Ok, the title is slightly tongue in cheek, but I shall explain it later in the article. In the meantime, let’s talk about something I like a lot… burgers! One thing that’s fairly apparent is that the quality can vary considerably between different burger joints. Some use the highest quality beef, which you might find in a good quality steak (I’m thinking of Minetta Tavern in NYC here). Some might use American cheese, others mature Stilton and so on. Ultimately though, whenever I order a burger, I’m expecting it to taste, well, like a burger. Provided chefs don’t go absolutely crazy with ingredients, you could argue that burgers exhibit a certain robustness. Even with some variation, it should be recognisable as a burger. If they are wildly different from a traditional burger, such as a dessert burger, they tend to fail (believe me, from personal experience, a brioche dessert burger, with a chocolate patty is somewhat revolting, as I’m sure you can imagine). Also simply adding more and more ingredients doesn’t make a burger better – it simply makes it too rich to eat. When it comes to creating trading strategies, burgers could well teach us a thing or two! I’m going to go through a shortlist of a few ideas below!
Only add ingredients/complexity for a reason
My pet hate hate with burgers, is adding additional ingredients for the sake of it (usually to charge more for the burger). The same is true of a trading strategy. If you are adding complexity is there a good reason for it? If there is a strong rationale that’s great! Or is it, so you can tell your boss and your investors that you’re super smart? Markets like to teach folks a lesson, who think they are super smart too! Don’t be the person making a chocolate burger to appear different…
Don’t overcook the burger/strategy
No one likes a burger which is so well done it’s burnt. Sure, you won’t get food poisoning, but you’ll also get an awful burger. If you spend ages and ages trying to search for a trading strategy something in a specific area, it simply increases the chance you’ve got into the realm of data mining. Note, that I’m not talking about the data cleaning/preprocessing step (and aggregation to some extent), which can be very time consuming in particular for a large dataset. It’s more applied to the construction of a trading rule.
Can we swap our ingredients/parameters in the burger/strategy without a big impact?
If we swap out ingredients in the burger does it still taste good? In the context of a trading strategy this can for example involve changing the parameters and observing how it impacts our returns. If our strategy is fairly robust if we change the parameters quite a bit (ie. choosing them unwisely!), then it bodes well. This can involve changing parameters in our trading rule or varying transaction costs in a backtest. If we change our pricing data source, are returns impacted somewhat? I like doing this when I’m backtesting intraday FX strategies, trying to swap out different datasets and seeing how robust my strategy is (how does it change if we’re using indicative data versus executable quotes or traded prices?). Higher frequency strategies can be much more sensitive to small changes in transaction costs. If a subtle change totally destroys our returns and we have to be incredibly precise in which parameters we pick, you have to question how robust the strategy is and it’s somewhat tricky to have confidence in running such a strategy live. As with everything there are caveats. The more niche the strategy the more likely that subtle variations can break this parameter question. Let’s take for example, a strategy which involves forecasting Apple earning by estimating iPhone sales. By construction this is only going to work on Apple stock (or possibly related stocks). There’s a good reason, why this shouldn’t be useful for any random stock.
Don’t ask for a beef burger in a vegan restaurant/is your strategy appropriate for your company?
You wouldn’t go to a vegan restaurant, if you’re looking to eat a beef burger. Far too often, it can be the case that the appropriateness of a strategy depends on who you are. If you are massive hedge fund, with masses of capital to deploy, it doesn’t make sense to develop a niche strategy which you can’t allocate much capital too (unless you have a very large team, in which case, you can make lots of niche strategies to deploy!). This is one of the first questions I need to ask any client, what are the type of time frequencies and assets they trade, so any research project I do for them, can be used by them in practice.
The more experience chef will probably cook a better burger!
A chef who has cooked many different type of meals and burgers, will likely cook a better burger. It’s the same for a quant who has developed many successful trading strategies over the years. He or she will know the types of pitfalls which can trip up the process of developing a trading strategy and also the problems you might encounter with running such strategies live. The way I view markets has evolved over the years, and I’ve (tried!) to learn from mistakes. I’m pretty sure that in years ahead, there’s a massive amount I can learn about markets too. Something I like to repeat a lot is, you can’t backtest pain.. you’re probably going to learn more from pain than a backtest too!