Take Your Back Test Into the Future

So you’ve developed or learned about a great strategy. You back tested it, tweaked the rules, and re-tested. Maybe the results are looking great – now what?

All the information that you have is history. Is it reliable and repeatable in the real world? Here are few things that I do before running a strategy live.

1) Make sure the strategy works in different market environments – not just the most recent time frame. A few years that I like to spot check are 2005, 2008, 2011, and 2013. You’ll get choppy years, trending years, big up moves, and big down moves.

2) Be realistic about frequency and screen time. If you can’t take every signal because your list is too long or the frequency is too quick, then you won’t be able to replicate the results of your back test.

3) Simulate the strategy in real time. Real time data and electronic paper trading allow you to gather important information without capital at risk. You won’t really know about liquidity in a simulated environment, but electronic simulation often keeps track of commissions and spread at the time of execution. You can also see if the future P/L curve resembles your back test.

4) Testing with small live execution allows you to see the true impact of liquidity and transaction costs. The real world is the only way to measure these vital components that can easily be removed or forgotten in a simulated back test.

5) Re-run your back test in the future and compare it to your simulation and small live testing. This step is vital to know if you are replicating your trade assumptions in the real world. After gathering a few dozen forward tested trades, you can run a new simulated back test over the same period. The closer your real time information matches your simulated back test, the more reliable your system is for executing at larger size.

6) Place a stop loss on your equity curve. If a future draw down exceeds the historic draw down, then your strategy is not performing as expected and its time to allocate to something that is still working.

7) Avoid “fixing” a broken strategy. If it needs to be fixed, then your original assumptions are not holding true in the real world. Assuming that you found the one thing that caused the problem is not a reason to place real capital at risk. Your fixed version should be treated like a brand new system – taking you back to the beginning of the simulation process.

Experience will help you avoid strategies that are doomed to fail. Until then, keep your risk small while you let the process and the market give you the evidence that your back test is working in the real world.

Send me an email if you would like to learn about the new SMB Systems forum. This is a place to share ideas, back testing reports, forward testing performance, and learn about what is working.

-Andrew Falde
[email protected]


No relevant positions.

Leave a Reply