Modeling Trading System Performance
hoW can i Tell When The SySTem iS Broken?
Encountering a large drawdown is one of the primary reasons trad- ers stop trading. Understanding, estimating, and limiting drawdowns, and determining whether a drawdown is within reasonable expecta- tions of the system or a symptom that the system is broken, are among the primary goals of this book.
The price and volume data consists of a combination of signal and noise. The trading system logic has been designed to recognize the signal portion of the data in anticipation of profitable trading oppor- tunities. Systems rely on the continued synchronization between the logic and the data. As variations occur in the data stream, the degree of synchronization shifts. When they are in sync, the system is profitable; when they are out of sync, the system is unprofitable. One of the most difficult questions traders face is how to determine whether the system is working as expected, or is temporarily out of sync, or is broken.
Determination of whether the system is working or broken relies on comparison of recent trading results to some benchmark. Comparison implies that there is some metric, such as profit per trade, available for both the recent results and the benchmark. There are two easily estab- lished benchmarks:
Previous performance – compare recent results against actual trades previously made with the system, or against the out-of- sample results from the validation process.
Random – compare recent results against random results.
Statistical analysis is used to evaluate the comparison. Several sections of MTSP explain statistical testing procedures, suggest appropriate tests, and give rules for actions.
iS iT PoSSiBle To meeT The goal?
Having a terminal wealth 10 times the initial wealth in 4 years requires an annual compounded rate of 78% per year. That is, 1.78 ^ 4 = 10.0.
Whether this is possible depends on several factors:
The distribution of trade results, including the average profit
per trade and variation among trades.
The number of trades per year.
The portion of the active account used to take each position.
Copyright © 2011 by Howard Bandy All rights reserved This document is a chapter of “Modeling Trading System Performance” Published by Blue Owl Press, Inc. www.modelingtradingsystemperformance.com