proportional cost, a leverage cap and second-moment utility gives exact lower and upper decision envelopes, each attained by at most two posterior magnitudes. A finite-grid calculation carries a deterministic approximation allowance. For a fixed finite signal partition, independent calibration gives a safe cost-aware policy, an oracle-regret bound and a simultaneous confidence interval for information. Equal information can then support either zero trading value or positive value under the same admissibility and cost restrictions. Gaussian filtering and vector portfolio identities are retained as supporting benchmarks, not as claims of a new universal information inequality.
Author: Noguer i Alonso, Miquel
Language: English
Subfield: Finance
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7236653576
