The trend filter

November 29, 2015 12:00 PM

Beating the index isn’t very difficult. There is one simple rule that can be added to practically any equity strategy to dramatically improve results. It might not be what you think. In fact, relative to this one point, the rest of the strategy doesn’t really matter that much.

You may be used to discussions of whether to use technical analysis or fundamental analysis. Should we use value investing or growth investing? Trend-following or counter-trend trading? Systematic or discretionary decisions? These may be interesting talking points, but they are merely minor ones in the context of equity strategies.

So if I am not talking about the stock selection criteria, then what about position sizing? While this is an important factor, it’s still not the most critical one. There is something much simpler that can in one fell swoop improve your equity strategies: 

Don’t buy stocks in a bear market.

That sounds so easy, doesn’t it? Let’s make a completely random strategy and take a look at just how much of a difference such a little detail can make. 

Random stock picker

Supposedly, an average monkey throwing darts at a list of tickers can pick stocks better than most investment professionals. Due to a shortage of nearby primates willing to take on the task, it will be delegated to a random number generator instead. At the start of every month, all stocks will be sold and then a random set of 50 stocks from the S&P 500 Index constituents will be bought. Realistic trading costs are added, both commissions and slippage.

Note that to make a fair simulation, stocks will be picked only from the S&P 500 members as it looked on the day in question, even if those stocks no longer exist. This is a critical point because the current index members are in the index only because they had strong historical performance. Any long-term simulation using current members would show unrealistically positive numbers. All cash dividends and corporate actions are taken into account and adjusted for. 

Some may think that picking stocks at random seems a little irresponsible. To remove any doubts about that, this simulation will also use random weights. That is, all stock selection as well as position sizing will be totally random. There will be no constraint to position sizes, other than that they must all add up to 100% exposure at all times. The only fixed factors are that the strategy will always hold 50 stocks and it will always be fully invested. Holding this number of stocks should ensure a sufficient diversification.

As crazy as it may seem, it’s a money-making strategy. No one is seriously suggesting that you would trade like this, but it does make for an interesting learning exercise. It gives some idea of what kind of returns you should be getting on average, regardless of stock picking strategy or position size approach.

Running a random simulation like this once or twice will not provide any useful information. After all, if you throw a die once, you may get any number but if you throw it 500 times the average will approach three. Therefore a total of 500 iterations of this random strategy will be run to make sure we have enough of a basis to draw conclusions.

The power of random

It might come as a surprise to many people, but this utterly random strategy has a very high probability of outperforming the index. It’s not that this is a great way to trade, but rather that the index in itself is a horrible way to trade. The primary culprit in the poor performance of the index is that it’s heavily overweight in large caps. Because the S&P 500 Index is market-cap-weighted, huge companies like Apple (AAPL), Microsoft (MSFT) and Exxon (XOM) completely dominate the index. The six largest companies in the S&P 500 have together as much impact on the index performance as the bottom 250 stocks. 

Because the largest stock in the world is unlikely to be the best performing company in the world, investing with the index might not be the most rational approach. The largest companies in the world are the largest now because they had spectacular performance in the past, but once a company has a market value of half a trillion dollars, 
it’s not as easy to double as it was when it was worth a couple of billions. Chances are that some other company will be the top performer next.

“Strategy performance” (below) shows the performance of the random strategy. The thick black line shows the S&P 500 Total Return Index—that is, with reinvested dividends included. Because our simulation benefits from dividends just as in real life, it would be unfair to compare with the usual price index. The commonly used S&P 500 price index disregards dividends and will therefore show worse performance than you would get in reality. 

As can be seen, a few of the random iterations failed to beat the index, but more than 90% of the time, the random strategy wins.

While our proverbial dart-throwing monkey in this demonstration is beating almost all mutual fund managers in the world, the obvious problem seen in the figure is the drawdowns. The random strategy will at times take large losses and spend extended periods recouping those losses. If you have the luxury of being able to close your eyes for a couple of decades and wait, this might be fine. However, the drawdowns mean that if you invest for shorter periods you may incur significant losses. This is where the magical rule comes in: Don’t buy stocks in a bear market.

A simple filter

Now take that same random strategy and add a trend filter. For demonstration purposes, we’ll keep it as simple as possible. At the start of every calendar month, we sell all holdings and buy 50 random stocks. The weight of each stock is still completely random, but they all add up to 100% to make sure we’re fully invested. The only added rule here is that if the index itself is below its 200-day moving average, we don’t buy. This means that if the index is below the average on the first of the month, we spend the entire month in cash. Now we do another 500 random iterations to see what comes out.

In “Avoiding the bear” (below), the results are clear. Even the worst of these random simulations clearly beats the index in the long run. On average, there’s a very large out-performance. That, however, is not the most important part of this picture. What’s really interesting is how this filtered version avoided any serious losses. While the index itself took large hits at times, the random approach did not, thanks to the trend filter.

Note the long flat periods for the random strategies during the bear phases of the markets. It’s not easy to sit on your hands and watch during these market declines, but it is the most prudent course of action.

The scatter plot in “Risk and reward” (below) makes it easier to compare. First look at the big red dot. That’s the S&P 500 Total Return Index showing an annualized return of 4.4% and a maximum loss of 55% over this period. Not very encouraging numbers. The lower cluster of green triangles shows the result of the random strategy without any trend filter. It’s quite a wide spread, but an overwhelming share of these data points shows higher returns than the index, and most of them also show lower drawdowns. 

The cluster of purple diamonds at the top reflects the random strategy with the trend filter. Now we’re seeing a tighter formation and, more important, considerably better results. Every single point shows much higher return than the index at much lower drawdowns. 

As odd as it may seem, such a random strategy is likely to outperform the index in the long run, something that the mutual fund industry has so dramatically failed to do.

Is random the answer?

Trading a random strategy is absolutely not the answer. The point of the demonstration was to use random stock selection and random risk allocation to isolate the effect of the trend filter. The key point to take away is this: Practically any equity investment strategy will be greatly improved simply by avoiding buying when the markets are falling. That may sound obvious, but in practice it’s not a common method.

Performing random simulations like this should give you added confidence for your own approach. If you understand what a random selection is likely to yield, you will have realistic targets to beat. Clearly picking stocks at random doesn’t make much sense. Allocating risk to the positions on a totally random basis makes even less sense. But it still works.

What you should do is keep using a trend filter. Exactly how you implement it doesn’t matter much, as long as it tells you when the markets have been declining. The 200-day moving average is fine, but so are many other ways of doing the same thing. Don’t focus on indicators; focus on getting the task done.

Now add a rational way to select your stocks. This can, for instance, be based on fundamental value factors, on growth analytics or on quantitative momentum rankings. To that, add a method for exiting the positions. That could be based on traditional stop-loss mechanisms, rankings or on other criteria. 

Finally, add a method of deciding position sizes. A common approach in the industry is to use volatility parity or variations thereof, but there are several ways that this can be done.

With the trend filter kept and the stock selection and allocation added, you will end up with a strategy that has a high probability of outperforming every single mutual fund in the world.

About the Author

Andreas F. Clenow is a hedge fund veteran and CIO of ACIES Asset Management, based in Zurich. He’s the author of best-seller “Following the Trend” as well as the recently released “Stocks on the Move.” You can reach him at