The Renaissance common-sense test
Email 
Print
Robert Frey, formerly of Renaissance Technologies, has a new Fund of Hedge Funds, and good for him. He also lifts the kimono ever so slightly on how Renaissance works:
At Renaissance, models had to meet four principles, says Mr Frey. These were (and maybe still are): simplicity – “don’t make it more complicated than it needs to be”; commonality – “make it as broad as possible”; stability – “models you have to readjust constantly probably aren’t as good as ones that stand the test of time”; and rationality – “it can’t just be statistically valid”. You have to employ reason to identify a statistically significant but spurious pattern.
I think that quants in general would pay lip service to these principles, but they wouldn’t necessarily give them such a central importance. Of course, the proof of the pudding is in the eating: what counts as simple and stable for Renaissance’s purposes would probably be considered nothing of the thought by mere mortals.
But I do like the final “common sense” test. “This strategy works, but I don’t know why” is always a bad way of trying to make money, because it’s very likely to be a statistical fluke. Ideally, of course, there would be a sequencing test too: it’s not enough to come up with a strategy which works and then try to work out why. You have to start with a theory of why a certain strategy might work, and then test it. I wonder whether Renaissance does that.
Comments RSSPost Your Comment
Another useful step might be to work out why the strategy would continue to work in the future.
This is not particularly revealing– you need to see how it all unpacks in-real-life. So, suppose the model says ‘X’ but your gut says something else. Do you 1)ignore the model?, 2) ‘fix’ the model?, 3) follow the model? 4) rephrase the question? 5) change the subject?
The broadest way of putting it might be, when it works, to go “I meant to do that.” When it doesn’t, blame Europe.
MattF, first thing you do is try and identify the mismatch between your gut and model and why they are disagreeing. Second is to see what your gut says and what the model says vs the real-life. If one disagrees with the data then it is wrong. Most of the time the disagreement boils down to some market microstructure that you forgot to model – ie tax or regulatory differences or liquidity assumptions. Of course alot of the time it is not that clear cut in that case you need to use your experience and skill to make the decision….
The main that was interesting for me was the focus on not losing money – which sounds like an obvious statement. But most people don’t focus on this, they focus on MAKING money. Fact is that if you made market rate during the boom years and zero losses in the down years you are over time going to have pretty decent performance.
The Economist, amusingly, is saying the reason RenTech succeeded so wildly is because … they don’t obey any common sense rules, and do, in fact data mine to their heart’s content. We all know those crazy physicists and astronomers are big on mindlessly using false correlates.