[kevstev]

They make a lot of those resources available for free. I worked in algo/hft for about 11 years, then left and started building my own tools for building factor based algos using my interactive brokers' apis and my own toolkit. I was doing it on the side for about 3 months before I came across quantopian- and immediately felt like an idiot because they had done so much of the work for me- and it was available for free. It would have taken me probably 2 years to get anywhere close to their offering, and that's just for the pure trading/event processing side of it, let alone their backtesting, and risk measurement framework.

I have been using it for a factor and value based microcap strategy for the last 4 years and have been beating the market's total return by about 5% on average each year. I trade a relatively small universe, its not really scalable to a large fund.

I stopped applying for "quant" jobs in 2007- everyone wanted a PHD even though I was doing a lot of the same work- I got lucky and landed a job in Options AMM and for awhile was doing both dev work and what I later found out was considered quant work in most other places.

My point being though is they provide a lot of infrastructure that would normally only be available to a very small number of people in banks and the like and those tools are gate-kept to PHDs and those deemed worthy of even getting a look. I don't put my algos in the competition and actually trade them manually so they are likely completely under the radar from quantopian, and aside- they look for very specific characteristics in their algos that mine don't fit. There is opportunity out there though, especially if you are trading in small size (lets say under 5-10M),

[c128]

2007, that's when most quant shops shit the bed, you didn't miss much goodtimes. I'm curious, how many out of samples did you backtest your algo through? And what time frames were used ?

[kevstev]

In general, I backtested for the longest period the platform would allow, which is unfortunately only back to about 2002 (I really wish it went back pre-.com boom, to 1997 or so). I would then often do separate tests for 2007-2013, 2012+, etc, just to see how returns were affected by market timing- you can kind of eyeball this, but over long timeframes, sometimes unacceptable underperformance during a volatile period can be a bit hidden over a long period.

[c128]

We easily go over a hundred years of testing.

[kevstev]

If you have the data, that's great. Is there really value in that though? Markets have changed so much, especially with the rise of index funds and etfs and such. I couldn't imagine much value going back to pre WWII, or even before 1970. I would be satisfied if Q went back to the early 80s

[c128]

I'll put it to you this way, I was stunned at how much hasn't changed.

[yosyp]

Could you elaborate on quant jobs requiring PhDs? Do they look for a very specific PhD (i.e. statistics), or is it just a blanket requirement to get past the hiring screens?

[kevstev]

This was all about 10 years ago, the market has moved since then, and now it seems you can slip under the PHD radar and do similar work if you classify yourself as a "data scientist." Some of the mythos around quants and technology has been uncovered, 10 years ago quantitative and electronic trading was still a newish thing and at a backseat to traditional traders and PMs. Many firms these days are technology first, though there are often still walls between those who write "models" and those who write "infrastructure."

Anyway, its still largely true that for "quant" roles you need a PHD. It doesn't need to be in anything particular, but Physics/Math/Statistics are strongly preferred. This is just a hiring screen thing. Its not impossible, but quant types tend to have a big head and be elitist, and I personally found them to just be real jerks in the hiring process, and there were plenty of opportunities opening up in the then lucrative algorithmic/HFT space so I went down that road. By jerks, there were just several opportunities where you could just tell they didn't like my lack of credentials and that I went to a state school (on a scholarship, but still they can't have non ivy leaguers stinking the place up), and it was also fairly easy for them to just throw advanced math problems at me and knock me out of the running- regardless if this was something that was ever used in their actual work.

I don't really regret it, but it would have been nice to be more heavily involved in pure finance stuff- I find the markets fascinating, I was very happy when I wrote some of the first TWAP and VWAP strategies out there, and then later (surprise!) started getting edged out of that space by "quants" with PHDs.

[c128]

In the early 90s, I was lucky. I pitched on the phone, they called me in, and the following day I had a corner office overlooking the beautiful statue of liberty. We laughed, dined, and drank often after work, it was fun and not a snob in sight.

[solaarphunk]

machine learning or statistics with a programming background.