It is quiet start to the markets on this lovely morning in London town, so on Friday I published a few thoughts about model based trading as an alternative/additional strategy to the conventional discretionary style in FX, and I would like to follow that up. It is necessary in a fund environment to recognise the difference between the two styles, and to keep the two disciplines totally segregated from each other. Having a model based strategy with an over-ride of discretion, is like being a little bit pregnant – It`s impossible; well, certainly impossible to sell! When a fund sells an individual offering, it can only be sold as a single defined strategy, simply because investors typically want a spread of different types of trading styles within a portfolio, and any deviation from a model based trading strategy makes it a discretionary offering; simple as that.
The problem is building a strategy that can be left alone to run – a mini `black box` scenario. Many traders that I know, run a single model, alongside a discretionary portfolio. Generally this model will have a lot of iterations and filters on it, and will therefore trade infrequently, coming alive only when multiple conditions necessary to satisfy the code are in line. Building this type of model can be very time consuming; simply to find the `right` combination of parameters to produce triggers at a manageable frequency can, unless you are very disciplined, often result in a model that is curve fitted, and under performs. The alternative however can easily be less attractive – a model that is simple in approach, but is very active, and very volatile. It may be profitable in back-testing, but the actuality of running it day to day needs too many filters, and this can quickly erode its usefulness.
When an idea comes to fruition, and a well behaved trading model is built, the final check for me is – does the strategy actually make sense? This is where I have a problem with complex single models, it is impossible to do anything other than rely on the mathematical methodology. Simple strategies that evolve with observation of price action – supplemented with anti volatility filters – can be subjected to a logical assessment to understand just why this combination might make sense! Take a very simple example: let`s assume we noticed from our observations and analysis, that there was a period during the day when the price pattern over a period of time, typically set the tone for the rest of the day, and by studying that small time period, we could take a position which after applying optimum filters (take profit/stop and reverse/end of day look back etc.) looked good. Before adopting this model, to me it is necessary to understand if the basis of reasoning makes some sort of sense! So if that model was in cable, and that time period was first thing in London trading time – no problem, a case can be made for the model.
Many readers will have developed their own successful model based styles of trading – others distrust the whole concept, and I have heard it likened it to bringing in management consultants (!); in other words, moving the decision making process to someone else! I would be interested to hear your views.