Testing Algorithmic Trading Software
Feb 8th, 2010 by Smelly Brain Articles
When developing algorithmic trading strategies, many brokers find that one of the longest parts of the development and delivery cycle is the testing phase. Of course, the length of time it takes depends upon how sophisticated the algorithm actually is. There is always the question of relativity.
Relativity is how close or how far the algorithmic trading strategy deviates from a standard. So if a broker has developed a standard strategy and the new strategy is an iteration, or a customization of that, the testing time will be determined by how many iterative steps the new strategy is away from the source strategy. Each iterative step has to be fully tested and back-tested.
Then, once the new algo trading strategy has been fully tested by the broker, it needs to be tested in the client environment too. generally, the client has a number of revisions and again, each of these revisions needs testing.
This is one of the challenges facing brokers who provide algorithmic trading services to their customers. The client generally has a specific requirement in terms of trading parameters. Often those requirements are very quantitative in nature, particularly if the client is a quantitative hedge fund. The broker builds the algo (or customises an existing one) and delivers it to the client, who then generally asks for various tweaks to be made. There could be many iterations of such tweaking before the algorithm performs exactly as the client requires.
Then there is the whole issue of data provisioning. Most brokers will back test the algorithms in-house with their own data, but increasingly clients are want to see the algos back-tested with historical data from their own environment.
So testing plays a big part in the delivery of algorithmic trading software and algo trading strategies. The brokers who can work with their clients and their technology providers as partners are the ones who stay ahead of the game.
For more on this topic, visit the Algo Trading Podcast