Finding Edges

The article “Finding Edges” was originally published on Robot Wealth blog.

How do we find edges?

First, we must be clear about what constitutes a good idea. It isn’t as simple as it having to make money. The risk profile must also be tolerable. This is a personal preference.

Next, we need to be able to trade it.

Robinhood won’t let you sell naked options. You can’t trade the Indian markets or crypto derivatives from the US. Retail can’t trade OTC.

These restrictions will eliminate many things, particularly from academic studies, where they don’t seem to recognise this problem. (As an aside, I recently had a very well-respected academic pitch me a short volatility ETF idea. He thought his observation that the VIX basis was over-valued was the secret sauce. But he also thought that marketing was a trivial problem, and that people would prefer an active solution over a passive one. Basically, he thought the easy bit was the hard bit.)

Here’s what I think is important for finding ideas.

Stay current with academic research

Academic papers are worth reading.

You will need to read dozens to find anything useful, and even then, it will probably only be the basic observation that is illuminating.

I browse SSRN habitually. And I think browsing is the key. A targeted search is almost guaranteed to limit you to areas you have already considered. SSRN finds will have references, and you can then put them into Google Scholar and see what papers have cited those. Hours of fun ensue.

Don’t start with data

There is too much stuff to test until you have an idea from somewhere else.

This is related to the naïve view of science, where we start with an observation, then make a theory. But without a theory, what should we observe? Is the sun rising in the east relevant to the double slit problem? Seems unlikely, but we only know that from theory.

Treat anecdotes as evidence

I like talking to market-makers and floor traders.

Even now, the really good ones aren’t particularly quantitative. Instead, they notice stuff. They see patterns a long time before they are picked up by statistics.

Of course, the obvious downside is that they will also see stuff that turns out to be noise, but good “old school” traders are great observers.

As one said to me, “I can tell you stuff that you can go and nerdify.”

The best situations occur when there is a mismatch between what theory predicts and what the trader observes.

Assumptions we forget are assumptions

Think about the assumptions that are so well ingrained that people have forgotten they are assumptions.

People commonly apply put/call parity to American options. This is wrong due to carry costs, obviously, but more interestingly, it is wrong due to volatility when there is a steep term structure. The option that is expected to be exercised will see a very different volatility to the non-exercise candidate.

Everyone has software that can price American options, but no one uses different volatilities for calls and puts of the same strike.

Cast a wide net

People often ask me, “What option book do I read next?” They seem to think that there is some deeper knowledge out there. But it only takes three or four books to get to the end of the relevant option knowledge.

At that point, you should read about lots of stuff.

Your brain works in a non-linear manner, and giving it random things to ponder often leads to new relevant ideas. Read about gambling, sports, military history, Silicon Valley…

Just read widely.

What to do with your ideas?

Once you have an idea, be prepared to give up on it quickly.

If it doesn’t clearly point to money, no amount of tweaking is likely to produce a robust trading idea. You will either over-fit to the data or not even manage that.

This is particularly true of “interesting ideas”.

You may really want the Parrondo’s paradox to be the basis of a successful strategy, but what you want doesn’t matter. You might sometimes prematurely discard a good idea, but that is less likely than wasting time on a bad one.

Next, apply the rule that you need to be able to give a one-sentence reason why the idea might work.

Think of yourself as asking for funding for a start-up. Give a convincing elevator pitch to yourself. “I provide liquidity when markets are distressed” could be convincing. “I have a 4-layer neural net” isn’t.

It doesn’t hurt to ask other people for feedback at this point. It is very unlikely that you have a totally new idea, and that telling someone will kill it.

If your idea passes these stages, then it is a potential strategy.