Information about the Chinese financial markets is often fleeting, sparse, and long after the fact. This week, however, Bloomberg brought a piece on a recent Quant Quake, a headline too good to pass up.
The piece talks about how Chinese quant hedge funds—hedge funds that manage portfolios based on rules-based (quantitative) models—got caught in a market storm that not only rendered their models useless, but outright wrong.
Chinese quant funds have far outperformed managed or discretionary funds and their peers across the globe in the past years because the computer-driven (and in some cases allegedly AI-driven) models can consume more market information and react faster and more precisely to market events than human traders.
That apparently went topsy-turvy in the two weeks leading up to the Chinese New Year.
Chinese stocks have been in a slump lately which the quant funds were navigating just fine by continuously calibrating a blend of long and short positions that kept them market neutral with an upside edge.
The slump turned into a downward slide, and to prevent the stock market from free-falling, the Chinese authorities intervened by propping up government-led funds and imposing restrictions on short-selling. Market movements became erratic and unpredictable. Models were trained on calmer historical data. That did not turn out right.
The few comments that have emerged from parties involved refer to this as a massive black swan event. As mentioned before, a black swan event is something so unprecedented, unlikely, and unimaginable that we would have to revise our understanding of the world.
Like your grandpa enjoying a beetroot latte.
Please note the word unprecedented as in has not happened before. As in never. As in not in a million years. Only, it kind of did. During the Quant Quake of August 2007, some US quant funds experienced similar surprise losses because their models did what they were trained to do, only it wasn’t the right strategy anymore.
Our point is: keep your grandparents away from pretentious hot beverages and dark birds.
Also, although modeling tools improve all the time, models are still only as good as the data they are trained on, and there is an inherent trade-off between having models predict what is normal and what is extreme.
Regitze Ladekarl, FRM, is FRG’s Director of Company Intelligence. She has 25-plus years of experience where finance meets technology.