By now, the February trials and tribulations of the CTA crowd are well documented.
For quite some time, folks had been warning that systematic strategies were running high exposure and could be vulnerable to forced de-risking in the event market conditions were to suddenly deteriorate and that’s precisely what happened in early February.
On February 8, we asked if CTAs (which have become the market bogeyman par excellence over the past several years) might have just suffered one of their worst 5-day drawdowns in history.
According to BofAML, CTAs and risk parity were forced to offload some $200 billion in equity exposure thanks to the harrowing declines that began to materialize in the week leading up to the February 2 AHE beat, declines that accelerated following that print, and ultimately spilled over on Monday, February 5, when the dreaded VIX ETP rebalance risk was realized leading directly to the most dramatic VIX spike on record.
Fast forward two weeks and JPMorgan’s Nikolaos Panigirtzoglou asked whether CTAs were facing an existential crisis. To wit:
But the most recent underperformance during February is more problematic for the CTA universe because it casts doubt on the idea that Quant funds such as CTAs exhibit higher convexity and lower beta during market corrections. In other words, it casts doubts on the idea that during market corrections, Quant funds such as CTAs tend to outperform as their drawdowns are contained. In fact, the opposite happened during February’s correction. In our mind, February’s correction shows that CTAs can become victims of even modest profit taking when momentum or trend following signals become too strong and momentum positions become too crowded.
Well if you thought the pain was acute for the trend followers, consider what happened to AI funds. Here’s what Panigirtzoglou wrote in his latest note, out Friday:
Systematic fund unwinding was not only confined to CTAs and Risk Parity funds. Artificial Intelligence (AI) funds also slumped in February as revealed by the monthly Eurekahedge AI index which performed even worse than CTAs. This index went down by 7.3% the worst monthly decline since the index began in 2011.
AI funds are exactly what they sound like – that is, they’re hedge funds that try and leverage (figuratively and literally) artificial intelligence and machine learning in the course of attempting to, as Panigirtzoglou puts it, “predict and/or adapt more quickly to market trends.”
Well clearly, Sophia didn’t do a particularly good job “predicting and/or adapting” in February or if “she” did, that “adaptation” involved forced selling, which just serves to underscore the notion that quant funds may be dangerous.
Which reminds me of something I tweeted on February 19:
this is fucking awesome. now you'll be able to have a robot overseeing your smart beta ETFs.
"it's a factor rotation!!! sell Sophia!!! sell!!!!"
— Walter White (@heisenbergrpt) February 20, 2018
Panigirtzoglou goes on to ask three questions. To wit:
- Have AI funds also played a big role in February’s de-risking?
- Have the strategies AI funds typically employ become more crowded over time?
- Have their strategies become more correlated with those strategies employed by other systematic funds such as CTAs and risk parity funds?
Spoiler alert: yes, yes, and (probably) yes.
While JPMorgan admits they can’t definitively answer the first question, Panigirtzoglou nevertheless notes that in his mind (where there is real as opposed to artificial, intelligence) “the significant underperformance of AI funds during February, surpassing in magnitude even that of CTAs” is suggestive that AI funds were at least partially to blame.
As far as the second question goes, the preponderance of AI strategies points to “yes” and as far as the third question, well, here’s a correlation chart between AI funds and risk parity and between AI funds and CTAs:
So there you go. If you’re still looking to point fingers for what happened in early February, here’s a possible scapegoat: