Skip to main content
Labs
FINANCEGPT LABS RESEARCH · HISTORICAL CASE STUDY

Synthetic options chains and portfolio research in sparse-data markets.

This 2024 FinanceGPT Labs research case study describes an experiment using a VAE-GAN framework to generate synthetic options-chain features where observed options data was sparse, then compares a portfolio backtest with and without those synthetic features.

In one sentence: This is a historical research and backtest case study. It is not a live performance claim, investment forecast or representation that synthetic data is observed market data.

Portfolio construction and management from natural-language objectivesEquityGPT connects objectives, research, portfolio construction, risk, review and rebalancing. Portfolioobjective + constraints Objective Research Construct Risk Review Rebalance
ANSWER-FIRST DEFINITION

This is a historical research and backtest case study. It is not a live performance claim, investment forecast or representation that synthetic data is observed market data.

THE RESEARCH QUESTION

Can synthetic derivatives features improve research where observed options data is sparse?

The source case study focused on the JSE and described a VAE-GAN approach for producing synthetic options-chain information from sparse historical data and correlated market inputs.

Sparse observed options data
Synthetic feature generation
Quantitative validation constraints
Portfolio backtest comparison
HISTORICAL BACKTEST

The research backtest reported 50.48% versus 42.46%.

Those figures describe the historical research backtest only; they are not current returns, expected returns or live customer performance.

Synthetic-feature portfolio: 50.48% source-reported backtest return
Baseline portfolio: 42.46% source-reported backtest return
Historical research only
Synthetic and observed data must remain visibly distinct
MODEL FRAMEWORK

Generate, critique and validate rather than treating synthetic data as observed fact.

The source page described a Variational Autoencoder for probabilistic generation and an adversarial validation layer, with financial constraints such as put-call relationships used as part of the realism checks.

Ingest sparse historical data
Generate candidate synthetic features
Validate quantitative consistency
Keep provenance and synthetic labels attached
HISTORICAL BACKTEST VISUAL

Source-reported portfolio comparison.

The visual below reproduces the two backtest figures stated in the supplied historical case-study page. It is not a forecast or a live-performance claim.

Synthetic-feature portfolio50.48%
Baseline portfolio42.46%
FINANCEGPT APPLICATIONS

Use the application designed for the work.

FinanceGPT supports financial creation; EquityGPT supports portfolio construction and management; FinanceGPT Labs provides the complete platform; FinanceGPT Developers provides APIs and model tooling; FinanceGPT Tools provides focused utilities; FinanceGPT Chat provides conversational finance.