I mentioned long/short equity strategy a lot in the previous articles. This investment strategy is very important and it is one of the most prevalent alternative strategies. (This strategy is generally associated with hedge funds.) A long/short equity strategy seeks to minimize market exposure while profiting from stock gains in the long positions and price declines in the short positions. So the total returns from this strategy is a combination of the return from market exposure(β) plus any value-added from stock-picking or market timing(α). As long as the long positions can generate more profit than the short positions or the other way around the strategy will be profitable. Typically, equity long/short strategy is based on ‘bottom up’ fundamental analysis of the stock.
Long/short equity is also one of the market neutral strategies(refers to hedging out market risk). One of the popular variation of the long/short model is that of the ‘pair trade’ (which involves offsetting a long position on a stock with a short position on another stock in the same sector) that I introduced and used before.
When it comes to the practical phase, the first thing we should do is to build a ranking system to separate the stocks based on specific ranking scheme( we often use the returns as the criteria). Generally, the equities whose rankings are high are expected have a higher expected returns on average(we believe they are undervalued now). Moreover, we need to create two baskets for both long and short positions. We also need to decide how many equities(positions) we would like to hold in total before setting up the baskets.
For instance, the strategy goes long the top n equities of the ranking system and goes short on the bottom n while maintaining equal dollar volume between the long and short positions. The total positions that you hold here are 2n. You actually will sell [(1/2n) * d](total dollars that you plan to invest) of equities ranking low and then long same amount of (in dollar value) of equities ranking high. I will also illustrate the principle behind the strategy here with a simple example:
Assumes that you have ranked equities based on your ranking system and set up your short and long baskets respectively based on your budget and your total positions plan. The average expected return of the short basket(based on the ranking system you build) is -2% and the average expected returns of the long basket is 2% under an ideal situation. In the end, your total returns would be 2% — (-2%) =4%. The following graphs demonstrate this principle in the visual and statistic way(generating the values by simply simulating a dataset which is normal distributed; then generating the return series by adding some noise on the value series in order to make them to be highly correlated with each other so it is easier and more direct for us to understand):
The beauty of long/short equity strategy is that it is statistically robust and market neutral as by ranking stocks and entering hundreds or thousands of positions, you are making many bets on your ranking model. You are betting purely on the quality of your ranking scheme and your total returns of a long/short equity strategy are dependent on how well the ranking spreads out the high and low returns instead of how the market looks.
As you can see the key component of long/short equity strategy is to choose the right ranking scheme(which can be value factors, technical indicators, pricing models or a combination of any of the above). Since choosing a good ranking scheme is quite tricky, the good starting point for me is to pick existing know techniques and see if I can modify them slightly to get increased returns.
Another concern related to the strategy is rebalancing frequency. I will just briefly introduce it here. All the preparation that we did for constructing long/short strategy is based on how we predict the future return of our equities, however, every ranking system will be predictive of returns only within certain timeframe(such as a priced based mean reversion may be predictive over a few days while a value-factor factor model may be predictive over many months). It is critical to decide the timeframe over which your model will be predictive and you also need to verify it by conducting the statistic tests before executing your strategy. Once you have determined the timeframe on which your ranking system is predictive, trying to rebalance the portfolio in order to take full advantage of your models based on your estimation and test result.(Attention: if you are going to trade many equities, you probably need to consider the transaction costs related to the trading. It is dangerous to change your positions by long/short equity strategy too frequently since the transactions cost would be really high.)
I would like to plug in another concept ‘Capital Capacity’ which usually implies the range between the following two criterion:
For Long/short equity strategy, it tends to have very high minimum capital capacity since we are betting on the average and diversity of investments in order to limit the market-related risk. If the volume of our investments(or we can call it long/short equity strategy’s positions) is too small, the risk would be much higher and it requires our predictive model to be very accurate(which is not very difficult like I discussed before) to some extent.
The long/short equity strategy also tends to have high maximum capacity since we only have very little capital per position and very large amount of positions in total. Only if you make a very high percentage of the trading volume of any given security, you probably need to consider the outcome of the slippage.
In short, there are three key factors here we need to take into account when we are going to measure the capital capacity of the strategy.
Here is the backtesting output of the last 6month( I did this few days ago so the date ended by 12–11–2015 which is last Friday). I modified the algo based on equity long-short algorithm that posted before on the forum.
You can just have a look about the result of the algo. As for what I changed exactly, why and how did I change it, I will make a detailed explanation as well as upload the code in my next article which I may post later today or tomorrow. I will also extend the testing period in order to get a better understanding about the quality of the algo on the historical data and post it later.
As usual, leave your comment if you find anything I explained here is inefficient. Thank you for your time and hope you enjoy the content of the article.