V.7:2 (40-42): Creating prices for system testing by D. Howard Phillips

Creating prices for system testing by D. Howard Phillips

Technical trading systems are evaluated using historical, real-time or computer-generated data. All three data sources are useful, but computer-generated data can be the most powerful. In particular, using computer-generated price data that has a known trend percentage allows you to determine the performance of a trading system statistically. These "trend price streams" are helpful in evaluating any technical trading system, but are essential to evaluating, adaptive trading systems that learn from experience and change their trading rules according to the characteristics of the price stream. If you want to evaluate an adaptive system, you must know how well it adapts to a known trend

Random number generators

I created the price stream in Figure 1 using a random number generator to assign either a plus or minus (+ or -) to each number in the DELTA column, which shows the change in price.

Now, a simple definition of a trend is "something repeating." The interesting thing about random number generators is that they do generate trends. For each new price in the DELTA column, the probability of a +0.25 DELTA is equal to the probability of a -0.25 DELTA. Further, it is probable to have two, three or four successive price increases (or successive price decreases). This leads to trend generation by all random number generators.

In Figure 1, the AVDELTA column gives the average of the five preceding DELTA values. The sign of this average is the trend indicator, showing that the trend is either positive or negative. You can use the AVDELTA sign to create price streams having 10% trend, 20% trend or any other trend percentage, as illustrated in the equations in Figure 2. These price streams will never repeat and, therefore, are ideal for evaluating technical trading systems. They also do not contain sine waves or other periodic waveforms which can lead to unrealistic test results