Mettle et al 2024
“Analysis of Investment Returns as Markov Chain Random Walk” is a peer-reviewed open-access article by Felix Okoe Mettle, Emmanuel Kojo Aidoo, Carlos Oko Narku Dowuona and Louis Agyekum (the first three affiliated with the Department of Statistics and Actuarial Science, University of Ghana), published in the International Journal of Mathematics and Mathematical Sciences (Wiley/Hindawi, 2024, Article 3966566; DOI 10.1155/2024/3966566). It is included in this vault as a clear example of how a discrete Markov Chain Trading Model is, in much of the academic literature, used purely descriptively — to characterise and rank markets — rather than as the engine of a tradeable strategy.
The model is a “Markov chain random walk.” Its defining choice is the state definition: rather than the usual fixed up/flat/down triple, the state is the number of consecutive same-direction changes in the investment return, giving a countably infinite state space (a run of one up-month, two up-months, three, and so on, and symmetrically for down-months). The transition matrix is estimated from the empirical frequencies of these consecutive return changes, and standard Markov-chain machinery is then applied: powers P^t for t-step probabilities, the limiting (steady-state) distribution for an aperiodic irreducible chain, mean recurrence times derived from the limiting probabilities, and a defined six-month moving “crush” probability capturing the chance of adverse runs. The data are 450 monthly returns spanning January 1976 to December 2020 for five countries — Canada, India, Mexico, South Africa and Switzerland — drawn from the Federal Reserve Bank of St. Louis (FRED).
Mettle et al 2024 [proposes_model] Markov Chain Trading Model Mettle et al 2024 [relates] Regime Classification Mettle et al 2024 [defines] State Definition Arbitrariness
The finding is comparative and descriptive: the Mexican market had the lowest limiting probabilities for all negative states and the highest for positive states, while the Indian market was the reverse — so over the 1976-2020 window “the Mexican market performed better than the others … whilst India performed poorly.” The authors frame the output explicitly as information “for market regulators and investors in setting regulations and decision-making” — a market-quality / risk-assessment tool, not a timing signal. The paper itself describes the model as a way to “assess the performance of the markets,” not to forecast a tradeable entry or exit.
Mettle et al 2024 [supports] Regime Classification
For grading this is unambiguous: there is no trading strategy, no backtest, no positions, no P&L, no transaction costs, no out-of-sample test, and no trading benchmark. The only “benchmark” is the five markets compared against one another. A finding that Mexico’s chain had favourable steady-state probabilities over 1976-2020 is a retrospective descriptive statistic; it carries no claim — and supports no inference — that trading on the chain would have been profitable, especially after Transaction Costs and Slippage. The correct grade is therefore inconclusive: the paper is a legitimate, peer-reviewed application of Markov chains to Regime Classification / market-risk ranking, and is genuinely useful for that descriptive purpose, but it provides no evidence of tradeable alpha and makes no such claim. It should be cited as confirmation that the discrete Markov chain’s robust, defensible use is descriptive, not as support for any profitability claim.
Transaction Costs and Slippage [contradicts] Mettle et al 2024 Mettle et al 2024 [contradicts] Markov Chain Trading Model
Connections
- Markov Chain Trading Model — proposes_model, 2024, source: https://onlinelibrary.wiley.com/doi/10.1155/2024/3966566
- Markov Chain Trading Model — detects_regime, 2024, source: https://onlinelibrary.wiley.com/doi/10.1155/2024/3966566
- Regime Classification — detects_regime, 2024, source: https://onlinelibrary.wiley.com/doi/10.1155/2024/3966566
- State Definition Arbitrariness — relates, 2024, source: https://onlinelibrary.wiley.com/doi/10.1155/2024/3966566
- Random Walk Benchmark — relates, 2024, source: https://onlinelibrary.wiley.com/doi/10.1155/2024/3966566
- Out-of-Sample Backtesting — lacks_live_evidence, 2024, source: https://onlinelibrary.wiley.com/doi/10.1155/2024/3966566
- Transaction Costs and Slippage — excludes_costs, 2024, source: https://onlinelibrary.wiley.com/doi/10.1155/2024/3966566