Paper: arXiv 2310.04536
Authors: Piotr Pomorski, Denise Gorse
Abstract
This work extends a previous work in regime detection, which allowed trading positions to be profitably adjusted when a new regime was detected, to ex ante prediction of regimes, leading to substantial performance improvements over the earlier model, over all three asset classes considered (equities, commodities, and foreign exchange), over a test period of four years. The proposed new model is also benchmarked over this same period against a hidden Markov model, the most popular current model for financial regime prediction, and against an appropriate index benchmark for each asset class, in the case of the commodities model having a test period cost-adjusted cumulative return over four times higher than that expected from the index. Notably, the proposed model makes use of a contrarian trading strategy, not uncommon in the financial industry but relatively unexplored in machine learning models. The model also makes use of frequent short positions, something not always desirable to investors due to issues of both financial risk and ethics; however, it is discussed how further work could remove this reliance on shorting and allow the construction of a long-only version of the model.
Complexity vs Empirical Score
- Math Complexity: 6.0/10
- Empirical Rigor: 8.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced statistical and machine learning concepts (Markov switching models, random forests, feature selection with BorutaShap) and presents a rigorous backtest with multiple asset classes, realistic transaction costs, and benchmark comparisons.
Research Flowchart
flowchart TD A["Research Goal: Ex Ante Prediction of<br>Financial Regimes to Improve Portfolio Performance"] --> B["Data Inputs: Equities, Commodities, Forex<br>4-Year Test Period"] B --> C["Methodology: Novel ML Model<br>with Contrarian Strategy & Short Positions"] C --> D["Computational Process: Compare against<br>Hidden Markov Model & Benchmark Indices"] D --> E["Key Findings: Substantial Performance<br>Improvement over HMM & Benchmarks"] E --> F["Outcome: Over 4x Cumulative Return<br>vs Benchmark in Commodities"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a novel ex‑ante financial regime prediction model (KMRF) that extends prior regime‑detection work and uses a contrarian trading strategy across equities, commodities and FX.
Source sentence
This work extends a previous work in regime detection, which allowed trading positions to be profitably adjusted when a new regime was detected, to ex ante prediction of regimes, leading to substantial performance improvements over the earlier model, over all three asset classes considered (equities, commodities, and f
Assumes 6
- Transaction costs are as listed in Table 1 and are realistic estimates obtained from industry experts. execution
Source sentence
The costs assumed in this work are listed in Table 1; these are realistic estimates which were adopted after appropriate consultation with industry experts.
- Frequent short positions are allowed in the trading strategy. model
Source sentence
The model also makes use of frequent short positions, something not always desirable to investors due to issues of both financial risk and ethics
- Data are pre‑processed with fractional differencing to balance stationarity and non‑stationarity. data
Source sentence
after first having gone through scaling process known as fractional differencing [20], which balances stationarity and non-stationarity within time series.
- A contrarian interpretation of regime signals is used: bullish predictions trigger short positions, bearish predictions trigger long positions. execution
Source sentence
When a bullish regime is predicted, a short position should be assumed in the underlying asset. When a bearish regime is predicted, a long position should be assumed in the underlying asset.
Show 2 more
- Each asset class has a separate benchmark used only for out‑of‑sample performance comparison. data
Source sentence
Note that each asset class has a separate benchmark used only in out-of-sample model performance comparison
- All series, including benchmarks, end on the same date (29/04/2022). data
Source sentence
all series, including the benchmarks, end on the same date, 29/04/2022.
Data
multi-asset
- Three‑asset‑class price series (equities, commodities, FX) market-prices commercial
Source sentence
Regimes will be predicted for three asset classes: equities, commodities, and foreign exchange (FX) pairs. The component assets for each class are listed online [19], as well as the start dates of the data, after first having gone through scaling process known as fractional differencing
- Benchmark series for each asset class benchmark commercial
Source sentence
all series, including the benchmarks, end on the same date, 29/04/2022.
Costs
Modeled
Two‑way costs per trade: Equities 0.40 % total, Commodities 0.27 % total, Currencies 0.13 % total (brokerage, spread, market impact).
Source sentence
Table 1: Two-way (buy (long), sell (short)) trading costs for each asset class Asset class Brokerage commissions (%) Bid-ask spread (%) Market impact (%) Total cost (%) Equities 0.07 0.065 0.265 0.40 Commodities 0.14 0.13 0 0.27 Currencies (FX) 0 0.13 0 0.13
Finds 3
- The commodities KMRF model achieved a cost‑adjusted cumulative return over four times higher than its benchmark index. performance
Source sentence
in the case of the commodities model having a test period cost-adjusted cumulative return over four times higher than that expected from the index
- The equities KMRF model attained the highest annualised Sortino and adjusted Sharpe ratios among the three asset classes. performance
Source sentence
the equities KMRF model achieved the highest annualised Sortino and adjusted Sharpe ratios
- All three KMRF models outperformed their respective benchmarks and generated substantial risk‑adjusted returns net of costs. performance
Source sentence
each of the KMRF models, based on its generated signals, was able to outperform the selected benchmarks, as well as generate substantial risk-adjusted returns, net of costs
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 11 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.