Markets continue to drift sideways, and so I thought today, I might finish my discussion about macro funds and model trading. The comments from you yesterday seemed to reflect a healthy scepticism about model building, and in that, you reflect my own feelings. I have been involved in the building of FX models, and I know just how much work and attention to detail these endeavours require, and I also understand that there is never a finished product. All models need constant attention to understand and monitor how they are coping with a changing marketplace. Rather like changes over the years to a golf course, after a period, so many piecemeal changes are made, that the original concept can be lost. So it is with a single model platform, tinkering with the many filters constantly can sometimes improve the short term performance, but over a while, the changes can radically change the original ethos of the model.
One approach, and in my opinion the most workable (also requiring the most work!) is to build a suite of fairly basic models, all of which have credibility/stand alone characteristics of their own, and then learn how to manage the performance of the suite as a whole. This approach is common enough to have some merit, and is worth a quick look. At the basic level, you could have 12 models, and a minimum requirement to have 6 of them in phase for the concept to be active. Each model would have a `corridor` of past performance to indicate when it was going out of phase, and would be withdrawn from active duty if performance showed signs of deterioration. At any one stage, there should be enough models performing within their known parameters for the offering to be credible – rather like substituting footballers when their performance tails away, one comes out and (hopefully) another comes in. In the above scenario, if fewer than 6 are active, then the system temporarily goes into cash, because a message is coming loud and clear that there are forces around that are pretty unusual. As you will have gathered, this is no easy concept. The way that the inclusion of a model in a suite can either calm or increase volatility – almost irrespective of the individual performance of that particular model is fascinating – again from a football analogy, there are certain players that complement each other, and others that you just can`t play together because their characteristics are too similar.
So now to try and summarise thoughts about FX modelling. There is no doubt that this style of trading is alive and well, and in its many different guises, has many supporters; the work required to build and nurture models is huge; model based trading must be separate from discretionary trading, but the two styles can work very well side by side; in discretionary/prop trading all the thought goes into a position before the deal is done ie. what would constitute a change of heart, where are the exit levels, when, if successful do I increase the position etc. All this is done to make as little as possible a surprise, because surprises can make you act irrationally – in a model environment this is precisely what you are trying to build.
Nobody said it was easy, so in this case, nobody was wrong…..