Memory Production under Coarse-Graining: A Balance Law for the Indivisibility Deficit

Memory Rank and Liftability of Transition Families Without a Generator: A Projection Theory for the Indivisibility Deficit

We call a two-time family of transition kernels on a finite state space flat when it satisfies the Chapman–Kolmogorov property, and we read flatness as a flat transport over the complex of time intervals. This paper develops the geometry of families for which flatness fails; such a family admits no time-local Markov generator on the observed state space. The defect Δ(t,s,r) is the intrinsic failure of the Chapman–Kolmogorov property; it is a kernel-valued 2-cochain, and associativity of composition imposes on it a coherence identity of Bianchi type. A first result closes the natural route to an invariant: under the invertible mass-preserving re-dressings admissible for a family, the orbit closure of every family contains flat families, so no orbital minimization of a flatness distance defines an invariant. The invariant is built from dilations instead. Within a declared dilation class, F_d denotes the compact semialgebraic closure of the set of compressions of flat transports on d hidden states with positive block masses. The liftability profile h_d(V), the attained distance from V to F_d in a weighted metric on families, is nonincreasing in d and 1-Lipschitz in the family, and its zero stratum consists of the families approximable to arbitrary precision by such compressions. On a grid of m+1 times with n observed states the model sets stabilize: every conditional family of a process on the grid lies in F_D for an explicit D ≤ D_max = ∑{k=0}^m n^{k+1}, so F_d = F{D_max} for every d ≥ D_max, the union of the model sets is closed, and the profile of every family in the closure of the process-induced families vanishes at a dimension at most D_max. The memory rank, the minimal vanishing dimension, is thus finite exactly on that closure. Two linear programs bound the profile from below: the marginal-consistency defect and the joint-law defect; the second vanishes on the closure of the process-induced families, characterizes the process-induced families through the support of its optimal laws, and detects families that the first does not. First-order conditions at the projection are stated on the tangent and normal cones of the strata. The profile is class-relative: passing from time-homogeneous to time-inhomogeneous lifts weakly decreases it and decreases it strictly on explicit families at the dimension-matched level. Exact examples close the paper: a family whose profile is bounded below at every dimension by a direct evaluation of the obstructions, a family with consistent marginals and no joint law, a lumped chain whose memory rank lies strictly below its generating dimension, and the class gap in closed form.

Selection Functionals on Path Space for Stochastic Processes Without a Generator: Finite-Grid Gibbs Selection and Self-Normalized Estimation

We develop an estimation theory on finite-grid path space for stochastic systems whose two-time transition families violate the Chapman–Kolmogorov factorization and therefore admit no generator on the observed state space whose evolution reproduces the given family. Candidate paths supplied by an external proposal mechanism are reweighted by the exponential of a discretized Onsager–Machlup (OM) action computed against a tractable surrogate model that is explicitly not the truth, and expectations under the resulting Gibbs-selected measure are estimated by self-normalized importance sampling (SNIS). Up to a constant and the OM divergence term, the selection weight is the surrogate's own discrete path density raised to the temperature, so the selected measure is a product of two path densities rather than the density ratio used by Girsanov reweighting; this identity is the construction's point of departure from the path-reweighting literature. For Gaussian pairs the product adds precisions to leading order in the grid step δt: the selected measure acquires the diffusion coefficient 1/σ_eff² = 1/σ_Π² + λ/σ_ν² + O(δt), different from the proposal's for every temperature λ > 0, and this quadratic-variation mismatch is the mechanism behind every degeneracy the paper reports—the chi-square divergence between selected measure and proposal grows exponentially under grid refinement, the selected measures converge to a continuum law that is singular to the proposal's, and the finite grid is the primary setting because the importance-sampling representation of the selected measure from the proposal does not survive refinement. The estimator theory is standard SNIS theory specialized to this weight and is stated as such: strong consistency, a central limit theorem, finite-sample concentration in two tiers (polynomial under finite chi-square divergence, exponential under bounded weights), and degeneracy diagnostics organized by the chi-square divergence. The relation to the true process is a conditional transfer under an explicit total-variation hypothesis, flanked by an unconditional kernel-level obstruction bounding, under a contraction hypothesis, every divisible surrogate family away from an indivisible target; the two are logically independent, and under grid refinement the hypothesis is attainable only for proposal–surrogate–temperature triples designed to the target's quadratic variation. Surrogate calibration is formulated as Euler quasi-likelihood estimation, and the inconsistency of calibrating by the OM action itself is recorded. Two fully worked one-dimensional examples are given: a Gaussian pair in which every quantity is in closed form, and a double-well surrogate with non-constant drift divergence in which the benchmark is obtained by transfer-matrix quadrature and the divergence term measurably changes the estimand.

Convex Perturbation and the Suppression of the Indivisibility Deficit

We study how the Chapman–Kolmogorov (CK) deficit of a column-stochastic two-time family {Γ(t←s)}{s≤t} on a finite state space responds to convex mixing with a divisible reference family, Γ_g = (1−g)Γ_0 + gΓ_ref, g ∈ [0,1] — the simplest fixed-weight convex perturbation of an indivisible family. The organizing result is an exact identity for the deficit operator of the mixed family, Δ(t,s,r;Γ_g) = (1−g)[X − gK], where X = A − CE is the substrate deficit and K = (C−D)(F−E) is a substrate–reference interaction term assembled from the kernels on the three subintervals of the triple; the operator-norm bound D(t,s,r;Γ_g) ≤ (1−g)D(t,s,r;Γ_0) + g(1−g)M(t,s,r) follows. Because the identity is exact, the deficit up to temporal scale δt equals (1−g)S(g) for a convex function S; the response need not be monotone in g, can vanish at one interior coupling, and can rebound. This forces a two-threshold formulation of detectability: a first-crossing threshold g_first(δt;ε) and a terminal suppression threshold g_term(δt;ε) beyond which the deficit remains below a detection level ε at every stronger coupling. We prove 0 < g_first ≤ g_term ≤ g_1 ≤ g_0 < 1 with g_1 = 1 − [D_CK + M* − √((D_CK + M_)² − 4M_ε)]/(2M_) and g_0 = 1 − ε/(D_CK + M_), where D_CK is the unperturbed deficit up to scale δt and M_* the interaction scale; the bound g_1 is attained by an explicit configuration, so no bound in D_CK and M_* alone improves it. The sufficient couplings tend to 1 as ε → 0; exact vanishing is guaranteed at g = 1 and possible at no more than one interior coupling. A two-state family exhibits the interior zero, the rebound, and, for tolerances below the rebound peak, the separation of the two thresholds at the scale level. Secondary results, stated in the body, are an interval-local stability lower bound on g_first, quadratic suppression of the integrated deficit, and a rigorous bound for state-dependent coupling. The families are treated as abstract two-time transition families; no claim is made that the mixture is realized by a particular measurement or control protocol.

Koopman Spectral Theory for Indivisible Stochastic Processes

The operator-theoretic analysis of stochastic dynamics standardly assumes a transition kernel satisfying the Chapman–Kolmogorov (CK) factorization, which equips the two-time operator family with the cocycle structure on which generator-based spectral theory rests. We develop spectral machinery for two-time transition families that violate this condition irreducibly—indivisible families—for which no time-ordered-exponential generator in the Dyson (Bochner-integrable) class exists. The framework operates on the forward family {P(t,s)}, s ≤ t, on L²(Ω,μ) through the singular structure of its composition deficit Δ(t,s,r) = P(t,s) − P(t,r)P(r,s), not through eigenexpansions. Two results organize the paper. Locally, an indivisibility tensor β(s) = C(s) − L(s)² − L′(s), computed from second-order Taylor coefficients at the diagonal, governs the exact leading-order scaling Δ(t,s,r) = (t−r)(r−s) β(s) + o((t−s)²); its vanishing is necessary for divisibility of a jointly C² family, and β(s₀) ≠ 0 forces CK failure near s₀. Globally, closed deficit subspaces V_in, V_out concentrate the failure of factorization—every deficit operator vanishes on the orthogonal complement of V_in—and, under a standing contraction assumption, a universal three-pair bound places every divisible comparison family at operator-norm distance at least D(t,s,r)/3 from the family at the worst single time pair, a constant that cannot be improved. Covariant eigenfamilies with non-multiplicative multipliers witness indivisibility; an integrated deficit on the trajectory space L²([0,T] × Ω) vanishes exactly under divisibility; each two-time map admits a Stinespring dilation on a fixed dilation space, jointly continuous in (t,s), with a measurable formal Hamiltonian generator under explicit regularity, and divisible families admit an exact cocycle dilation on every finite time grid. In any selected unistochastic gauge the deficit splits exactly into an interference cross-term plus a non-composition term; only the sum is gauge-invariant. Worked two- and three-state families, including one with V_in ≠ V_out, exhibit the constructions in closed form.

The Physics of the Instant: Analysis from Limits, Relativity, and Renormalization

Much of theoretical physics rests on the idealization that a physical system possesses a complete state at each instant of time, a state sufficient to determine its future evolution. We argue that this completeness is a physical assumption rather than a consequence of how the instantaneous state is defined, and that established physics provides no basis for it and positive reasons to doubt it. Three independent lines of evidence are assembled. First, operationally the state at a temporal point is the continuous extension of descriptions over finite intervals: the ε-δ definition of a limit assigns a value at the point from its punctured neighborhoods, and whether that value is dynamically sufficient is a property of the dynamics that the construction neither supplies nor guarantees. Second, quantum mechanics and general relativity, applied jointly, bound spatial and temporal resolution at the Planck scale, and the Salecker-Wigner clock bound, combined with the requirement that the clock itself not undergo gravitational collapse, yields the same temporal bound from the properties of the clock rather than of the probed region. These bounds are not theorems of a completed theory of quantum gravity, but within the semiclassical domain in which they are derived they obstruct the limiting process δt → 0. Third, in quantum field theory products of field operators at coincident points are singular, an effective description carries a validity scale beyond which it fails, Wilson's analysis makes renormalization a feature of the theory rather than a defect of perturbation theory, with a continuum limit requiring an ultraviolet fixed point of the flow, and the algebraic formulation assigns observables to open regions rather than to points and fixes the dynamics from any neighborhood of a Cauchy surface while assigning no algebra to the surface itself. Taken together, these show that completeness at a temporal point is not delivered by the construction of the state, is not reachable by measurement within the domain of established physics, and is not where established theories place their physical content. Dynamical completeness is defined for the deterministic and the stochastic settings, and the descriptions that presuppose it, the uniquely solvable initial-value problem and the Markov property, inherit the assumption. For an irreducible finite-state Markov chain in its stationary law, the composition defect of its temporal window average vanishes at least linearly with the window width, which yields a one-sided test of memorylessness on the observed state space. The memoryless description is the special case that requires justification.

The Standard Model of Complex Economic Systems: A Unified Framework for the Adaptive Resolution, Reconstruction, and Evolution of Market Signals

(This paper presents the comprehensive theoretical and empirical framework underlying Adagio Group's investment strategies.)

We present the Standard Model of Complex Economic Systems, a purely theoretical framework in which the forecasting problem for financial signals factorizes through three stages. First, an adaptive spectral resolution, characterized axiomatically rather than by any basis fixed in advance, resolves the observed signal into frequency-ordered components and a residual. Second, each component is treated as a scalar observable of an underlying subsystem and lifted to a reconstructed state space by delay coordinates, with validity governed by embedding theory and, for stochastic components, by minimal Markovian realization. Third, each reconstructed component is evolved through the spectrum of a finite linear representation of its Koopman operator, and the composite forecast is recovered by recombination. We prove that the prediction error of any causal forecast admits an irreducible floor that depends on the process and the horizon and not on the architecture. The functional extensions of the Omega ratio and the Summers Total Risk-Adjusted Performance Measure serve as the framework's evaluation functionals. The framework is presented at the level of the architecture: particular numerical instantiations are proprietary, and no empirical results are reported.

An Intuitive True Total Risk-Adjusted Performance Measure and Characteristics Matrix

Keating and Shadwick’s Omega ratio captures all statistical moments of return distributions but has practical limitations with respect to portfolio optimization. Kapsos et al. simplified the Omega ratio into an expression with better practical application, but its focus as a portfolio optimization tool represents a trade-off with respect to absolute risk-adjusted performance measurement. First, we've transformed the Kapsos form of the Omega ratio in an analogous manner to what Modigliani did with the Sharpe ratio to create an intuitive percentage output scaled against the market. Second, we’ve deconstructed the Kapsos form of Omega to create a 3 x 1 matrix that measures and intuitively communicates the three fundamental characteristics that define an investment: risk, return, and liquidity.

Be a Better Fiduciary: Private Structured Products & Quantitative Risk Analytics for Financial Advisors

Most financial advisors adhere to a very traditional asset allocation model built entirely upon public securities. Outside of the fact that a set of relatively vague, qualitative criteria govern the literal value of their clients’ life work, the substance upon which those models are predicated is a set of assets completely dependent upon schizophrenic secondary markets...

Investment Banking for Private Real Estate Operators

Investment banks are intermediaries that, amongst many other functions, help typically large companies raise capital by advising on and underwriting new securities issues. To prepare for a new issue of securities, investment banks first advise their clients on considerations such as capital structure (how much debt vs. equity should be issued; what types of equity and debt should be issued, etc.), the strategic use of other financial instruments (such as warrants), and...

Investment Clubs: Gain the Exclusive Access of the Top 1%

Constructing Alpha: An Introduction to the Fundamentals of Risk

From Wall Street to Main Street

Introduction to Options in Real Estate

Determining Equilibrium Value for Residential Real Estate

Deciphering Monetary Policy as a Means to Beat the Market

Consonance vs. Dissonance: A Physical Description