Why I Quit Algorithmic Trading to Do Web Startups

It's hard to underestimate the impact that working in a secretive industry has on your career and tools you can build on (points 4 and 8). Trading technology is driven by a really fantastic set of challenges. There are only a few other problem domains where you are forced to react semi-intelligently to millions of messages each second with microsecond latency. So there's some real innovation going on. But what Jason says about the closed culture is spot on.

Information only flows from one firm to another through hires. There is little open dialog between developers at different firms. For most of the last decade that I've been in the industry, the sheer scale of the technical challenges we were facing kept most of us innovating ahead of the consumer web (in terms of infrastructure tech -- don't ask us about UI). But now that the consumer web is going real-time, I expect that edge to evaporate quickly. Unless you want to play the high frequency game (and have access to the requisite large amount of capital), I would recommend a web start-up over a trading firm. Your salary will likely be lower, but you'll be happier, and may even have more upside.

- Ben

Disclaimer: I haven't fully made the leap myself yet. But I'm doing my best to spin out a real-time web start-up from our trading firm.

Interesting post.

It is good to see that the author sees that there is no free lunch.

My semi-outside perspective is that there are firms in an arms race to be the quickest at providing liquidity, but his is a small fixed pie. Everything else is being paid to take risk. It is a tough game, especially for folks with skills marketable elsewhere.

It takes a while, but great for the poster to realize money isn't everything.

I really enjoyed your post; I, like many programmers I would assume, look at anything like financial trading or sports betting and wonder if it's really possible to build an algorithm that could consistently beat the market.

The assumption is that once logic is removed, a machine could execute based on data. But humans crunch data so much faster than CPUs at this point (at least until around 2020 probably), it's still hard to pinpoint the minute logic that goes into a successful trading strategy.

Your post really shows the realities of how challenging it is, even when guided by humans who have past success at trading.

This was very interesting. I would also recommend reading the article about increasing your luck surface area (http://www.codusoperandi.com/posts/increasing-your-luck-surf...). Luck is a concept people think about a lot in terms of startups and "luck surface area" is a great way to quantify this.

i came to the same conclusions as this fellow, and left electronic trading for the startup world almost exactly a year ago today.

i couldn't be more pleased with my decision.