pith. sign in

arxiv: 2607.01705 · v1 · pith:QZF3AQDSnew · submitted 2026-07-02 · 💱 q-fin.MF

Portfolio Optimization under Fast and Slow Latent Mean-Reverting and Momentum Drift

Dannin J. Eccles , Roger Lee This is my paper

Pith reviewed 2026-07-03 02:21 UTC · model grok-4.3

classification 💱 q-fin.MF
keywords portfolio optimizationpartial informationlatent factorsmean-reversionMACDexponential moving averageutility maximizationstochastic control
0
0 comments X

The pith

Filtered estimates of latent mean-reversion equal fast-minus-slow EMA differences plus a Volterra correction.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper studies portfolio optimization when an investor observes only prices but the asset drift is produced by two unobserved factors that mean-revert at two fixed but different speeds. It derives that the investor's best estimate of the current mean-reversion level is exactly the difference between a fast exponential moving average and a slow exponential moving average of past prices, together with one deterministic correction term that does not depend on the data. For investors who maximize log, power, or exponential utility, this estimate produces explicit rules for the optimal dollar amount held in the risky asset at each moment. The construction supplies a precise mathematical route by which MACD-type trading signals appear as the correct response to partial information about drift.

Core claim

Under partial information where the drift is driven by two latent factors at distinct time scales, the filtered estimate of the latent mean-reversion level equals the difference of fast and slow EMA processes of the price history plus a deterministic Volterra correction. This structure produces candidate optimal strategies in explicit feedback form for logarithmic, power, and exponential utility, together with admissibility and verification results. The results establish a mathematical foundation for the endogenous emergence of MACD-type signals as estimators of latent drift information contained in observed price paths.

What carries the argument

The difference between fast and slow exponential moving average processes of the trailing price history, which drives the filtered estimate of the latent mean-reversion level.

If this is right

  • Optimal investment strategies take explicit feedback form that depends on the MACD-type signal and the Volterra correction.
  • Admissibility holds for the derived strategies under logarithmic, power, and exponential utility.
  • Verification theorems confirm optimality of the candidate strategies.
  • MACD-type signals arise endogenously as the correct estimators of latent drift.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same reduction may appear in other control problems that involve filtering multiple latent scales from a single observed path.
  • One could test whether the Volterra correction remains negligible when the two time scales are close rather than widely separated.
  • The explicit form invites direct comparison of the derived rule against standard MACD implementations in controlled numerical experiments with known two-factor drifts.

Load-bearing premise

The risky asset drift is generated by exactly two latent stochastic factors that evolve at distinct, fixed time scales and the investor observes only the price path.

What would settle it

A direct computation of the filter under the two-factor model that yields an estimate different from the fast-minus-slow EMA difference plus the stated Volterra term would falsify the reduction.

read the original abstract

We consider a class of partial-information portfolio optimization problems in which the drift of a risky asset is driven by two latent stochastic factors evolving at distinct time scales. We show that the filtered estimate of the latent mean-reversion level is driven by the difference between fast and slow exponential moving average (EMA)-type processes of the trailing price history, yielding a Moving Average Convergence Divergence (MACD)-type signal, along with a deterministic Volterra correction. Under logarithmic, power, and exponential utility, we derive candidate optimal strategies in explicit feedback form and establish admissibility and verification results. In particular, the results provide a mathematical foundation for the endogenous emergence of MACD-type trading signals as estimators of latent drift information contained in observed price paths.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper considers a class of partial-information portfolio optimization problems in which the drift of a risky asset is driven by two latent stochastic factors evolving at distinct time scales. It shows that the filtered estimate of the latent mean-reversion level is driven by the difference between fast and slow exponential moving average (EMA)-type processes of the trailing price history, yielding a Moving Average Convergence Divergence (MACD)-type signal, along with a deterministic Volterra correction. Under logarithmic, power, and exponential utility, candidate optimal strategies are derived in explicit feedback form and admissibility and verification results are established.

Significance. If the filtering reduction and verification theorems hold, the work supplies a rigorous stochastic-control foundation for the endogenous appearance of MACD-type signals as optimal estimators of latent drift under a two-factor linear-Gaussian model observed through prices alone. The explicit feedback forms for standard utilities and the accompanying admissibility/verification results are concrete contributions to the partial-information portfolio literature.

major comments (2)
  1. [§3] §3 (filtering equations): the reduction of the filtered drift estimate to a two-dimensional Markov process consisting of fast EMA, slow EMA, and deterministic Volterra term holds only when the two latent factors are linear-Gaussian with known, fixed speeds; the manuscript must state explicitly that the Kalman-filter innovation process yields precisely this state without residual non-Markovian terms under these assumptions.
  2. [§4] §4 (verification theorems): the admissibility and verification arguments for the power-utility case rest on the candidate strategy remaining in the claimed explicit feedback form; the proof must confirm that the Volterra correction does not violate the integrability conditions used to justify the HJB solution.
minor comments (2)
  1. [§2] Notation for the two time-scale parameters should be introduced once in §2 and used consistently thereafter to avoid redefinition.
  2. [Abstract and §4] The abstract states that 'admissibility and verification results are established'; the corresponding theorems in §4 should include a brief statement of the precise integrability conditions imposed on the strategies.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading, positive evaluation, and constructive suggestions. We address each major comment below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [§3] §3 (filtering equations): the reduction of the filtered drift estimate to a two-dimensional Markov process consisting of fast EMA, slow EMA, and deterministic Volterra term holds only when the two latent factors are linear-Gaussian with known, fixed speeds; the manuscript must state explicitly that the Kalman-filter innovation process yields precisely this state without residual non-Markovian terms under these assumptions.

    Authors: The model is formulated under the linear-Gaussian assumption with known, fixed mean-reversion speeds. We agree that an explicit statement should be added to §3 clarifying that the Kalman-filter innovation process produces precisely the claimed two-dimensional Markov state (fast EMA, slow EMA, and deterministic Volterra term) with no residual non-Markovian components. This clarification will be inserted in the revised manuscript. revision: yes

  2. Referee: [§4] §4 (verification theorems): the admissibility and verification arguments for the power-utility case rest on the candidate strategy remaining in the claimed explicit feedback form; the proof must confirm that the Volterra correction does not violate the integrability conditions used to justify the HJB solution.

    Authors: We will augment the verification argument in §4 to include an explicit confirmation that the Volterra correction term satisfies the integrability conditions required for admissibility of the candidate strategy under power utility, thereby ensuring the feedback form remains valid for the HJB solution. revision: yes

Circularity Check

0 steps flagged

No significant circularity: derivation derives MACD form from two-factor filter rather than assuming it.

full rationale

The paper posits a two-factor linear-Gaussian latent drift model with fixed distinct speeds, observes only price, and derives that the Kalman-type filter for the mean-reversion level reduces exactly to fast-EMA minus slow-EMA (MACD) plus a deterministic Volterra integral. This is a direct consequence of the linear-Gaussian structure and known parameters, not a redefinition or fit. Optimal strategies and verification theorems are then obtained from the resulting finite-dimensional Markov state; the reduction is model-dependent but not circular. No self-citations, no fitted parameters renamed as predictions, and no ansatz smuggled via prior work appear in the provided text. The result is self-contained against the stated assumptions.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Abstract-only review; no explicit free parameters, axioms, or invented entities can be audited beyond the model setup stated in the abstract.

axioms (1)
  • domain assumption Asset drift is generated by exactly two latent stochastic factors evolving at distinct fixed time scales under partial information (price path only).
    This is the modeling premise stated in the abstract that enables the MACD reduction.

pith-pipeline@v0.9.1-grok · 5648 in / 1312 out tokens · 27952 ms · 2026-07-03T02:21:43.897310+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

16 extracted references · 1 canonical work pages

  1. [1]

    EMA-type trading strategies maximize utility under partial information , journal =

    Chen, Xiaodong and Lee, Roger , keywords =. EMA-type trading strategies maximize utility under partial information , journal =. 2023 , issn =. doi:10.3934/fmf.2023005 , url =

    work page doi:10.3934/fmf.2023005 2023

  2. [2]

    The Journal of Finance , volume=

    Asset pricing in a production economy with incomplete information , author=. The Journal of Finance , volume=. 1986 , publisher=

    1986

  3. [3]

    Stochastic processes and their applications , volume=

    Utility maximization with partial information , author=. Stochastic processes and their applications , volume=. 1995 , publisher=

    1995

  4. [4]

    Stochastic Processes and their Applications , volume=

    Optimal trading strategy for an investor: the case of partial information , author=. Stochastic Processes and their Applications , volume=. 1998 , publisher=

    1998

  5. [5]

    The Review of Financial Studies , volume=

    Dynamic nonmyopic portfolio behavior , author=. The Review of Financial Studies , volume=. 1996 , publisher=

    1996

  6. [6]

    Option pricing, interest rates and risk management , pages=

    Bayesian adaptive portfolio optimization , author=. Option pricing, interest rates and risk management , pages=. 2001 , publisher=

    2001

  7. [7]

    Mathematical Methods of Operations Research , volume=

    Optimal investment under partial information , author=. Mathematical Methods of Operations Research , volume=. 2010 , publisher=

    2010

  8. [8]

    Stochastic processes and their Applications , volume=

    Portfolio selection under incomplete information , author=. Stochastic processes and their Applications , volume=. 2006 , publisher=

    2006

  9. [9]

    International Journal of Theoretical and Applied Finance , volume=

    Portfolio optimization under partial information with expert opinions , author=. International Journal of Theoretical and Applied Finance , volume=. 2012 , publisher=

    2012

  10. [10]

    Journal of Applied Probability , volume=

    Portfolio optimization with unobservable Markov-modulated drift process , author=. Journal of Applied Probability , volume=. 2005 , publisher=

    2005

  11. [11]

    International Journal of Theoretical and Applied Finance , volume=

    Pairs trading of two assets with uncertainty in co-integration's level of mean reversion , author=. International Journal of Theoretical and Applied Finance , volume=. 2016 , publisher=

    2016

  12. [12]

    Communications in Mathematical Sciences , volume=

    Filtering and portfolio optimization with stochastic unobserved drift in asset returns , author=. Communications in Mathematical Sciences , volume=

  13. [13]

    Applied Mathematical Finance , volume=

    A mathematical analysis of technical analysis , author=. Applied Mathematical Finance , volume=. 2019 , publisher=

    2019

  14. [14]

    Finance and Stochastics , volume=

    Asymmetric information in fads models , author=. Finance and Stochastics , volume=. 2006 , publisher=

    2006

  15. [15]

    2008 , publisher=

    An introduction to stochastic filtering theory , author=. 2008 , publisher=

    2008

  16. [16]

    2009 , publisher=

    Continuous-time stochastic control and optimization with financial applications , author=. 2009 , publisher=

    2009