| Avoiding Bias in a Trading System |
Retail traders, or traders responsible for their own account, are generally looking for a system or trading strategy (or multiple strategies) that will help them make decisions, manage risk and ultimately turn a profit. Most traders around the world deal with these struggles and successes, in all markets.
I would broadly define a “trading system” as an approach that prescribes entries and exits, position sizes and expectations. In this, and subsequent lessons, we will look at each of these three components and also discuss the impact of investor psychology, technical analysis and fundamental analysis on the creation and implementation of a strategy.
Building a system is not complicated, but testing and execution can be difficult. This means that proper preparation is key to success. As we progress through this course, we will build a few systems and walk through the decision making process together.
Article Continues Below. Click on the pairs below for analysis, charts and data
At PFX we naturally have a longer term bias toward forex investing and system building, because it minimizes the retail trader’s biggest enemy – trading costs – without adding significant disadvantages. However, the ideas discussed herein are just as valid short-term. The two common issues faced by System Building and Trading: 1. Attribution bias 2. Over-optimization
Attribution Bias Attribution bias is not an issue unique to traders. It’s experienced by people in all aspects of life, but can be detrimental to trader’s success. The issue appears when you attribute a causal relationship between two things that have no definite relationship.
Here's a specific example: Imagine you are a new trader. You see the amount of time and news spent on the US non farms payroll report each month, and therefore perceived a causal relationship between the report and price movement that follows it. You may naturally assume that if the report misses expectations, or reports results lower than the previous month, that the USD would weaken and that movement would be predictable.
If this sounds like an assumption that you have ever made, then you've been subject to attribution bias. In fact, the NFP report has almost no predictive value for the USD. Looking back at historical results, a down NFP report resulted in a down month for the USD just slightly over 50% of the time. The monthly chart below shows a good illustration of this random movement.
The blue arrows indicate whether the NFP report at the beginning of the month exceeded (up arrow) or disappointed (down arrow) expectations. The most recent month look pretty good but once you go back a full twelve months, the report loses its trend-predicting ability. The more data you include, going back in time, the more random the results are. While the Non-Farm Payrolls report is one fundamental variable that may influnce price movement, the actual report gives no immediate trading indicator.
There are many variables in the forex market that cause price movement, and traders should be very cautions when using single variables to forecast and trade. System traders often base systems on single indicators that seem to predict results, but most of the time, these indicators are just one of thousands of variables moving prices, making a system based on an incorrect indicator flawed before it is ever executed.
Over-optimization Over optimization is related to attribution bias for forex traders. Over optimization is the act of trying to fit past data into a predetermined model. The problem usually appears when traders are back-testing a system, and begin adjusting the system’s parameters until past results are maximized.
There are a couple problems this leads to. First, over-optimization leads to results that are skewed, innacurate or just plain false.
The second is that even accurate past results are not necessarily indicative of future returns. If that phrase sounds familiar – it should. It is included on every prospectus you will see from a legitimate advisor or money manager. This is a trap for system builders as well as traders hoping to buy a system from an advisor. Beware of overly smooth equity curves with multi-year returns that seem abnormally high. They are probably optimized-backtested returns. In the video, I will run through an example of exactly this situation and why the trading industry unintentionally encourages this behavior. This is found often and easily illustrated in systems that perform well in certain market conditions but not in others. An upward trending market may make a system look very profitable, but it may be a big loser when the market is channeling or down-trending.
What can you do about this information? I begin this course with these warnings, because it is best to go into portfolio strategy and system design aware of the issues that you will face. This knowledge can keep your mind open, and your logic sound, to opportunities that we will be demonstrating in the rest of the course as we build a few systems together.
Let’s use Non-Farm Payrolls as an example again. You may be able to show that NFP has no ability to predict the intermediate trend for the USD, but it does do something else quite reliably. Because of the amount of trader attention on the report and the anticipation before its release, options volatility and premiums run up before the release. An FX options system trader can definitely use that information in a system.
Similarly, experimenting with optimization is not all bad. It can tell you a lot about the sensitivity of a system to market conditions. If a variable you are adjusting in the back-testing process shows that small changes yield very large differences in overall returns, then this is a bad sign for implementation. That means that normal market slippage could ruin the system and make it unexecutable. Likewise, the opposite is also true and could identify a great system.
Comments deemed inappropriate will be removed
3.25 Copyright (C) 2007 Alain Georgette / Copyright (C) 2006 Frantisek Hliva. All rights reserved." |
| Learning Markets Partners More Partners | Become a Partner |
 |  |  |  |  |
 |
| Learn to Invest | Reviews of Stock Brokers | Stock Picks | Technical Analysis | Broker News | Investor Education | What to Invest In | Live Market Analysis | Should I Invest? |
