Magazine article Modern Trader

Sound Off!

Magazine article Modern Trader

Sound Off!

Article excerpt

Monte Carlo confusion I am perplexed by Wolf von Ronik's article about "System Testing via Monte Carlo" (April 2001). He uses a moving average crossover system to illustrate Monte Carlo testing, but the example he describes uses random data sets that mimic the distributional characteristics (mean and variance) of the real market to be traded. I thought moving average systems were specifically intended to take advantage of (putative) serial dependencies or (hypothesized) persistent temporal behaviors (momentum) of markets. Regardless of its shape, a random distribution by definition has no serial dependence and no persistent temporal behaviors. How can such a testing procedure tell [us] anything about the utility of a trading system when the test purposefully ignores the very market properties that the trading system attempts to capture?

Richard B. Hoppe

Via e-mail

Wolf von Ronik responds: As I understand Mr. Hoppe's argument, a moving average analysis is undertaken for the purpose of discovering tradable patterns inherent in a market. I will add that if such patterns are discovered, then the market cannot be random in nature. Mr. Hoppe also feels that a pseudo market" created by using the mean and standard deviation of a known market would be wholly random. I will add that such a random market would not exhibit any preference for any trading technique be it moving average or buy and hold.

My rebuttal is via proof by contradiction. If the pseudo markets created from real market data were truly random, and therefore devoid of the patterns inherent in the original "real" data then surely repeated simulations using such randomly created data would result in each technique winning out about half the time. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.