Strategy's returns and P&L

"Developing a trading strategy is an itertative process; it begins with forming few trading ideas, back-testing, and drawing conclusions or forming new ideas and starting over. This paper discusses the back-testing process of the strategy; we demonstartes the steps, issues and assumptions made during the preparation of the data sample and the calculation of the hypothesized startegy's returns. Finally, we turn our attention to risk aspect of the strategy and application within an established trading environment with common risk management policies."

Overview

You have a trading idea, and you are anxious to see whether it actually work. You may wonder also how would your strategy have done if we can rewind the time (back-testing). To do so, obviously we need to acquire necessary data, define hypothetical strategy buy/sell signal, and compute the returns.

Trading strategy

A trading strategy defines the followin:

• The underlying instrument(s).
• Enter decision - When, How, and at what condition(s) (e.g. price) we should open a position
• Exit decision - Given an open position, when, How, and at what condition (e.g. price) we should open a position

Strategies vary significantly in their complexity. For our discussion here, we construct a simple startegy that buys S&P 500 Index at market open, and sells it at market close on a daily basis. Furthermore, we assume that our trading capital remains the same.

Raw market-data

There are many data vendor offering real-time and historical prices. The decision of where we get the data from depends on its availability, quality and cost. Many data vendors offer historical data for one asset-class (e.g. equity, fixed income), frequency (e.g. daily, 15-minutes delayed, tick-data), and in raw or preprocessed form (i.e. scrubbed from flaws).

Another important factor in the decision is the actual data offered; does the data includes bid-ask? HI-LO? Limit book? etc.

For this strategy, we used S&P500 daily index prices at maket open and at market close in the last 6 years (2003-2009). Fortunately, this data is publically available.