And so it begins

Today I am taking my first step toward building a quant hedge fund completely from scratch. I have no partners or intellectual property, and only enough savings to support my family for five years at our current burn rate. I know it will very likely not work out, but I’m still going for it.

NLP carve-out

I don’t want to quit my job in a way that is disruptive to my employer, especially because I’m working on a project that is nearing production. So, my boss and I have arranged for me to drop down to half-time while I finish my project and train my replacement. His lawyers have written up a side-letter that will permit me to do research on my own time and equipment, in one narrowly-defined area I am fascinated with: natural language processing on company news and announcements. I am excited to begin developing my own intellectual property. I studied some NLP during my PhD, but I need to catch up with the deep neural network research that has taken over the field.

Fool’s errand

Evidently, what I’m attempting is impossible:

If you’re going to take the leap to start a fund, you can’t expect to succeed with “a guy, a Bloomberg and a dog.” That may have worked ten or twenty years ago, but it won’t work today. Competition with the big boys is fierce, regulation is more onerous, and allocators are more sophisticated than ever.

-Ted Seides

These quotes are everywhere:

To start a true, institutional-quality hedge fund that uses the LP / GP (Limited Partner / General Partner) structure and has large external investors, such as endowments, pension funds, and funds of funds, you’ll need to raise hundreds of millions of USD.
The bare minimum to get noticed is $100 million, but realistically it’s more like $250 million+, and ideally more like $500 million – $1 billion.

You have no chance of accomplishing that unless you have deep connections to potential Limited Partners and a great track record over many years at an existing fund.

… high returns on small amounts of capital (i.e., millions of dollars or less) do not mean that much.

… results from “personal accounts,” no matter the account size, are not taken seriously.

… if you start with, say, $5 million, you will not have enough to pay yourself anything, hire others, or even cover administrative costs.

Survival strategy

I hope to succeed by following this path:

Live lean – Keep costs to an absolute minimum. I will work from home, including building a makeshift datacenter out of refurbished computers. I will buy low-cost tick data, and I will start by looking at the tens of millions of free, publicly available SEC filings.

Low capacity niches – I will begin by searching for market inefficiencies in areas ignored by sophisticated players. No institutional investor would be interested in even a high-Sharpe trading strategy that only had the capacity to earn $100k/yr. But a strategy like that would be very valuable for extending my runway. So, I am going to search out strategies with short hold times, relatively infrequent trade opportunities, or in small-cap stocks. Only after I have a little success will I attempt to find higher-capacity trades.

Incubator fund – This will make my track record more official than just trading in a personal account. An incubator fund is half-way to a hedge fund. It will only hold my personal capital, so there is no need of a private offering memorandum, partnership agreement, or subscription documents. Incubator funds only cost about $5k to set up, and they can be converted to a full-fledged hedge fund later, with the track record intact.

Possible outcomes

Most likely I will fail, tuck my tail between my legs, and go try to find a job. But if it does happen, success could look like any of the following:

Hedge fund – I’d love to be able to attract capital and some teammates to launch a small hedge fund. I hear the 2-and-20 fee structure is dead, but I might still be able to charge 0-and-20 while keeping my costs low. I understand this is a long shot, particularly the fund-raising part.

Funding deal – There are companies out there that will let you keep your IP and entity, but they become your source of funding. These include groups like Paloma and Schonfeld. This would take away the entire burden of fund-raising.

Multi-strat fund – Some hedge funds and prop trading firms will bring your group in-house to trade their money. You share your IP and get paid a percentage of the profits, but you would be part of their entity. These funds also have sophisticated trading infrastructure you will integrate with. Millennium Partners is one example. They have a reputation for giving you a chance, but dropping you as soon as you lose money.

Portfolio manager – By this, I mean just taking a job at an existing hedge fund. I’ve known several people with nice arrangements where they work as part of a team, but their strategy’s performance is tracked separately (although the desk’s capital is pooled), and good performance leads to a good annual bonus. This can be a nice arrangement, especially if you can get a contract that allows you to jointly own improvements you make on your strategy while employed there.

Managed accounts – Investors may be uncomfortable wiring money to a small hedge fund manager. So, rather than taking investor money and adding it to a pool, it is also common to just manage multiple investor accounts individually. I would have trading access to the accounts, but not the ability to transfer funds. I would still be able to charge a performance fee. The main disadvantage is the added hassle. For example, each trade needs to be broken up and allocated among all the accounts, and regulations require the allocation process be fair. Allocation differences will eventually lead to differences in account performance, which complicates the calculation of the track record. Interactive Brokers has built systems for helping registered investment advisors manage accounts through their platform.

PA – Perhaps it will be possible to support my family just by trading my personal account. This might make sense if I can get leverage, or if my research yields only low-capacity strategies. It sounds a little isolating long-term, though.

Why blog?

Yeah, that’s not totally clear to me. I will obviously have to be very careful for this to not become some sort of advertisement to solicit investors, as that would be illegal. This blog also won’t be a place where I share proprietary results from my research.

I think it’s partly a bit of academic culture that remains with me. There is some satisfaction and sense of accomplishment in publishing, even if it is just on a blog.

It is also sort of a journal. Whether I succeed or fail, I want some record of this journey to remain.

If I’m honest, I suppose it is also a tool to help me find a job if I fail. Looking for work after giving up, after struggling in obscurity for years, after burning through all my savings will not be so easy. Hopefully my technical writing will give hiring managers an idea of what I am learning. For example, in writing this post, I learned how to overlay meme text onto an image in linux!

convert -font impact -pointsize 75 -fill white -draw 'text 360,460 "AND SO IT BEGINS"' helms_deep.png and_so_it_begins.png