Artificial market making with neural nets: an application to options

Abstract

Empirical research on option pricing has uncovered systematic deviations between market prices and the predictions of the well-known Black-Scholes formula (Rubinstein, 1985). If the Black-Scholes model were true, then the market prices of all options on the same underlying asset would correspond to the same Black-Scholes implied volatility. In fact, Black-Scholes implied volatility varies with time to expiration and strike price, a phenomenon commonly known as the "volatility smile". The aim of our research is to test whether neural nets are able to predict bid-ask spreads, by examining the market for S&P 500 index options. Subsequent research will expand the problem to simultaneously predict the price and the bid-ask spread. We describe the data and summarize previous findings concerning the dependence of the bid-ask spread on various inputs.