Thread by @RobinWigg

Good piece on a really cool idea that unfortunately fizzled. Short thread on what I thought might have been the Achilles heel (which I didn’t appreciate enough when I wrote about the phenomenon a few years ago on.ft.com/3696uEP )

https://twitter.com/justinaknope/status/1339231403465117697

Basically, I think I (and Quantopian obviously) underestimated the compounding power of collaboration and institutional knowledge. Essentially, you probably had thousands of people independently coding very similar algos on large cap US equities.

But that is just ferociously difficult. If you were a novice quant joining say DE Shaw or Two Sigma and tried to find a way to trade S&P stocks, you’d get laughed at. All the simple stuff a novice might naturally try was thoroughly mined decades ago.

The big quant outfits probably have vast libraries of now-defunct or never-worked signals that stretch back decades. Even if you found something that worked, it probably only works at a scale where it’s utterly uninteresting for a bigger shop, or is just a mirage of data mining.

Or you discover stuff that is large capacity but ubiquitous, such as factors that are completely mainstream - and have had a rough time for a few years now.

Without getting too philosophical, it probably shows the limits of lone-wolf, atomised knowledge work, and why WFH will become a bigger thing but has limits. There is value in being part of a broader organisation that might seem hard to discern, but is very real.

I thought quant trading would be more conducive to this crowd-sourced model, but I increasingly think quant is another area where scale is paramount. It’s an arms race where DYI algo traders are competing with shops investing in gargantuan data, quantum computing and MITs finest.

Anyway, I’m sure @jaredbroad and @richardcraib will have more nuanced and intelligent views on what went wrong!