1 min read

Quantifying Mud

There is an urge that I and many highly discretionary traders are guilty of: taking muddy inputs that make our business what it is and trying to "clean it up" with quantification. Assigning numerical values for every sub-factor of a trade that you can then run regressions on to achieve some incredible insight of: "oh ya you kept  doing x or y".

The saying goes your outputs are only as good as your inputs, a lesson that is increasingly important in the world of big data and modeling everything. However, the average discretionary trade is a summation of both explicit (market-generated information) and implicit (experience, subconscious intuitions, subjective patterns). This means whatever model you try to build is going to be filled with subjective interpretations of both those implicit and explicit factors, and those interpretations will in most cases take place in a semi or fully-tilted frame of mind. (Grading a trade after you get your ass handed to you will not yield objective remarks.)  That is to say, your inputs are at best – muddy.

Then there are the outputs and your subjective interpretations and course of action. The model says you are taking too many trades with a narrative < 3... okay so now only look for narratives >3. Ezpz. Oh wait... those were graded with hindsight and impacted by the outcome of the trade, and look at all those <3 trades that worked out beautifully. Surely if I had just taken them I would have made some green. Yeah... okay.

You are discretionary when your toaster now has AI. Still track core metrics and don't throw the baby out with the bathwater. But embrace the mud.