How to Look Into Potential Performance of your Strategy before Live Trading

The first question that I always ask myself is "do I have a friend who I can call and he will tell me how is it right to trade the market and whether it is going to be an up or down day tomorrow" and the answer is always definitely no and probably no one has one like that.

So how can we make our trading less risky or at least when it comes to loosing money, how can we make it so that it is psychologically  justified when that does occur because it will. How do we build out our "pain levels', levels where we would say stop and turn the strategy off?

This is all a big discussion that I would like to have today, it is a great topic and let's get down to it. I will share my personal approach. I am not a quant, all I do, is I try things and make them work without too much interest in what can be the undercover super hidden mathematics  although I have great respect to this discipline but as long as it works and makes money why not just make it run!!!

So we all have heard about Robert Pardo and his book on optimization. Well at least that was the first book I read about in sample and out of sample periods that we use in order to get at least some sort of projection into the future. 

Basically, we optimize the strategy on one set of data which is called the in sample, and we run the strategy on a different set of data with the settings we found before on the in sample. That other sample is called the out of sample. You can get quite creative with this approach and run multiple variations and combinations and test the "test" out of your strategy until it fails eventually and never hits the live). I used to say if you never want to trade the strategy live just keep testing it)!

So if we were in the perfect world of endless lifetime and limitless resources for testing we would run many many in sample and out of sample iterations and then put together a big out of sample piece to see how our strategy would have performed over a large set of data with periodic re-optimizations.

Now according to Robert Pardo, we should also compare performance of each particular in sample period with its corresponding out of sample period and score out each result and sum it up eventually. The book says that if at least 50% of the results are greater or same as expected the strategy is believed to be quite robust. By the way, this absolutely does not guarantee that it won't fail over the first few trades on live)))).

So next is the question what are the pain levels? When do I stop the strategy and either replace it or re-optimize it? 

I personally have a few in favor such as: max draw down, max consecutive losses, period without new equity high, max loss in one trade. So if either of these are demonstrated on live trading with performance that I have never seen on the test, I would become very cautious.

Also, there are few more that I like to keep an  eye on such as average win and average loss. Your strategy should show something similar to what it was on the test, otherwise it would be a different strategy or the volatility has changed and the strategy is not doing what it was doing anymore.

Another good one is max win because if you see very good results on live such as a huge win that you never saw on historical it might be a good idea to turn the strategy down after that because a loss can come that you never seen just as easily.

One more technique that is quite effective is to continue your in sample equity curve into your out of sample equity curve and then further into your live trading equity curve. Just by looking into this chart you can easily spot the mood change of your strategy. I would also recommend the same for the draw downs curve.

Now finally, what should be said is that if you run a test and then go live you should expect to see what was there during the test unless you are trying to "hype" the strategy and not keep it running for a long time. What I mean is that sometimes you can turn the strategy on during the period of a peak in draw dawn with the expectation of that there is high probability that it would actually start climbing out of it right now.

In other words if your maximum consecutive number of wins is 5 and you turn your strategy when it has historically just made those 5 wins most probably your next few trades are not going to be very profitable!

I my opinion it is quite important when you turn the strategy on, it should be at some sort of equilibrium and not on the edges of its performance.

Now coming back to the psychological side, if you see 25 consecutive losses on the test but you still like the strategy and you are ready to trade it you have to be ready to sit through those losses to see its performance after that.

To finish up I want to mention that if you have 5 strategies that you validated, most probably all of them won't fail so when you start your trading game you need to plan your capital and manage your risks. Risk management sounds very posh but all it is, is take all your 5 strategies, out your max draw downs, sum them up, if you are ready to loose that amount of money to see if these strategies work, go for it!

Hope this helps!) Cheers!