Intuition and Trading

There are numerous problems system designers face when tasked with developing IT Expert Based Systems. Such systems have to rely on a rule base from which inferences are made about a problem area(s), this is in a way the core idea behing an Expert System (ES). However the main problem is the acquirement of expertise or knowledge, if you like, into the rule base.

Numerous academic projects point out the fact that when traders were tasked with explaining some of their trading decision making processes, they were useless. The traders often explain that much of their decision making is based on intuition and that pin pointing explicitly, lhow they arrive to trading decisions is very difficult.

In a recent scientific experiment, two men with amnesia were asked to remember pairs of 8 objects (Nature July 28iss). Each man was shown the 8 pairs 5 times a day for numerous days. The man eventhought they could not remember why, learned to distibguish the pairs.

An excerpt follows:

When one of the men was asked if he picked an object because he remembered seeing it before, he said, "No. It just seems that's the one. It's here [pointing to his head] somehow or another and the hand goes for it."

Without being aware of it, the men learned the task after approximately 1,000 trials. This is about the same number of trials needed by monkeys with similar lesions in the medial temporal lobe. Normal monkeys take about 500 trials to learn the task.

The implication is that humans have an unconscious ability -- comparable to that in animals -- to mold our behavior in response to a consistent stimulus. "Habit learning is going on all the time. All of us are acquiring habits, " Squire said. "What this is showing is that habit learning is well-developed in humans, and that it works independent of consciousness."

This seems to indicate a strong relationship to intuitive learning, such knowledge that is acquired by traders via repetitive trading behaviour and performance feedback. In fact this is feedback learning and in IT terms it is synonymous with Unsupervised Learning, for cluster detection of patterns and inference. At least this is my interpretation, I am quite interested into past research of this kind... anybody who has a valuable opinion please bring your comments to the table :-)

Correlation

I just noticed today how good the wikipedia article on correlation is. This is worth a read, if not only for nice concise expressions, of the form such as "correlation or co-relation refers to the departure of two random variables from independence" and many others. Anyway, as we are talking statistics, do check out the highly related articles on Pearson's Correlation, and Standard Score.