Hi all,
I have been trading using mechanical systems since the last few years on my own account. I started by trying various technical analysis indicators but found that most of them don't make money directly and then moved on to stat arb. Quite a bit of edges have been reduced to almost zero if not gone entirely. However, a few changes to the original unprofitable strategy can turn it into a very profitable one. A flawed strategy by itself is unlikely to make money no matter what you do, but I've found that good performing strategies stop working and some bad performing strategies perform very well at times. This is a (long) collection of what I learnt and hopefully will save you some time when you experiment :)
Any strategy can be broken down into
- Entry conditions
- Risk management
- Exit conditions
Portfolio construction
If a significant number of positions are not showing a Mark to market profit at any point until they are closed, then the entries are either too early or too late. Entry conditions should be either based on a model or data. When you trade on a model, remember that it is ultimately a representation and is unlikely to ever be accurate 100%. Putting too many parameters in a model will make it harder to debug and take longer to fix when it breaks down. Data based strategies have the risk of a variety of biases. It is essential that if you are using patterns in data to trade, there should be a sound reason for the pattern to stay consistent over time.
Keeping stop losses too close to the entry, or too wide from the entry point will result in hitting the stops frequently or big drawdowns. I think risk management matters more to the overall Pnl in the long run than the specific strategy. Trading using algorithms is a great way to get rid of emotions affecting the trade. A stop loss should not be placed based solely on risk tolerance or performance constraints. If you can tolerate a maximum loss of 5%, keeping the stop at 5% regardless of the strategy or security is a bad idea. Consider the hypothesis for the trade. You should ideally exit a position only when the reason for entering the trade has been invalidated. So this could be 2% below the entry price, or even 20%. After considering this, size your trade size accordingly so that the stop loss is still within risk limits. Volatility should also affect your stop. As an example, oil futures typically move 2-3% on a volatile day. Currency futures rarely move 3% a day. Keeping the same stop for both is not optimal. This is especially true for stocks.
I think most strategies fall ultimately into two categories- direction based and value based. If the strategy makes money betting on the direction of something, it is direction based. If you trade a spread or bet on the fair value of some security being a specific value, it is value based trading. The exit strategy should ideally be separable from the entry. If you have separate entry and exit strategies, you can then mix them to create new strategies.
Portfolio construction is running your strategy on multiple securities or building a portfolio of multiple strategies. Since some strategies perform well at some times and others poorly, by combining the strategies into a weighed portfolio, you can reduce the volatility and smooth the returns. A simple way to do it would be to use correlations between the strategies. Put together different strategies that are uncorrelated to each other in a portfolio.
Hope this helps!
