Initiated the Structure Arbitrage activity at Tykhe

The comparison between the Credit and the Equity Volatility worlds easily shows strong behavior likeness. A structure arbitrage activity between the two classes of products followed, which, unlike the usual approach, isn't based on deep puts or fundamental analysis.

After having parsed without success the research literature, I started my own statistical analysis. The preliminary results were interesting enough to justify a full-scale study:

 

Data collection

Contacted several data providers and effectively negotiated data acquisition, before going through a deep cleaning process.  

Alphas detection

The amount of data justifying an efficient environment, MatLab was selected to develop the analysis and simulation infrastructure.

This allowed the selection of the additional alphas, required for a proper description of the phenomenon. 

Analysis & Modelization

The relation between the different assets is well described with a descriptive statistical approach, bringing the following results: 

  • The at-the-money implied volatility of a single stock can be well described by a mix of the company's CDS, stock price, and a several other tradable instruments. 

  • This relationship is stable both through cross-sectional analysis (sub-groups of companies), as well as across time. The description parameters are stable, proving the effective existence of a mathematical relationship behind the data.

  • The R2 of the descriptive model is bound within a range of 60-80%, low enough to generate arbitrage opportunities, but high enough to avoid large marked-to-market fluctuations.

  • The residues, of 3 to 20 vol points in size and fully tradable, are mean reverting at a rate of 20% per week. Since the trading costs are minimal, the arbitrage strategy is viable.

This relationship can be used not only as an indicator of future implied volatility, but also an effective skew model - describing precisely why some stocks deserve a larger skew than others and quantifying the appropriate level. 

Trading Simulation

Whether as a volatility directional or a skew model, bets can be taken isolated (on a single stock approach), or by spreading them within sectors. In all cases, the strategy is largely scalable in size

Although the strategy was proven valuable in its initial form, more work needs to be done to determine which of the trading approaches offers with the best Sharp ratio and other risk measures.