Known-bias price streams for system evaluation

Evaluation on generated prices

Technical trading systems can be evaluated on historical, live, or computer-generated prices. Series built with a known trend-percentage make that evaluation statistical rather than anecdotal.

Unbiased increments still form runs

When each increment is equally likely to be an uptick or a downtick of the same size, runs of two to four consecutive moves in one direction still occur. Unbiased random generation already produces trends.

Averaging the five most recent increments and taking the sign of that average yields a trend indicator. That indicator can be used to build streams with a chosen bias, including 10 percent or 20 percent trend.

The trend-percentage is the imposed statistical bias that makes successive increments more likely to continue a signed direction than to reverse it.

Bounds that reverse the increment

Generated streams can wander without limit unless a high-price and low-price bound is applied, and such bounds are often desirable.

A range-constrained-stream is a generated price path forced back inside those stated bounds by reversing the increment sign at the boundary. After an initial signed increment is drawn with a specified trend bias, a breach of the low bound forces a positive increment and a breach of the high bound forces a negative increment before the next price is computed.

Illustrated streams that did not repeat

One illustrated construction produced 50 points with a 20 percent statistical-trend bias. A second family of 200-point streams started at 100 and compared 0, 20, 40, and 60 percent trend settings, and none of those streams repeated.

A known target for adaptive rules

An adaptive-rule-set revises its own rules from the observed character of the price stream. Adaptive systems that revise rules from the character of the price stream require a known trend target. Otherwise the evaluator cannot tell whether the system adapted.

Non-repeating generated streams that carry a known trend-percentage and omit periodic waveforms such as sine waves are presented as a suitable test bed for technical trading systems.