Hacker News new | past | comments | ask | show | jobs | submit login

There was a video interview with the founder where he explained pretty much how they do it. The employ a lot of bright PhD maths/physics types, get them to come up with all the algorithmic strategies they can think of, run tests with historic data and live trading to see which ones work and then scale up those.

There isn't one smart guy or one great strategy - there are dozens of smart guys and loads of strategies and they can win because they outsmart the city types.

That sounds too much like this:


As I remember, beaucoup back testing, hilarity ensues anyway.

The issue with LTCM wasn't just that they didn't test sufficiently - the problem is that traders could take positions without supervision and any accountability. My understanding is that any firms that survived that period (and I believe they were all pulling the same shenanigans as LTCM, just maybe to a lesser degree) now have risk management because unlike Lehman/Bear they can't just foist that risk off onto the public market. No one at the head of a multi-billion dollar hedge fund wants to stop being there if all they have to do to endure is pay for risk management.

*edit: I'm not a serious student of this aspect, but I recall that none of the top independent hedge funds took a bath in the 2007+ collapse (that is, those that weren't in-house funds from a major wall st. company). They all had their risk management in order and all did pretty well in buying distressed assets. Some like Bridgewater really managed to grow non-stop right through that.

As I remember the book, they back tested quite a lot. It's just that whenever they went live with trading reality changed on 'em. Go figure.

They were exposed to a margin call. Trading reality changing or no, if you can get taken out by your prime broker at the end of the day, you need to account for that too and they didn't.

One thing backtesting misses is the other market participants may see what you are up to in live trading and try to take advantage. That was a big factor with LTCM.

They could lie about that too and just use linear regression.

A linear model on a heretofore unknown predictor is basically how all hedge funds make money.

Coming up with the predictor is often the hard part. For example, take the tweets of a (sane) president and run sentiment analysis on it. If it is positively correlated with mentioning an equity, the sentiment of the tweet might be a good linear predictor of the stock price.

The math is simple once the feature is well defined.

Feature development is the current frontier, as I understand it.

That's my point. They hire the smart people so they can really abuse the simple stuff.

No mention of leverage?

I'd take leverage for granted but it only works if the underlying bet you are leveraging is in your favour.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact