pith. sign in

arxiv: 1907.03093 · v1 · pith:OJ7E6FKFnew · submitted 2019-07-06 · 💱 q-fin.PM · math.OC· q-fin.MF

Dynamic Mean-Variance Portfolio Optimisation

Xiang Meng This is my paper

Pith reviewed 2026-05-25 01:46 UTC · model grok-4.3

classification 💱 q-fin.PM math.OCq-fin.MF
keywords dynamic mean-variance optimizationtime-consistencygame-theoretic approachCEV modelportfolio optimizationtime-inconsistent strategiesmarket data fitting
0
0 comments X

The pith

A game-theoretic approach produces time-consistent strategies for dynamic mean-variance portfolio optimization.

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

The paper addresses time-inconsistency in dynamic mean-variance portfolio optimization by using a game-theoretic method rather than precommitment. This method yields a strategy that an investor can stick to without later regret. The strategy is then extended to economies where asset prices follow constant elasticity of variance dynamics. Testing on real market data and simulations shows that strategies whose assumptions match actual market conditions tend to perform better. The approach is also compared to a deep learning derived strategy.

Core claim

By applying the game-theoretic approach, a time-consistent optimal strategy for the dynamic mean-variance optimisation problem is derived. This strategy is extended to a CEV-driven economy. When fitted to both real market data and simulated data, the strategy performs better when its model assumptions are close to market conditions. A selected strategy is compared with one obtained via deep learning technique.

What carries the argument

The game-theoretic approach to resolving time-inconsistency in dynamic mean-variance optimization.

Load-bearing premise

That the game-theoretic method correctly produces a strategy that is time-consistent and that closeness of assumptions to market data reliably predicts better performance.

What would settle it

A backtest where the game-theoretic strategy shows no improvement or worse results compared to alternatives when its assumptions are aligned with the data.

Figures

Figures reproduced from arXiv: 1907.03093 by Xiang Meng.

Figure 4.1
Figure 4.1. Figure 4.1: Performance of the static MVO strategy, with different target return period (around October 2010) when they announced the successful experiment of unmanned vehicles [13]. The following plunge is due to the continuous holding of Google (it is a common rule in finance that the price must have a rebound after a continuous surge). Although we did not quit the market at the correct time, this still proved the… view at source ↗
Figure 4.2
Figure 4.2. Figure 4.2: Performance of the DMVO strategy in a GBM economy [PITH_FULL_IMAGE:figures/full_fig_p032_4_2.png] view at source ↗
Figure 4.3
Figure 4.3. Figure 4.3: Performance of the DMVO strategy in a CEV economy tiable. Overall, figure 4.3 shows that the strategy assuming CEV with α = 1 is the only one that generates positive wealth in the market. Furthermore, U-shape on the red curve proves the effect of hedging demand. We now proceed to present some interesting findings regarding the CEV application to the real world. We first show the result of CEV performance… view at source ↗
Figure 4.4
Figure 4.4. Figure 4.4: Application of CEV strategy, under real market data and it is indeed the case for the demonstration here. Recall that stocks which follow the CEV process have the property: The stock price has a heavy left tail when α < 0, and has a heavy right tail when α > 0 (Cox, 1996) [10]. We summarize the market index price as the weekly average (assume equal weight for each stock), and plot the histogram in figure… view at source ↗
Figure 4.5
Figure 4.5. Figure 4.5: Summary of market index price Certainly, the result has its limitation: Not only we did simplification to models, but also the data has bias itself. In order to obtain a more robust conclusion, data from different market (outside US) and diversified range (not restricted to top 20) could be chosen for separate experiments. 30 [PITH_FULL_IMAGE:figures/full_fig_p034_4_5.png] view at source ↗
Figure 4.6
Figure 4.6. Figure 4.6: Comparison between dynamic and static, under real market data 31 [PITH_FULL_IMAGE:figures/full_fig_p035_4_6.png] view at source ↗
Figure 5.1
Figure 5.1. Figure 5.1: The LSTM structure for time series [14] 34 [PITH_FULL_IMAGE:figures/full_fig_p038_5_1.png] view at source ↗
Figure 5.2
Figure 5.2. Figure 5.2: Prediction vs actual data for Keppel Corp [PITH_FULL_IMAGE:figures/full_fig_p039_5_2.png] view at source ↗
Figure 5.3
Figure 5.3. Figure 5.3: The distribution of selected stocks from SGX 35 [PITH_FULL_IMAGE:figures/full_fig_p039_5_3.png] view at source ↗
read the original abstract

The portfolio optimisation problem, first raised by Harry Markowitz in 1952, has been a fundamental and central topic to understanding the stock market and making decisions. There has been plenty of works contributing to development of the mean-variance optimisation (MVO) so far. In this paper, one kind of them, namely, dynamic mean-variance optimisation (DMVO) is mainly discussed. One can apply either precommitment or game-theoritical approach to address time-inconsistency in DMVO. We use the second approach to seek for a time-consistent strategy. After obtaining the optimal strategy, we extend the result to a CEV-driven economy. In order to prove the usefulness of them, strategies are fit into both real market data and simulated data. It turns out that the strategy whose assumptions are close to market conditions generally gives a better result. Lastly, a selected strategy is chosen to compare with another strategy came up by deep learning technique.

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

1 major / 0 minor

Summary. The paper addresses time-inconsistency in dynamic mean-variance portfolio optimization (DMVO) via the game-theoretic approach to derive a time-consistent strategy. It extends the result to a CEV-driven economy. Usefulness is demonstrated by fitting the strategies to real market data and simulated data, with the conclusion that strategies whose assumptions are closer to market conditions generally perform better. A selected strategy is compared to one obtained via deep learning.

Significance. If the empirical validation is rigorous, the work could offer practical guidance on model selection for DMVO by showing the benefit of matching assumptions to market conditions. The game-theoretic derivation and CEV extension follow standard techniques in the literature, so significance rests primarily on the quality and transparency of the data-fitting exercise and the deep-learning comparison.

major comments (1)
  1. [Empirical fitting and comparison section] The central empirical claim (that strategies with assumptions close to market conditions give better results) is load-bearing for the usefulness conclusion yet rests on unspecified elements: data sources and periods, definition of 'fit' (calibration vs. backtest), performance metric, out-of-sample protocol, and statistical tests. This directly affects the reliability of the practical-value assertion.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive report and the opportunity to clarify the empirical aspects of our work. We address the single major comment below.

read point-by-point responses

  1. Referee: [Empirical fitting and comparison section] The central empirical claim (that strategies with assumptions close to market conditions give better results) is load-bearing for the usefulness conclusion yet rests on unspecified elements: data sources and periods, definition of 'fit' (calibration vs. backtest), performance metric, out-of-sample protocol, and statistical tests. This directly affects the reliability of the practical-value assertion.

    Authors: We agree that additional methodological transparency is warranted. In the revised manuscript we will expand the empirical section to explicitly state: the precise data sources and sample periods used for both real-market and simulated experiments; the distinction between parameter calibration on in-sample data and subsequent backtesting; the concrete performance metrics (e.g., realized mean-variance utility or Sharpe ratio); the out-of-sample protocol (including any rolling-window or fixed-split design); and the statistical tests employed to assess differences across strategies. These additions will allow readers to evaluate the robustness of the claim that strategies whose assumptions align more closely with observed market conditions perform better, without altering the reported numerical results. revision: yes

Circularity Check

0 steps flagged

No circularity in derivation chain; empirical validation is post hoc

full rationale

The paper first derives the time-consistent strategy via game-theoretic approach for DMVO, then extends the result mathematically to a CEV-driven economy. These steps are presented as independent of the subsequent data fitting. The fitting to real/simulated data is described only as a way to 'prove usefulness' and compare performance, not as an input that defines or forces the derived strategy. No equations, self-citations, or ansatzes are quoted that reduce the central claims to fitted parameters or prior self-work by construction. The paper is therefore self-contained on the mathematical side against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No information available from the abstract regarding free parameters, axioms, or invented entities used in the derivation.

pith-pipeline@v0.9.0 · 5685 in / 1216 out tokens · 33294 ms · 2026-05-25T01:46:39.705061+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

14 extracted references · 14 canonical work pages

  1. [1]

    Markowitz, H. (1952). Portfolio selection. The journal of finance, 7(1), 77-91

    work page 1952

  2. [2]

    Richardson, H. R. (1989): A Minimum Variance Result in Continuous Trading Portfolio Optimization, Manage. Sci. 35(9), 1045-1055

    work page 1989

  3. [3]

    Li, D., & Ng, W. L. (2000). Optimal dynamic portfolio selection: Multiperiod mean-variance formulation. Mathematical Finance, 10(3), 387-406

    work page 2000

  4. [4]

    (1955): Myopia and Inconsistency in Dynamic Utility Maximization, Rev

    Strotz, R. (1955): Myopia and Inconsistency in Dynamic Utility Maximization, Rev. Econ. Stud. 23, 165-180

    work page 1955

  5. [5]

    Basak, S., & Chabakauri, G. (2010). Dynamic mean-variance asset allocation. The Review of Financial Studies, 23(8), 2970-3016

    work page 2010

  6. [6]

    & Murgoci, A

    Bjork, T. & Murgoci, A. (2009). A General Theory of Markovian Time Inconsistent Stochastic Control Problems

    work page 2009

  7. [7]

    Bjork, T., Murgoci, A., & Zhou, X. Y. (2014). Mean-variance portfolio optimization with state-dependent risk aversion. Mathematical Finance, 24(1), 1-24

    work page 2014

  8. [8]

    Sharpe, W. F. (1966). Mutual fund performance. The Journal of business, 39(1), 119-138

    work page 1966

  9. [9]

    Lindsay, A., & Brecher, D. (2010). Results on the CEV process, past and present

    work page 2010

  10. [10]

    Cox, J. C. (1996). The constant elasticity of variance option pricing model. The Journal of Portfolio Management, 23(5), 15-17

    work page 1996

  11. [11]

    Ait-Sahalia, Y., & Lo, A. W. (1998). Nonparametric estimation of state?price densities implicit in financial asset prices. The Journal of Finance, 53(2), 499-547

    work page 1998

  12. [12]

    T., & Litzenberger, R

    Breeden, D. T., & Litzenberger, R. H. (1978). Prices of state-contingent claims implicit in option prices. Journal of business, 621-651. 40

    work page 1978

  13. [13]

    Markoff, J. (2010). Google Cars Drive Themselves, in Traffic. Retrieved from: https://www.nytimes.com/2010/10/10/science/10google.html

    work page 2010

  14. [14]

    Piculjan, N. (2017). Introduction to Deep Learning Trading in Hedge Funds. Retrieved from https://www.toptal.com/deep-learning/deep-learning-trading-hedge-funds 41

    work page 2017