Question by RoastGoose: Options Trading Algorithm……?
how would i best go about developing an options trading algorithm?
Barry – i’m not sure what you meant by overriding strategy. Make as much money is the shortest time possible? :\
Would this be the general outline for starting from scratch?
1. Analyze the underlying –> Backtest correlations between the underlying and TA, FA and sentiment factors
2. Use results from 1 to predict the implied volatility and trend (?) of the underlying
3. Based on predictions, use the option strategy that’s most likely to generate the best profits
Thanks for your feedback
Best answer:
Answer by Barry789
There are different uses of options, not all of which are concerned with making the most money from the option trade.
You can use them to speculate (outright naked buy and naked sell of puts and calls), increase income (covered call writing), protect positions (buying puts on a long position that has profit), etc. There are a lot of options and stock, long and short, that you can use to replicate any outcome you want, assuming your assumptions on market and stock movement and volatility are correct.
Once you decide what you are trying to do, then you should back test the price movements to see if there are any tactics that work better than others *in the market conditions you are predicting*.
The second step in back testing is to extend the range of favorable and adverse price movements to see when the tactic would have turned against you. The troubles that LTCM had were a result of not testing their strategy into what amounts to 5 and 10 standard deviation events.
I wouldn’t worry too much about implied volatility. In the short term it doesn’t change much. You should know what it is today and what it has been in the past, but the volatility won’t be as big a factor as the movement of the underlying and its market.
You should understand that the distribution of price changes is not normally distributed, which is the assumption of most of the models people use. Mandelbrot’s work has demonstrated that these things are best understood using the stable distribution. One interesting factoid is that in the stable (also called the Pareto distribution), the standard deviation is undefined.
The assumption of normality or log-normality implies that all events occur within +/- 3 standard deviations from the norm and that events like the ones that buried LTCM and the events of 10/19/87 were once in a 3,000 years events. If you look carefully at price changes, you will see that there are a lot of outliers, price changes that don’t fit within the models, but which contain information.
Know better? Leave your own answer in the comments!