Why Build A Portfolio Of Quantified Strategies (Including Two Strategies)

Last Updated on December 7, 2020 by Oddmund Groette

The most important task of a quantitative trader is to find trading edges and subsequently turn them into good stand-alone strategies. However, many traders neglect to test how those strategies perform together as a portfolio of strategies. Just as a long-term investment manager put together a portfolio of stocks, a short-term trader needs to evaluate how the strategies perform together as a portfolio.

You want to build a portfolio of strategies that both complement each other and make the portfolio diversified. In order to this, you preferably need to test and simulate on a trading platform. You need strategies that differ in markets, time frames, and types.

An investor doesn’t compose a portfolio or basket of only oil stocks. The investor looks for stocks that both diversify the portfolio and complement it. A trader needs to have the same mentality.

It’s impossible to predict the future value of a stock, and likewise, it’s impossible to know the future predictive power of a quantified strategy. This is the reason why you want to have a portfolio of many strategies. Some strategies will gradually deteriorate, some might blow off spectacularly, and others perhaps perform much better. You need to diversify your strategies just as you diversify your stock holdings.

Why is the composition of the strategies important?

Five individual strategies that perform well on their own, might not complement each other well. There could be several reasons for that, like for example:

  • You have overlapping trades in the same instrument/asset class.
  • The strategies might correlate too much.
  • Strategies in the same instrument might not overlap, but the winning trades might overlap a bit, making the average gain substantially less.
  • The type of strategies is too similar.

Why you shouldn’t spend too much time on one strategy

We know many traders who spend literally years fine-tuning just one or a few strategies. They are trying to hit the jackpot by adding more variables or filters for their quant strategies. But we believe this is a very bad idea. It’s a bad idea because:

  • You end up curve fitting the strategy to the past and lose predictive in the future.
  • All strategies, sooner or later end up arbed away or become victims of changed market behavior.

We believe in this: It’s better to have a portfolio of many sub-optimal strategies than one or a few “perfect” strategies.

Why have a portfolio of strategies?

The most important factor in quantified trading is having unrelated strategies, ie. strategies that don’t correlate with each other. That is of course no easy task and requires a lot of work. For example, finding strategies that don’t correlate among stocks is next to impossible. Stocks move up and down mostly in tandem and thus trading a mean-reverting strategy on a basket of stocks gives many signals at the same time. You need to diversify away from stocks to avoid this.

Another obstacle is that during panics the correlation among all asset classes increases. During a short-term panic, as we saw during the GFC in 2008/09 and the Covid-19 in March 2020, everything is sold down except Treasuries and some select commodities. If there is a long-term bear market, like we saw during 2000-2003, the correlations are much weaker.

How can you make uncorrelated strategies?

You need to look at different factors to differentiate your strategies:

Different markets and products

The best way is to look at different markets. Commodities, for example, often go the opposite way of stocks. Thus, one of the first things you should do is to look at different markets.

Time frames

Time frames can also be helpful. Day trading, for example, should not correlate much to the overall trend of the market. Trading books like to mention you should always go with the underlying trend, but much of this is nonsense and never tested in any quantitative way. The fact is that longs could be fantastic for short-term gains in a bear market. We will in a later article bring forward some interesting facts about the three last bear markets we have witnessed in the stock market.

Type of strategy: Mean-reverting, trend-following, or momentum?

Mean-reversion means a move either up or down is followed by a move in the opposite direction. Stocks are typical very mean-reverting in the short-term.

There are three main ways to trade the markets: mean-reverting, trend following, and momentum. Mean-reverting and trend following are opposites: when one is not working, the other should absorb the losses.

Trend-following strategies are very difficult to trade because they have mostly a low win rate, thus you need to prepare for absorbing many small losses while waiting for the one big winner that recoup all the previous losses. Very few traders can withstand many losses in a row before they either abandon the strategy or start tinkering, usually at the worst moment. This is exactly why you shouldn’t be either a trend-follower or a mean-reverting trader only. You can use both strategies. There are no bad or good type of strategies, you have to be open and adaptive and use all the options there are in the markets.

We can add breakouts to the list, however, breakout strategies are quite similar to trend following strategies, and sometimes you can’t separate the two. Both trade on market strength and assume that once momentum has picked up, the market is likely to continue in the dominant direction. Breakouts refer to the price when it breaks out of a range or a price level, and you want to be long or short this breakout in the anticipation of further movement in the same direction. You define the breakout yourself, but breakouts might be a bit complicated to formulate by code and exact math.

Be both long and short

Some investment managers focus only on the short side. The reason is simple: even if short for example has a CAGR of 4% annually, it might offer tremendous value in a portfolio. It gives downside risk reduction, and at the same time might contribute to better overall CAGR of the portfolio.

As a trader you need to think very much the same. Unfortunately, finding short strategies are very difficult. In most markets you can’t reverse long strategies and expect them to perform well on the short side.

Strategies deteriorate or get arbed away

One piece of advice is something you will never read on the internet:

Strategies stop working!

If you manage to find one good strategy you can be pretty sure it will stop working at some point in the future, or at least have a long period where it performs badly. No matter how prepared you are, markets change and evolve. This is yet another reason why you should have an arsenal of strategies that varies in markets, types and time frame. Not all strategies will stop working at the same time.

You have to look at failed strategies as the cost of doing business.

Why do strategies ultimately stop working? There are many reasons for that:

Curve fitting

No matter how much you try to avoid curve fitting, it’s inevitable that you to a certain degree curve fit. Most of us test a lot of ideas and strategies, and some of them perform well simply by chance. If you do proper out-of-sample testing, this fallacy can to a great extent be avoided.

Avoiding many variables in the strategy is also a good start to avoid curve fitting. You must always have in the back of your mind that most movements are completely random and just noise. Ask yourself this: Why should the strategy work? Is there any logic?

If you test many strategies, some of them are surely just a product of luck and noise.

Strategies deteriorate

Strategies usually deteriorate over time. Why does this happen? It’s inevitable because markets change and the ever-changing market cycles take years to develop. And many get arbed away:

Arbitrage

Let’s use an example from the investment world:

Factor investing has gained a lot of attention over the last 5-6 years. We believe many of these strategies are likely to be unprofitable over the next decade. Take low volatility stocks for example. Because of their popularity, their earnings multiple has increased a lot. This type of investing is attracting a lot of capital and thus getting “crowded”. When something gets too crowded and too popular, it stops working. Think of strategies the same way.

An example of two strategies and how they perform together

Let’s make an example of two very simple strategies and look at how they perform together.

The first strategy is this (in Amibroker):

Buy= RSI(2)<15;
buyPrice=C;
Sell= C>Ref(H,-1);
sellPrice=C ;

In plain English it’s like this:

  1. Buy when the two-day RSI indicator is below 15. Read here for how the RSI indicator works.
  2. Enter at the close.
  3. Sell when today’s close is higher than yesterday’s high.
  4. Exit on the close.

The strategy performs like this on the S&P 500 (100 000 compounded from 1993 until November 2020):

And the equity curve looks like this:

Not bad for being invested 24% of the time!

But let’s look at another strategy that is somewhat similar in nature, ie. mean-reverting:

range=MA((H-L),100);

Buy= ( (Ref(C,-1)-C) > Ref(range,-1) ) ;
buyPrice=C;
Sell= C>Ref(H,-1);
sellPrice=C ;

The strategy goes long when today’s close has dropped more from yesterday’s close than the 100-day average difference of the high minus the low. The strategy is of course very similar to the RSI strategy and returns a pretty similar result:

But how do these two strategies perform together? Let’s assume you want to trade both strategies, but only have one position in SPY at all times. Hence, you only trade when you have no position, you don’t enter an additional position when you are already long. Combined the strategy is like this:

range=MA((H-L),100);

Buy= ( (Ref(C,-1)-C) > Ref(range,-1) ) OR RSI(2)<15 ;
buyPrice=C;
Sell= C>Ref(H,-1);
sellPrice=C ;

Combining the two strategies we get this result:

All parameters deteriorate: average per trade and the profit factor both decline while the time spent in the market increases from 24% to 31%. Obviously, the strategies are very similar in nature and have a lot of overlapping trades. But that is the point of showing these two strategies: you need to have strategies that are different. You can’t add strategies that are reasonably similar, you need to complement the portfolio.

How many strategies do you need to succeed?

There are in practice no limits on how many strategies you can run via automated software. We at Quantified Strategies use Amibroker and Tradestation, and the biggest limitation is our own imagination and creativity. For example, we have at the most run about 55 strategies via both platforms without any problems at all.

Conclusion:

A portfolio of strategies might smooth your equity curve and make you avoid behavioral mistakes like abandoning trading or acting irrationally during a market panic. You need an arsenal of strategies to keep you diversified.

True inefficiencies don’t happen often. A strategy doesn’t need more than 10-20 trades per year to be fruitful. If you for example have 20 strategies which are all somewhat different in markets, time frames and types, you generate about 200-400 trades per year. As a swing trader that should get you a good start.

However, it takes time to build a portfolio of strategies. Take your time and be patient. Make sure you enjoy the process!

We have a landing page of different strategies that might offer you some inspiration: