Andreessen Horowitz, Point72 Invest in Crowd-Sourced Quantopian | Hacker News Reader

Andreessen Horowitz, Point72 Invest in Crowd-Sourced Quantopian

bloomberg.com

97 pointssdouglas20 comments

> “I think of Quantopian as the next-generation BlackRock,” said Alex Rampell, a general partner with Andreessen Horowitz, who will be joining Quantopian’s board. “They can become a very large asset manager, but hopefully in a way that no one else can replicate by aggregating and harnessing the best talent.”

Hmmm.....

The thing is, until you actually manage money you can't really tell if you have alpha or if you are one of the million monkeys typing out a novel on a typewriter.

Are there people who aren't currently at funds who could be awesome quants?

  Absolutely.  Just working on the markets as a full time job is a large part of succeeding.

Can some of these strategies that quantopian users are coming up whit make money?

  Of course, alpha is everywhere.

The biggest hurdle is again taking the theory and putting it into practice. And then redoing it year over year. I mean making money from 2008 until now isn't really proof of much as the market has been up and to the right almost the entire time.

I don't want to sound really dismissive, but until a quantopian fund successfully manages say 200 million with returns of S&P + 5% over a period of say 5 years its probably best to cool the hype:)

And I say this as someone who is a big fan of what they are doing. It's just in finance you see alot of hype and you learn very quickly, that the scoreboard is everything, nothing cuts through bullshit faster than actual returns data net of fees.

I've written alot about quantopian and them turning themselves into a fund of sorts if anyone is interested.

https://news.ycombinator.com/item?id=12335272#12336086

https://news.ycombinator.com/item?id=12171843#12173142

https://news.ycombinator.com/item?id=11207569#11211122

Agree with your points. And also skeptical of the BLK comment - though for a venture capitalist joining the board, making this investment sound like it's a future $60B market cap firm is par for the course.

I'm in the industry as well, and I see recruiting and training of new quants to be very inefficient and very expensive right now. I've met for coffee with PhDs in Physics with job offers already from HFs in hand and their first question to me is "what does a quant HF trader do?"

What are your thoughts of Quantopian becomes used more for recruiting and training quants for other HFs than as an asset manager itself?

from 2008... as the market has been up and to the right almost the entire time

Was there a time that the market did not go to the right since 2008?! Where's Dr. Strange?!

something learned during the recent campaign was where Donald Trump's inherited money originally came from: the Yukon gold rush! but not from gold, it was bled from all the people chasing after the gold, most of whom lost their meager investments. This is the business model A16Z et al is pursuing here. And if gold is discovered, they'll be first on the scene.

I'm a fan of Quantopian's product but I don't understand why anyone would invest this kind of money in their solution. They'd have to manage an enormous amount of money to have any reasonable growth. Add to this that the hedge fund world is consolidating, I just don't see the value. What's the out here? I think management fees are going to close to zero soon (if they are not there already).

A) Recruit, hire, and train a handful of quants, pay them $250k/yr+ each, and hope they bring a positive edge to your algorithmic trading.

B) Award $10,000 to the winner(s) among hundreds of participants aggressively competing to solve the same problem as your expensive employees from A.

Which would you choose?

You assume the outcome would be the same - I see no evidence to support that B > A. Given that A is the incumbent business model and form, B would have to be run for 5+ years to prove that there is anything there that can win consistently and also scale to the needs of the institutional investor.

Indeed, if you are a new fund starting out and seeking institutional money and are not coming from an established firm, you can expect to have to develop 3-5 years, minimum, of track record before anyone will generally invest in you. Given that any trader on this platform might disappear overnight, I don't see institutions rushing to put money into it.

Perhaps (just perhaps), Quantopian is gathering (or has gathered) the evidence that B > A. If so, this is a great investment.

Regarding institutional money, they have been around for a few years. I'm not sure about how long they've been running the competition, but they certainly have been running an asset management service for some time - and would have impeccable records regarding this.

In other words, there is a lot about Quantopian that we don't know and so it would be difficult to make a judgement call on the quality of AH's investment.

My point is that I think it has a fundamental problem - if someone has a successful algo, they take it off the platform. Institutional money would be very leery (indeed, this isn't the first time this model has been tried).

So even if I'm confident in their overall model, based on, say, 5 years of track record, I'd still be concerned over algos just disappearing. Also, they've only been running their asset management business for a relatively short time via the investment capital from Cohen's fund.

I think when you submit an algorithm you give quantopian non exclusive rights to use it indefinitely. Even if you stop participating the best algorithms will still be used.

Update: looking at the FAQs that applies to contest submissions only. They state in the FAQ that all algorithms developed and evaluated on the site are by default owned by the developer. Entering the contests give them license to use.

I worked with them for a little while. At a fundamental level, its a brilliant model. They're basically looking for strategies that have little to no correlation to the S & P. Also, any strategies that short the S & P market successfully are welcomed. They already have enough of the trend following, or mean reversion strategies which are found everywhere in quant trading books.

Has anyone participated in one their programming contests? Just curious how approachable they are to someone without experience on Wall Street.

Their contests evaluate your algo on a number of heuristics, namely high alpha and zero beta. They are agnostic to you or your background.

Is Quantopian a good idea for an entry-level quant-wannabe?

Why not?

Yes. At least for two reasons: 1) great community from which you can learn a lot. Which will also give you a feedback if you post your ideas. 2) quick and easy way to see which of your ideas do not work.

no, its nowhere near professional level. for that you'd need clean price data, which is a business in itself.

Clean price data is one of Quantopian's features. They've put effort into normalizing price data for backtesting. For example, they deal with stock splits for you, so that you don't accidentally treat an 10-way split as a 90% loss.

Quantopian will bill you immediately for the first month's subscription.

How much does a month's premium data subscription cost?

Each premium data set is priced differently. Prices are set by the data vendor, ranging from totally free to $150/month. Most are in the $5 - $50/month range. (I work at Quantopian).

Reply on news.ycombinator.com