The Financial Industries Only Performance Based Publication on global markets which include futures and stock indicies. We are never compensated in a manner that approaches the value we bring to the investment performance of our clientele."

“The Infirmities of Algorithms”

“The Infirmities of Algorithms”

Algorithm’s- “Why they do not work over the long haul”

“An Engineer’s Perspective”-

Algorithms are used to process data for financial trades to recommend buy, sell, hold, etc. The actual algorithm is developed by a person so the algorithm is only is good as the person writing it and the data set that person feels is sufficient to execute the algorithm. So the decision to b/s/h becomes a black box approach after the data is entered.

This approach is sufficient and reasonable for simple desired results where there is no “human factor” involved. For example, if a baker wants to bake 1, 5, or 10 cakes, an algorithm can tell him the quantities of flower, sugar, yeast and other ingredients to make the desired quantity of cakes.

Well, what if one day I decide I want sweeter or a more buttery flavored cake and I go to the bakery and buy my cake without having any input on the baker’s recipe, I will be dissatisfied because the baker is making the cakes for the masses and not just me.  The human factor is missing here.

Market movement is intrinsically based on the human factor. There is a terrorist attack in the Middle East and the price of oil jumps followed by gas prices. Is there any rational reason for it when the oil fields are producing the same the day before and the days after the attack? Did someone not trust the algorithm and manipulate the b/s/h results? How could you program in the magnitude of the event and peoples response to the attack?

Additionally, we all know that the data sets are influenced by the human factor and they change all the time based on how society may feel this year versus next year on any given social issue and what is of prime importance now may fall to the bottom of the list next year.

Look at the pollsters predictions in the last presidential election for example. No one besides Trump (and not surprisingly) was predicting a Trump victory. Here is a prime example of how algorithms used to predict the election outcome in favor of Hillary Clinton was grossly influenced by a few people who clearly had entered biased data into an algorithm and we know how that turned out.

Short term use of algorithms in general is far more accurate than long term use of the same algorithm because the data sets and the human factor are constantly changing. You can perform data mining to routinely input new data and also revise the algorithm but how do you predict the human factor that can change in an hour, a day, or a week?  The only thing that is constant is change itself.

A very smart guy once told me that “we don’t tell the markets what to do, the markets move based on X and the human factor.” Being able to predict market movements is more of an art than a science, which is the exact opposite of an algorithm.  You need to watch market trends for sure but more importantly you need to understand how the market trend is/will be affected by the human factor.

Having a great mind focusing on a fewer number financial opportunities will yield far better predictions and profits than massive amounts of trading signals that are based on algorithms. Take your doctor for example. Do you want to go to the doctor’s office to be seen by a nurse or do you want to see the doctor or a specialist? The doctor/specialist has far greater knowledge than a nurse and the same analogy applies to the future direction of the financial markets.  Individual attention by an expert will yield far better results.

In the end algorithms are a weapon of math destruction… it is just not that easy.  Until humanity is reduced to little more than drones this will never change.  

Colin J. Dahlen P.E. Engineer  University of Massachusetts (Summa Cum Laude) -M.S.C.E.

Leave a Reply