|  MAR 1996

Measuring System Performance by Lars N. Kestner

TRADING TECHNIQUES Measuring System Performance Want to improve the evaluation of your trading systems? Here's a new performance measure that uses some statistics to measure profitability through time. by Lars N. Kestner Starting with an idea and developing a trading system is a relatively simple and straightforward process. Because it is a simple process, it is possible to create numerous systems in a short time. Since trading is a profit-oriented process, our interest lies in the performance of the best systems created, not the number of systems created. How, then, can performance be determined in order to select the optimal system? Looking at total profits is a major part of this question, but other factors are perhaps as important, if not more. Consistency of profits is probably the most influential factor in performance. Basing performance evaluation on consistency of returns can provide better insight to our ultimate question: How will the system perform in the future? Many measures exist to quantify trading system performance, but all have certain weaknesses. Some can be biased by one or two big trades, while some do not account for risk properly. Some are biased by correlation in returns, while others are not comparable over different periods. I have developed a new method of evaluating performance that is more robust than current popular techniques. This method, which maintains the idea of measuring reward as compared with risk, utilizes more advanced statistical techniques to quantify performance. Rather than simply looking at returns independently, consistency of results through time will be the focal point of this new performance method. The new measure, which I call the K-ratio, works with the equity curve of a system. An equity curve is simply a chart of accumulated profits through time. A mathematical model must be created for equity to calculate the components of the new method. In the model, equity on any day is explained solely by its distance from the start of the test.

by Lars N. Kestner