They use big data techniques to find weak correlation in data and invest moderate amounts of money in them. They never make a large return on any given investment, but they make very persistent returns; over the long run this adds up.

The limiting factor behind what they do is that it doesn't scale.

> How did these guys make 98.5% in 2000 and 33% in 2001 in the depths of the .com collapse of Wall Street?

The name "hedge fund" comes from the concept of a hedge; a strategy that will underperform in good times, but over perform in bad ones. 2001 is the exact time you'd expect a hedge fund to do well, especially one focused on dispassionate data analysis.

It was said before, they don't predict every word in a book. But if they see "New York" in a sentence they will bet the third word is "City".

This is good answer.

It's also how profitable sports betting syndicates work. Lots of small bets with relatively low expected value. 1% expected return per week is 70% return in a year. You can't scale the syndicate above certain limit.

There are many ways of achieving extremely high Sharpe ratios (e.g. exploiting market microstructure inefficiencies), with following caveats:

1) They are not scalable (for each such strategy, market turnover capacity is very limited, and increasing it requires coordinated efforts of many smart people). This is the reason they are not taking additional investments and strictly invite-only.

2) Operations efficiency becomes the key. Execution capability has to be consistently at top of the market, and margin for error becomes nearly nonexistent. It is very easy to screw things up. Hence the need for top technical and engineering talent.

Actually that is the least suspicious. They knew it was a bubble and bet against it. Like the guys in The Big Short. You make money when the market is far from reality.

You might want to take a look at a book entitled "The Quants" by Scott Patterson. He talks extensively about Medallion. Inception, speech recognition people, etc.

I dropped that book. Fortune's Formula is much better.

For those interested in either, can you give more information/opinions to sway one way or the other?

I can't say anything about the other book, but "The Quants" is a good read and dives into the history of these funds.

I also liked the style for what is not there. Usually, people addressing these questions try to make it look like magic and easy. I hate the 5 minute montage in movies with cool music that condenses 5 years of hardship. You can pick it up, read some pages and put it down if you don't like it.

The author seems to have had access to the people he talks about, and had help from the right people (Mandelbrot, Taleb, Ed Thorp, to cite a few). For me, this increases weight and it makes it cool.

An edX course about Discrete Time Signals? That's nice. Finding out that the instructor is Richard Baraniuk knowing that I used a paper of his (an adaptive optimal kernel) [to do time-frequency analysis for multiphase flow pattern recognition]? That's cool!

You're not supposed to ask those questions. ;)

However making money in those two years is nothing unusual.

This is categorically incorrect. They are not a high frequency shop, and they hold positions overnight. Two Sigma, however is a high frequency shop.

(I work in the industry.)

Renaissance Technologies is, somewhat famously, not a HFT shop.

The only remaining option is pair correlations / volatility arbitrage. :)