Last week we looked at the randomness bias, which is the tendency people have to seek patterns where none exist and to invent the existence of unjustified causal relationships. Because people attempt to understand and make order out of the market, they assume that the longer a trend continues, the more likely it will suddenly turn around. This manifests into the “gambler’s fallacy” which is a very common trap that traders fall in to and lose money when they do.
This week we will cover the topic of data reliability and biases that come up in this area.
Reliability. When people obtain information, they fail to assess how reliable their data is, where reliability refers to the degree to which information reflects what is really happening. What traders observe in the market, with the possible exception of floor traders and other market makers, is not the market, but some sort of visual representation of the market. Thus, you are responding to a bar chart or a candlestick chart, or a point and figure chart, or to a representation of the market profile, etc.—and not to the real market. Furthermore, few people make decisions from that information alone. Instead, they distort the information even more by using indicators. These indicators are essentially shortcuts or heuristics that people have thought up to condense, organize and make sense of the data. Interestingly, there are hundreds of possible indicators—in fact, hundreds of thousands if you count various permutations and combinations—but most traders use only about 20 of the most common ones in their decision making.
Market information is certainly distorted, and thus less reliable, when it is transformed into various indicators. The less reliable the information is, the less value it has for predicting. Using our example from last week when Jack observed patterns in the market, reliability is a measure of how accurately Jack’s pattern actually predicts a sharp move in the market.Many people might notice a pattern or relationship in the market and then use it in developing a system without ever determining how reliable the relationship is. Accurately knowing how well the pattern predicts the move is very important information for any person wanting to develop a trading system.
A lot of the biases people have in their decision making tend to distort reliability in some way.For example, we have many biases keeping us from knowing the true probability of an event happening. The true probability refers to the actual probability of the event occurring as opposed to a statistical estimate of the probability from a small sample.
One such bias that keeps people from developing a good trading system is called the representation bias. We tend to imagine that what we see or expect to see is typical of what can and/or will occur. Thus, if you observe a pattern in the market, you expect it to occur. If you develop some concept about the market, you will look for data to support that concept in the market, and you will probably find it whether it exists or not.
Once again, if you do not test objectively, and understand the results of the testing, you will probably find that your observations, in developing a trading system, tend to confirm what you expect to find. Thus, the representation bias is particularly important when it comes to assessing various trading signals. Are you considering the true probability rate in assessing your indicator? That is, are you considering the percentage of time a particular indicator is followed by the predicted outcome? Probably not!
I cannot overemphasize enough that trading indicators are merely ways of representing things of interest. Does a significant chart pattern actually mean that buyers are about to dominant sellers, or vice versa, and produce a significant price change? Of course not!It merely represents the possibility such an event might occur. Thus, any indicators you develop for buying or selling in markets are your way of representing potential trading opportunities. It is not the opportunity per se. Yet most traders, because of this particular bias, act as if the indicators are what they represent. It is like the indicators (be they stochastic, RSI, or moving averages) start becoming reality, instead of a representation of a concept or a belief in your head. When you realize this, you will become much more attuned to what trading is all about and less concerned about indicators and understanding the market.
Another bias that keeps people from understanding the true probability of an event happening, and thus distorts its reliability, is called the availability bias. We make predictions based upon how available the information is to us instead of the true probability rate in the population. Thus, when you first start looking at the market, the data sample you use will determine what you observe. In addition, strong emotional experiences, which affect how strongly information stands out in our minds, tend to strongly bias our decisions.
When people start to develop an estimate of how much a trading system can earn in a year or how many winning trades it will have, or any other estimate of its reliability, they tend to start with a set point. They then make adjustments to that figure according to anticipated changes in conditions. The initial set point is called an anchor. The dangers associated with using anchors in our decision making about trading systems (or anything else) is called the anchoring bias.
The first danger is that you assume there is some relationship between the anchor and what you are predicting. For example, in order to predict the price of the market a year from now, you would probably start making your estimate with the anchor of today’s price. Over a short period of time it may be an accurate basis for beginning to make an estimate (i.e., today’s price is a good starting price for forecasting the price in two or three days), but over a longer period of time the strategy does not allow for the unpredicted or the unexpected.That is why one of the most important parts of developing a trading system is extensive planning. And this extensive planning should include a careful consideration of everything that might go wrong.
The second danger in the anchoring bias is that people make an assumption that the initial set point or anchor itself is meaningful. For example, if you use the results of your testing to predict future results, you are assuming that those results are meaningful and will not change dramatically over time. This is probably true if your testing data is different from the data you used to develop the system and included enough samples to make future estimates reliable. But those are big “ifs.”
Another bias that tends to have a significant effect on trading decision-making is hindsight bias. People tend to see relationships in the market after they occur, and then assume they knew it all along. It’s very easy to point out such a relationship after it occurs. I’ve worked with a number of clients who claim that they cannot follow their signals. However, what tends to happen is that they do not recognize the signals while they occur. Instead, they see many possibilities in the data.But once the signal is complete, it is too late! They then criticize themselves for not taking it when it occurred. The typical response is, “I knew it all along. Why didn’t I take that signal?”
This problem will not occur if you write down your criteria for a signal in enough detail so that it could be entered into a computer.You can then make a checklist for your signal (or computerize it). Once you do, you will always see a signal when it occurs or the computer will see it for you. Thus, you really will know whether or not you actually knew it all along.