II blames style-drift for CTA woes

[Maverick1]

I was just listening to Leda Braga talk about "taking the emotion out of trading" with systematic approach and couldn't help but think what a load of wishful thinking. As has been pointed out on this thread, there are enough discretionary decisions being made over systematic strategies over time that this type of assertion is easily disproven. Discretion is everywhere, most significantly in the model development arena, where people are known to overfit their models to the last battle instead of truly coming up with robust parameters (which are hard to find!). Believe me, when the manager is looking at his trend following strategy return profile deteriorate and has to decide whether to reduce his allocation or not, he is sweating just as much as the discretionary pm.

There are so many variables that a discretionary manager can capture that a computer can't... Until the day computers can not only understand text but also read between the lines and make connections with other information like an experienced human brain can, I'm bearish on quants. Sorry, their ideas are just not original and most of them do not produce any alpha. The few that do will see their returns plummet just as the value guys saw their returns dwindle over time (btw the momo guys are next). Right now they're hidden by the high tide of a rising stock market and the "uncorrelated to the market or other funds" defense. You just wait till the next crash and watch the correlation go to 1 quicker than a flash flood on a Thailand cave... Then we'll find out who's been swimming naked.

The exception is Renaissance of course, but not everyone can be the Michael Jordan of trading

I mean look at Cubist right. So Stevie hires the best quants you can find in the business, they lose money for 2 quarters, now they're up decent. End result is probably flat to slightly up, hardly anything that wows...

 

[ironchef]

The exception is Renaissance of course, but not everyone can be the Michael Jordan of trading

What makes them different?

 

[Maverick1]

What makes them different?

I have some thoughts but wouldn't do nowhere close to justice... I think they're just a different kind of clever. Note clever is not same as intelligent, although they have plenty of that. Intelligence is everywhere but clever is not

 

[sle]

A long post deserves a long answer. There is nothing magical about systematic trading and some guys are going to be good and some guys are gonna suck. The issue of capacity is more serious on the quant side - the systematic edge leverages the activity of other participants. So all the good alphas are naturally capacity constrained and vanish quickly as more people enter the game. Trend following in commodities is a prime example. There are natural producer/consumer dynamics that cause commodities to trend, but these days there are probably too many speculative players and these dynamics are disrupted by their flows.

I was just listening to Leda Braga talk about "taking the emotion out of trading" with systematic approach and couldn't help but think what a load of wishful thinking. As has been pointed out on this thread, there are enough discretionary decisions being made over systematic strategies over time that this type of assertion is easily disproven. Discretion is everywhere, most significantly in the model development arena, where people are known to overfit their models to the last battle instead of truly coming up with robust parameters (which are hard to find!). Believe me, when the manager is looking at his trend following strategy return profile deteriorate and has to decide whether to reduce his allocation or not, he is sweating just as much as the discretionary pm.

Well, Leda is talking her book, of course, no surprise there. However, it's hard to deny that a scientifically developed model is far less likely to fall victim to cognitive biases and emotions than a the one that arises as a part of the discretionary process. At the very least, a systematic strategy has fewer opportunities to fail for cognitive reasons. It can be tainted either at the development stage or at the stage of decay (de-allocation trigger) while a discretionary trader has to make decisions about his actions every day. There is a reason why you don't see many quant blowups outside of systemic crises - systematic alphas fail in a non-spectacular manner (i.e. bleed out because other people started doing it better or faster). The renown quant blowups were non-systematic (e.g. LTCM).

As a aside, there is really no straight-forward division between discretionary and systematic, even though both sides would love to present it this way (well, you sure do ). Many discretionary guys have a "system they follow" and many systematic guys have a bunch of knobs (that go even beyond allocate or de-allocate) that they tweak on regular basis.

There are so many variables that a discretionary manager can capture that a computer can't... Until the day computers can not only understand text but also read between the lines and make connections with other information like an experienced human brain can, I'm bearish on quants. Sorry, their ideas are just not original and most of them do not produce any alpha.

Quant strategies are predatory by nature, in that they don't need to develop original ideas but rather leverage on what other humans are doing. A lot of the quant edge is, in effect, front-running the discretionary guys by understanding their flows, finding the trigger keywords in the news etc. The computers don't need to read between the lines, they just need to have a statistical sample of what happened last time these news items have arrived.

I mean look at Cubist right. So Stevie hires the best quants you can find in the business, they lose money for 2 quarters, now they're up decent. End result is probably flat to slightly up, hardly anything that wows...

Most senior PMs at Cubist have been with Stevie for over 15 years and SAC had a separate systematic strategies bucket for a long time, so there is nothing "new" about them. The fund, predictably, suffers from the usual curse of a large multi-manager shop - i.e. PMs want what's good for themselves and not what's good for the firm.