Improving the Robustness of Trading Strategy Backtesting with Boltzmann Machines and Generative Adversarial Networks

Discriminating modelling approaches for Point in Time Economic Scenario Generation

“…Classically, comparison between the similarity of distributional properties relies on statistical measures such as QQ Plots, Tail Behaviour measures, correlation measures etc as used in [11] or [30]. These approaches however might face a variety of problems as described in [18], given by different usages of the scenario generation (see examples from introduction), be it optimization with respect to certain options or a portfolio performance.…”

“…Additionally, it might also be interesting to verify the forecasting quality of an ESG, possibly by using its outputs for a stop loss strategy backtest as indicated in [30]. Difficulties arise within the choice of the stop loss criterion, which should be fitted to the modelled condition to utilize sufficient information from the model output to be a valid indicator for the model performance.…”

“…Similarly focusing solely on the use of generative models is [43], who research the performance of a conditional GAN to produce realistic scenarios to use on trading strategy calibration and aggregation. Another approach is given in [30] to compare the scenario generating abilities between a conditional RBM and a conditional Wasserstein GAN for the joint unconditional distribution of the S&P500 and VIX Indices, with conditional information consisting of 20 day previous data of both index values. The possibility of using the VIX Index as a forward looking indicating conditioning factor is only briefly touched upon with a recommendation for future research.…”

Scenario generation for market risk models using generative neural networks

“…As financial data, at least for liquid instruments, is consistently available, GANs are used in various fields of finance, including market prediction, tuning of trading models, portfolio management and optimization, synthetic data generation and diverse types of fraud detection, see Eckerli [2021]. Henry-Labordere [2019], Lezmi et al [2020], Fu et al [2019], Wiese et al [2019], Ni et al [2020] and have already used GANs for scenario generation in the financial sector. The focus of their reasearch was the generation of financial time series for a limited number of risk factors (up to 6) and / or a single asset class.…”

Simulating Multi-Asset Classes Prices Using Wasserstein Generative Adversarial Network: A Study of Stocks, Futures and Cryptocurrency

“…From the asset classes perspective, the Hong Kong stock market has less liquidity due to policy risks and expensive transaction fees of around 0.1477% of total transaction value per side, including buying and selling. Lezmi et al (2020) From the trading frequency perspective, simulating tick data, minute-level data, and daily-level data will enhance the robustness and scalability for multiple trading frequencies. Some asset managers prefer to buy and hold strategies, and others prefer transactions at a weekly frequency.…”