Alternative Title: What to Do if Your Boss is Terrified of New Ideas?
Several younger quants have asked this question:
“The culture of our quant group is very skeptical about new ideas. They are terrified of data-mining, and random factors. How can we innovate in such an environment?”
1. Idea-mining is safe. That’s called THINKING. No one can stop you from doing that.
The best way to gain support is to make your ideas as good as you can. Experienced researchers can tell if you are just dabbling for the sake of novelty or actually have identified something fresh.
Ideas are the most important part of investing, but as quants we spend only a tiny fraction of our weekly hours on developing this skill. Start every day building up your idea muscle. Before reading. Before technology. With no right or wrong answers. Just your favorite pen, notebook, and your brain. Perhaps music and a cup of coffee. Fill the pages with questions and ideas. Overtime, your idea muscle will grow and you will have notebooks filled with thoughts.
Occasionally, use the crowd to source and filter ideas. Look around your life, professional and personal, for a diverse set of acquaintances. Ask them for creative solutions and ideas about the topics that you are pondering. Run your ideas by them and test their quality.
Figure out your unique investment innovation style. While walking around town, pretend like you are an artist, contemplating your next idea, refining your style, associating with something completely unrelated. Start living in your ideas. Investing is not a job, it’s a mission. What kind of investor are you becoming? Don’t copy. Identify what is unique about the way your mind works. Wondering where to start - try here or here or listen to this podcast with Adam Savage. Or check out What is Creativity?
2. When communicating your idea, connect it to other credible ones.
For the idea to feel safe, it needs to be connected to something credible and trustworthy. Look for analogies in other areas. Find a supporting quote by a famous person. See if some fundamental analyst is already using a similar metric?
Compile a list of prior (ideally published) papers on related ideas, and conceptually anchor your idea within prior research. That’s what academics themselves do to get their ideas published.
Then articulate what is novel about your idea, best formulated and presented as an answer to a great question.
3. Time to test your idea?
Save up to 30-50% of all the data for the out-of-sample test. Be honest. Tell your boss that you will not touch the out-of-sample data until getting their approval to run it. This will feel brutal. Scary. What if the backtest comes back flat? Practice vulnerability.
Run the in-sample tests (including cross-validations):
Play, tweak, experiment, and create a composite of the best iterations of your idea.
Identify your preferred breadth to depth ratio that has the best chance of reaching the goal.
Remind everyone of the goal: to produce positive out-performance in live returns. The goal is not to minimize the difference between in-sample backtest and out-of-sample backtest or to avoid random factors all together.
If the in-sample results come back random and you found nothing, your boss will appreciate the depth of your original thinking and the integrity of your tests. Go back to step 1.
Run the out-of-sample tests:
If the in-sample results are positive, show them to your boss and ask for any feedback, tweaks, anything else to try before touching the out-of-sample. That will generate constructive engagement. Then hold your breath and run the model on the out-of-sample data.
Experienced quants would say that out-of-sample tests deteriorate by about 70%. It depends on the situation. Can be much worse or better. The point to remember is that we are not competing with natural sciences. We are competing with randomness. Who cares by how much the out-of-sample deteriorated vs the in-sample. All data work is data-mining, the question is whether all the mining ended up with just dirt or at least some gems.
Did you find at least something that was statistically positive in the out-of-sample? That’s great news. If not, go back to step 1.
4. Set up a paper portfolio tracking your ideas.
Even if your idea wasn’t accepted into the model, set up a production style realistic paper portfolio for your new ideas and run the performance reports. Over time, that will make the strategy look more real, especially if you are lucky and the near-term performance is good.
To conclude: the best quant models are always evolving and as “Creative Quants” we need to understand and adapt to the culture and the investment philosophy of the place we work in, while at the same time advancing our unique ability to innovate within it.
I was fortunate to have experienced working as a quant in both cultures: one that was very open to new ideas and one that was comparatively more conservative. The irony was that while working in the open culture, I craved more academic depth and technical rigor, and wished that we would hire some famous quant from another top firm so they could teach us the latest and greatest quant methods. In sum, I was wishing for less openness and more expertise. Not being able to have that at the time felt frustrating but with hindsight I realize it was also a blessing.
Then, after sometime of working at the conservative culture, where it felt like any good idea needed to have a Nobel Prize paper behind it, I started to crave the openness again. I started to miss the days where I could add any factor I wanted to the model and get instant approval to put it into production. (My freedom to do so expanded after several years of getting better at the four steps above, which for full disclosure I have not followed perfectly, but I continuously strive to).
So what’s the take-away for an aspiring Creative Quant? Don’t blame the circumstances. Everything has pros and cons. Find your strengths, your style and the styles of people you work with, and start coming up with something special that you are proud of. Over time, you will become a successful quant innovator and add real value to client portfolios.