Hedge fund quant AMA

Have you ever killed a man?

No one knows when the next crash will come or how it will come. If you have a well diversified portfolio and you're actually investing (i.e. not speculating or day trading), then just buy and hold.

Nope.

How can poorfags find and invest in a hedge find?

Thanks. What's your favourite company, and what are your thoughts on Novo Nordisk?

You cannot and should not.
en.wikipedia.org/wiki/Accredited_investor

I don't follow individual companies. The strategies and signals that I develop should work on an extremely broad set of companies or assets.

>en.wikipedia.org/wiki/Accredited_investor

In that case would you recommend just finding good etfs like on vanguard?

Depends where you are from. justetf is pretty good for EU investors

I'm from the US so I mostly just know about vanguard.

OP here (ID might have changed).

Yes. Find the ETFs that give you broad market exposure with the lowest fees/tracking error. Broad market exposure is probably the key here. Too many people think the S&P 500 gives you broad market exposure when in fact it gives you US large cap exposure.

I read World End Economica. The first part of the trilogy focuses on quants.
Do hot-shot stock traders tell you about having nightmares because you quantify their entire being into algorithms and greek symbols?

Not OP, but quants are why there haven't been hot-shot stock traders in about a decade. Also why bullpens are no longer.

I rarely interact with prop traders on the buy side, so I can't say. If anything, the sell-side traders are having more nightmares about the shift to electronic market making than about quants.

I've played WEE episode 1 and while it's relevant to my interests, it wasn't an accurate depiction of what we do at all.

are you the australian dude who did an ama like 2 weeks ago or so?

No. I'm American.

Financial world illiterate here..

What is an EFT?

A mutual fund by a different name?

And if they're really good to have money invested in, what's the best one that a Canuck can get into? (I have access to Canada, USA, and rest of America markets).

Do you work on any specific asset class?

What software do you use for models? Something in-house?

36mo beta?

Total n00b here, but what kind of information do you input into these models - is it total market value, is it stuff like how much cash is being printed? As I understand it's very macro yes? You're not looking at, say, how individual stocks are trading, you're looking at the flow of assets en masse?

Also how real-time are these models? Like are we talking, seconds, minutes? Or are these models built to last and it's like video processing where you need to "render" out so the computer can perform all the calculations?

How do you test these models? Like markets are the ultimate feedback environments so there's a real limit you can do to just simulations -- is every time you update or use a new model a huge risk?

Imagine if a mutual fund were a company that issued shares and you could buy/sell those shares like any other stock.

The main appeal is that they have lower expense ratios. Not sure about Canada, but you should do some of your own research if you're going to be investing your money.

Mostly equities and single name credit. All my work is done in Python.

Can't remember off the top of my head, but we're market neutral.

1. You can put anything you want that generates alpha. Fundamental measures, flows, satellite data, macro variables, etc. Ideally, signals should not be specific to too small of a subset. The greater your breadth, the greater your information ratio.
2. Also depends, holding periods can range from intraday to multiple months. I don't quite understand your analogy with video rendering, but machine learning based models will definitely need to be trained before being put into production.
3. Backtesting is the best you can do. New models are risky depending on what signals they use, so you probably want to give them very little risk to start with or trade on paper for a while first. In-sample and out-of-sample backtesting also helps.

Not OP, just gonna cover some shit cause I want him to give me a ballpark beta for last 3yr. It's a very uninteresting metric and I'm just curious how close to zero they got it.

He won't tell you shit about asset selection or anything proprietary obviously.

Any significant hedge fund will have an algorithm that runs many times a second.

Model optimization is usually approached through backtesting, where the algorithm is given developing inputs (the market environment) for a slice of the past as they happened, and told to react and make decisions based on those market conditions. See how well it does. Throw it stress events like the 07-08 crash, flash crash, stuff like that. It's a moving target; you want to look at the past to predict the future, but without overfitting and opening your system up to vulnerabilities, especially with black swan-type events. Backtesting is of course not reliable on its own, you always need to test out-of-sample as part of development.

You bitch. You beat me to it and stole my trips by 2 seconds.

... so I'm buying shares into brokerage that then just invests elsewhere the money I just invested and pays out dividends back?

So it would be like if I gave money to the Veeky Forums group, and they then said "hey Canuckanon, we gained 3%, here's your cut of the winnings".. and then what, just re-invests the winnings?

Oops.

etf.com/etf-education-center/21011-what-is-an-etf.html

It's cool. Broke 19-y/o NEET here. Been working on some models, but that doesn't pay jack shit until I make something good enough that it's worth licensing. Clearly debt is only solution short-term as there doesn't seem to be an entry level for quants w/o degree. Any degree / education-related recommendations?

It's tough even with a degree. Most places will want a masters or PhD from a top 10 or top 20 school in a STEM subject. There are a few good shops that take in quants right out of undergrad, but those are arguably even more competitive positions.

Go to a good college and do a STEM subject. Network hard, get good grades, and get internships/experience. If you don't make it straight into a front-office position, consider risk analysis or model validation quant positions too.

Wouldn't it be more worth my time to wageslave and develop a portfolio of valuable algorithms in order to earn a position in private equity through blood, sweat, and tears? This field will likely have been turned on its head three times over by the time I finish a masters / doctorate.

>The greater your breadth, the greater your information ratio.
Interesting.

> Also depends, holding periods can range from intraday to multiple months.
Ah okay. sounds like it's not like video rendering at all then.
thanks!

>He won't tell you shit about asset selection or anything proprietary obviously.
But are there things which are pretty universal to Quant? like variables or inputs that are pretty vanilla and always useful? Naturally the better models are the ones that have unique inputs and use them in new (accurate) ways.

>Model optimization is usually approached through backtesting, where the algorithm is given developing inputs (the market environment) for a slice of the past as they happened, and told to react and make decisions based on those market conditions. See how well it does. Throw it stress events like the 07-08 crash, flash crash, stuff like that. It's a moving target; you want to look at the past to predict the future, but without overfitting and opening your system up to vulnerabilities, especially with black swan-type events. Backtesting is of course not reliable on its own, you always need to test out-of-sample as part of development.
You've explained that very clearly. Thanks!

It might be more worth your time to wageslave, but developing signals that will get you noticed is unlikely. First, you're unlikely to come up with good signals without institutional support. Second, hedge funds simply don't go out and recruit quants like that. I don't want to discourage your passion, but being a quant is definitely not for everyone.

It's the Fundamental Law of Active Management.

Momentum and trend following strategies are probably the most universal.