Analysis Pitfalls

Back in the chapter on market analysis methods we mentioned the representation bias, which is the tendency to view the portrayal of something as the thing itself. The example we used was a map, which shows the terrain, but is not the terrain itself. As such, it presents a limitation on one’s ability to use it in the decision-making process.

There are other biases as well.

The Lotto Bias relates to control. Its name comes from those who play the lottery and their belief that picking numbers somehow improves their chances of winning. If the odds of winning are 1 in a million for a given set of numbers to come up, it does not matter a bit whether those numbers are picked at random or by some kind of system.

Lotto bias relates to trading in regards to entry signals. Traders often spend enormous amounts of time trying to find the perfect system for getting in to the market when the fact of the matter is that for any given trade the likelihood of profiting is essentially dictated by chance. That is why we have said a few times that at least as much effort must go in to determining the exit strategy, to include money management. It is also why the trader must listen to the markets and not succumb to the idea that imposing one’s will on them is possible.

Closely tied to the lotto bias is something known as the Gambler’s Fallacy, which is the belief that a string of losses increases the chance of experiencing a winner. If one were to roll a standard six-sided die, the chance of getting a five is 1/6, or about 17%. It does not matter how many times in a row the die comes up with something other than five, the odds of the next roll being a five are 1/6.

The problem many people have is in thinking of the odds of a specific run happening and then extrapolating that in to some kind of odds for a single event. For example, there is about a 1/3 chance of getting 20 non-five rolls in a row on our six-sided die if we roll it 100 times. Does that mean that if we see 19 non-five rolls in a row that there is a 2/3 chance of getting a five on the next roll? No! Why? Because each roll is an independent event. The 1/3 odds were for seeing a run of 20 consecutive non-five rolls, not the odds of getting a non-five if there had already been 19 non-fives in a row.

This discussion is all based on discrete, independent events assuming no memory and a fair die. One is likely to find that a trading system operates in the same fashion. Even if it does not, however, the point still stands. One must ask the right question of the statistics, and avoid using them to prove something, to get useful results.

Shifting back to biases, the Conservatism Bias essentially means being hard-headed. It is when one refuses to consider elements which do not mesh with one’s existing opinion or belief, or to change one’s position as circumstances dictate (or to do so only slowly). We have all seen this sort of thing happen, such as the case where someone refuses to accept a fact contrary to their belief until the evidence is incontrovertible. This kind of bias is particularly problematic when someone has expressed a view publicly, meaning they have to admit to being wrong. The point is to not to trade with a set notion, and to be ready, willing, and able to make adjustments to one’s view quickly.

Biases such as the ones we have mentioned are serious pitfalls for those trading the markets. It is hard to accept that one is going to be wrong a lot and that one has no control over the markets, but both are the case. The trader that can overcome these hurdles is moving in the right direction.

Another thing to worry about in one’s analysis, especially as it relates to system design and testing is postdictive error. This occurs when one uses information which would not have been available at the time to make decisions. Clearly, this is not something generally done on purpose. It can happen, however, as the result of incorrect referencing when doing calculations in spreadsheets and charting packages. Attention to detail is all that is required to avoid postdictive error creeping in to one’s work.