Travails of a Trend Follower

Over the last few decades, Trend following has really taken off with a lot of believers flooding the streets so as to say. After all, if I were to sell you with charts like the one below, why would you not be convinced about its benefit

But then again, I choose that particular chart out of the 1000’s available because I could use it to show what I wanted to show. Selection Bias / Survivor Bias and what not come into picture the moment I select a single or even multiple instruments from a set of data.

But trend following is tough and this makes one starting to question premises even when we can show from history that what is happening is exactly something that has happened in the past and will happen in the future as well. But, past is the past and the future is unknown. Present is the key and our emotions aren’t really concerned with either the past or the future but the moment on hand.

Yesterday for instance I had a debacle day for me. In the morning my stops got stopped out and I went long. Well before the close though, I once again got stopped out and went back to the original short mode. But markets had not ended and while I did not get any fresh signal markets did move a lot higher than the point that my longs got stopped out and I went short once again.

This is a rarity for my system with system exiting 2 trades in a single day being just 5% of the time, but it hurts and not surprisingly is the costliest (per trade) whipsaws as the chart below will showcase

What is striking in the above chart is that whenever the trend ends fairly soon, you end with negative results that are the primary contributor to the adverse win-loss ratio most trend following systems have. In the above system for example, if you are in the trade for more than 7 days, the probability that you shall still end up in a loss is pretty low. The chart below plots the same

Its amazing to see how not a single trade (out of 365 in above example) could close in +ve if the reversal happened in the first couple of days. Once that hurdle is crossed, the probability keeps going lower until it hits zero and stays there.

Since 1996, Nifty has moved up by around 8000 points. But if you were a trader, you could have gained that 8000 points by being rightly positioned on just 36 days (which is just 0.70% of the total number of trading days). If you were having a trading system that traded daily, if you slept for the first 10 years (1996 – 2006) and applied your theory on just those 36 days, you will have in theory outperformed all 99.3% of the other days (all this being theory, but please  bear it with me for a moment).

In other words, you could have been long since 1996 and gained exactly the same points as some one who entered and exited daily on just 36 occasions. Of course, if only we knew about these 36 days in ahead, why would we bother would be the question in your mind and you are absolutely right.

But think on the contrary you were long for all the time and yet were out of the market on those 36 days. What would you be staring at? You would be looking at having the same capital as you did 20 years ago.

Now, lets take a trend following system equity curve. What is the cost of missing a few trades (which inadvertently turn out to be the best trades you could have taken)?

As can be seen above, more the trades you miss, lower the returns. Since the number of trades scrutinized above is around 365, missing 15 trades is missing only 5% of the Signals and yet the returns can be disastrous. You could always argue that maybe the trades you missed weren’t the best but the worst and hence the returns actually are better. But if you could do that in real time, your success rate will be closer to 100% since you can easily over ride all bad trades similar to the way many advisors just remove their bad calls while showcasing their good ones.

So, given this relationship, why do traders still try to skip few trades in the hope that those skipped will be a loser and hence be advantageous. Think of a coin toss. Theoretically odds of a fair coin falling either on its head or tail is 50:50. But if you were to toss the coin n number of times, you can get streaks of heads or tails. But does that change the probability of the next coin toss to something like 40:60? Of course not.

Trend following systems have a average winning of 40 trades vs losers of 60 in a sample of 100. But that doesn’t mean anything since you can have 10 or 12 or even 20 consecutive losers without one single intermittent win. But does that really change the overall ratio? I say Nope. It still remains 40:60 in favor of Losers.

I keep hearing various remarks about how you should know when to ride the system and when to over-ride, how one should not take a trade before a big event (which is actually of a lower risk compared to the risk of the Unknown event we take when we carry positions home everyday), how January is not a good month for longs among various other theories.

When you have a loser after loser, its easy to believe a lot of nonsense that gets sprouted. But as showcased above, data indicates that missing can cause more harm than participating in each and every trade. All we can control is Risk and that is better controlled by modifying our position size than by skipping a few trades.