Strategy Optimization on Random in-sample and out-of-sample Periods

The "RandomMonths" function picks months depending on a percentage value passed to this function. If you specify 100% then the function returns TRUE for all months, if you specify 50% then the function returns TRUE for half of the months and if you specify 0% then the function returns FALSE for all months.


The idea was brought by "mdmd" and it consists of creating random in-sample and out-of-sample periods for trading system or rules optimizations.

Here is a basic trading system formula:
Optimize("a", 30, 90, 10);
Optimize("b", 30, 90, 10);
buy = rsi(14) > a and close > 10 and close*sma(volume,5)>100000;
sell = rsi(14) < b;

Buy a stock if RSI is higher than "a" and sell it if RSI is lower than "b".
We could backtest/optimize this strategy for a specific period (let us say 2000-2006 <- This is our in-sample period). Select the best performing strategies and then perform the backtest on an out-of-sample period (for example, 2006-2011).

Or we could use the "RandomMonths" function to generate random in-sample and out-of-sample periods.

To use the technique, we should add the following lines just below the "sell" rule of the previous trading system:
pass = RandomMonths(70, 15245);
buy = buy and pass;
sell = sell or !pass;

The second parameter of the "RandomMonths" is called random seed and it is a number that initialize a pseudorandom number generator. In order to get different months on each execution, you should change the seed value. A unique in-sample period corresponds to each seed number.

In the above example, the in-sample period contains 70% of the months as specified by the first parameter of the "RandomMonths" function.
We instruct the trading system to buy only if "pass" is true (We are in the in-sample period) and to sell all positions if the "pass" variable becomes false (We are out of the in-sample period). Note that "pass" condition of the sell rule can be removed if you want to keep trades open and use only regular exit rules.

It is possible to optimize the strategy using different in-sample periods by creating an optimizable variable and then adding the variable value to the seed number.

To backtest the strategy during the out-of-sample period:
- Use the same seed number
- Add "!" operator (logical negation) just before the "RandomMonths" function

Technical Analysis

• Vertical Horizontal Filter (VHF)
• chande moethan 1
• Keltner/Bollinger Bands Squeeze
• Percentage of Stocks Outperforming the S&P 500 Index
• TD Sequential Indicator
• Number of Stocks Above Upper Bollinger Bands
• EMAs Cross Strategy
• Output metric for each trading year
• Relative Strength TAA System
• Ascending and Descending Triangle Patterns
• More trading objects...

Fundamental Analysis

• Roubini Sentiment Indicator
• Company Revenues - US Stocks
• Dividend Frequency Indicator
• Industrial Production Index - IPI - Historical data
• FAO Food Price Index
• Earnings: EPS Estimates
• Splits Calendar - Historical data for U.S. Stocks
• Earnings Surprise Strategy
• Abu Dhabi Securities News
• Gross Domestic Product - Historical data
• More trading objects...