Categories

Every once in a while, the topic of random trading comes up. Normally, it’s part of a discussion about whether you could go long or short based on a coin toss and trade profitably because of a good exit and money management strategy. Let’s take a look and see if there’s any truth to that assertion by running some tests on EUR/USD daily data going back to when the euro was launched in 1999.

Random in, Random out
As a base line, I’m going to start with a totally random system – one which uses a coin toss to get into a trade and a coin toss as to whether to exit an existing position. The rules are very simple. Start with the coin toss to figure out long/short at the end of the first day’s trading. At the end of Day 2, we do a coin toss to see if we’re going to stay in the position we put on Day 1, or close it out. If we stay, we do the coin toss again the next day. If we exit, we start the process over at the end of that next day (so if we exit on Day 2, we do a coin toss as to whether to get long or short at the end of Day 3).

I ran 1000 tests on the data set to get enough information to make a reasonable conclusion. The results were pretty predictable. On average, the test resulted in a 26 pip loss, which is basically the same as being flat over more than 10 years of data. The standard deviation was 3668 pips, giving you an idea of how wide the distribution of results was over the 1000 test sample.

Random in, Strategy Out
The totally random system didn’t cut it, so let’s look at a random entry system that has a non-random set of rules for exit. I used the same coin toss entry as noted above, but for the exit I tested a reverse break approach. Specifically, the rule was that longs would be exited if the current day’s close was lower than the close from N days prior, and shorts would be exited on a close higher than the one from N days prior. I tested a range of look-back periods of 1 to 10 days. Here’s what it produced.

What the chart shows is the average result (the tick on the bar) and the range containing results one standard deviation above and below the average. So in the first bar we’re looking at an exit strategy which says get out of a position if today’s close is lower/higher (if we’re long/short) than yesterday’s. The average outcome was a loss of 3602 pips, with a standard deviation of 1846 pips. That means the 1-day test was a losing one in all or nearly all cases, and by a pretty sizable amount, generally speaking.

It is clear from this data that a random entry system can be profitable, though. We need look no further than the middle of the chart to see the performance of the longer look-back periods. The 6-day look-back provided the best result with a 5446 pip average gain and a 1236 pip standard deviation. Eyeballing the 1000 sample test results, I don’t see any negatives among them.

Maybe we’re looking at things backwards
Looking at these numbers, it’s hard not to think that maybe traders need to look at things the opposite way around from how they usually do – to think about exit first, rather than the entry. OK, I’m not really suggesting that we all just start trading random entry systems, but it certainly does provide fodder for further testing and analysis. We can use random entries to test the performance of different exit strategies. One caveat there, though. You have to make sure when you do something like that that you’re getting the same entries for each different exit approach, otherwise the results won’t be comparable.