Last week we decided to investigate buying when RSI signals an overbought market. Our task this week is to develop a strategy around the concept.

A natural starting point is to buy when RSI = 70, the traditional barrier for an overbought market.

The exit strategy carries an equally important load. Some will disagree, but I usually try to base my exit strategy on the same indicator used for entry. Let us exit the market when RSI closes under 65.

The idea appeared on a 60 minute chart. Stick with that for now until we analyze different time frames.

How do I test this idea?

• Manually backtest
• Use a computer

Manually backtesting does not require any technical skill. The disadvantage is that it takes an incredible amount of time and effort.

A computer can calculate a week’s worth of manual backtesting in a few minutes. There are many software packages available to do this. MetaTrader is a popular option for forex traders since it costs nothing. Other popular options require a monthly subscription like Esignal or TradeStation.

The next step is to program the source code (computer instructions) for the strategy. Computer programming is beyond the scope of the article. A company like OneStepRemoved.com helps traders skip this step by doing the programming for you.

My charting software reported profitable results on the intial tests, but the entry and exit points might produce better returns if they use slightly different values.

Curve Fitting versus Optimizing:

One danger of system development is curve fitting. If you know how the market behaved in the past, developing systems that made money on historical data is easy. The danger is that historical market behavior does not necessarily replicate itself.

The best way to optimize a strategy on historical data while avoiding curve fitting employs a method called walk forward. The idea itself is simple. Optimize a strategy using data that is:
a) more than a few years old
b) at least two years in length

You should use old market data for one simple reason: you most likely developed the idea trading in today’s market. If you want to trade the strategy for years and years, it needs to function in ALL market conditions. Short time periods do not vary enough to generate a firm opinion as to whether your strategy holds up in protracted, unfavorable market conditions.

Not knowing the future, systems developers do the next best thing: they look in the distant past to make sure the strategy holds up in different market conditions.

I chose to look at the EUR/USD 60 minute chart from October 2002 – April 2005 for my initial tests and optimizations. Ideally, I would use data from that was longer and from further in the past. Forex limits testing options because most charts only contain data from 2002 and on.

The tests returned a suggested entry point of RSI = 73 and left the exit point unchanged at RSI = 65.

The returns looked great, obviously, because they were optimized for that time period. The real test comes with the next three steps:
1) Walk forward
2) Testing multiple instruments
3) Analyzing multiple time frames

Next week’s article applies these three steps for the RSI strategy.