Whenever I travel I try as much as possible to explore a destination in full. I recently travelled to Nice for the WBS Quant Finance conference (I’m speaking at WBS Women in Quant Finance conference in London on 17 Oct 2018, if you’re interested in attending that). I had time to look around the area for a day before the conference. I had been to the area several times before, nearly 20 years ago. Rather than having some sort of strict itinerary, I simply started walking along the coast eastwards from Nice, till I got tired. By walking, as opposed to taking a train or car, it gave me time to absorb my surroundings and find some unexpected places. Obviously, there’s a limit to how far you can walk and hence, it only allows you to see a small area. However, it is probably the best way to explore a specific place in detail.
In a sense, we face this problem in when it comes to understanding markets and developing trading strategies. The first problem is which idea to spend time testing, which is akin to exploring an area by car. Then we have the problem that to do any sort of research takes a lot of time, the stage of walking. Clearly we can’t “walk” everywhere, instead we need to make a judgement call before we start that an area we wish to explore in detail is likely work it. How can we tell? Here are a few pointers that I think are useful at a high level for finding trading ideas to begin with, which we may later wish to investigate in more detail.
Come up with some general ideas to start
Before we can do anything, we need to come up with a shortlist of ideas we’d like to test. We have an infinite number of potential ideas to start, so we need to narrow it down first!
- One obvious source for trading strategies are books and papers. And, no, just because we can find that someone has written about a strategy doesn’t mean there’s no alpha in it! The market isn’t so efficient as to read every single paper and extract the alpha out of it immediately.
- You can also talk to your friends in the market and colleagues to try to understand promising areas. People are not going to tell you exactly every detail of their trading strategy for obvious reasons, however, a few comments here and there, combined with your own insights can be useful as a starting point.
- Also, people are often more open when it comes to discussing which ideas they’ve tried, which haven’t worked. If everyone is telling you a certain dataset has little or no alpha, then perhaps it’s not the best thing to try yourself first, unless you think you have a specific novel approach to looking at it.
- You might have ideas too! I’ve come up with many ideas basically through observation of the market. Many don’t end up working (my brain has just found the noise in charts), but sometimes they do work. In practice, your ideas are not going to be unique. Most strategies which I’ve run in the past, I’ve thought to be pretty unique, only to find that years later, yes, other folks were running the same strategies! If everyone has similar datasets and smart folks on their teams, they will probably end up in similar places (given enough time.. it won’t be the case on day 1!)
- If the idea is good enough in your view, add it to a list. I’ve kept a list of trading ideas to investigate for a couple of years. Whenever I hear about something I find interesting (or I come up with an idea), I simply add it to my list (and a URL to a paper if necessary) for future investigation
Choosing between ideas
When you have time for research a new trading ideas, you can take a look at your shortlist, which you’ve compiled over the years, the question is then which idea to look at first.
- First remove those ideas which are not executable for you. If you are a very large fund with lots of capital to deploy, a high Sharpe low capacity idea might not be as attractive to look at as a low Sharpe idea, with loads of capacity. By contrast a high frequency firm, with very good infrastructure, might favour a high Sharpe/low capacity idea, because it’s more likely to be executable
- Which idea you pick is also dictated by your resources. If it’s a really expensive dataset and particularly complex one, it’ll be very costly to research (and run potentially). Are you likely to find a trading strategy that can make enough money to offset this? (some expensive datasets might turn out “cheap” because they can be reused for many different types of trading strategy – by contrast some might be more niche, so the alpha content has to be particularly high to justify, eg. a dataset which is only relevant for a small number of stocks).
- What’s the delta (or the probably) you attribute to the idea working? Bear in mind even ideas where the outcome in negative could be “positive” – finding out that an idea doesn’t work is still valuable information. Experience does help significantly here to sort between ideas. Over time you get a feeling as to which ideas are more likely to work, and which are less likely to be useful (I admit this is not very scientific, but in my view developing trading strategies is more of an art than a science).
The next stage of research: number crunching
Once you’ve chosen a specific idea, the number crunching phase begins, best of luck with that! Of course, this phase is important for verifying your hypothesis, but in practice, coming up with a hypothesis is in my view much more important. Backtesting a trading strategy can be done with enough hours of coding. However, coming up with a promising hypothesis requires much more than an ability to code. We can also argue that having a lack of understanding of the market can render a backtest useless, because you will be backtesting an unrealistic strategy (eg. not taking into account proper liquidity). Once you have done your backtesting and additional research, you will know whether you want to try implementing it in a live environment (typically paper trading, followed by real cash). Alternatively, it could be the case (in all probability) that the results are not good enough to run (or it is too correlated to existing strategies etc). You can then update your shortlist with a brief comment and cross it off.
If you would like some ways of coming up with trading strategies from an independent source, obviously contact Cuemacro . We regularly do consulting in this area, researching novel datasets for trading purposes and developing trading strategies with clients.