This paper evaluates the profitability of applying four different volatility forecasting models to the trading of straddles on the German stock market index DAX. Special care has been taken to use simultaneous intra-day prices and realistic transaction costs. Furthermore, straddle positions were evaluated on a daily basis to preserve delta neutrality. The four models applied in this paper are: historical volatility, two ARCH models, and an autoregressive model for the volatility index. VDAX. The ARCH models perform best in generating profits for market makers. Forecasts based on historical volatility also produce statistically and economically significant profits over the two-year simulation period of 1993 and 1994. In general, a filter1rule with a small filter of0.5 per cent produces the best results for both the ARCH models and historical volatility. However, the VDAX-AR model generates much lower and usually insignificant profits, and for some filter rules this model even has cumulative losses for market makers. For non-market-makers and non-members of exchange, however, larger transaction\costs imply that no significant profits can be gained with any model of volatility forecasts.