Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
AI The Almighty Buck

$350M 'Quant Fund' Played the Stock Market Using Real-Time Data -- Then 2020 Happened (thenextweb.com) 72

An anonymous reader quotes a report from The Next Web: Coatue's $350 million data-driven 'quant fund' made a swift exit from the market in early April -- having realized the coronavirus pandemic had rendered its algorithm unreliable, Business Insider reports. The tech-focused fund, launched by billionaire Philippe Laffont just over one year ago, joins a growing list of "quant funds" to have failed to turn a profit due to 2020's unpredictable nature. Quant funds use complex algorithms to find hidden trade signals in a hyperconnected web of data. While Coatue's quant fund mixes old-school stock picking with quantitative analysis (a method dubbed "quantamental"), execs were reportedly concerned that data farmed in the midst of COVID-19 would confuse its in-house trading program.

One example cited by Business Insider's sources highlighted the way Coatue's program interpreted ecommerce data. It reportedly showed surges in website traffic for certain retailers as COVID-19 lockdowns spread across the world -- usually a positive sign for stock -- but failed to consider their dwindling revenues and closed physical stores. But Coatue's quant fund had underperformed long before COVID-19 hit. In February, reports surfaced showing it had posted only 2% returns since its launch in May 2019, and had actually lost money (1.2%) in last year's fourth quarter. On the other hand, Laffont's human-led fund bested the industry average to return 10% profit last year. Bloomberg has since noted that those profits have taken a hit, bringing its losses the year to roughly 6%.
While quant funds as a whole have had a really tough year, there were at least two that are doing just fine. "Toronto's Castle Ridge Asset Management, which trades some $100 million in assets, made 2.6% in March with its 'self-evolving' AI system that works with large-cap stocks," reports The Next Web.

"Over in Sweden, Volt Capital Management AB (in charge of roughly $30 million in assets) has returned a loin-tickling 24% to investors this year. Volt's fund was also reportedly prepared for those plunging oil prices."
This discussion has been archived. No new comments can be posted.

$350M 'Quant Fund' Played the Stock Market Using Real-Time Data -- Then 2020 Happened

Comments Filter:
  • this type analysis by feeding it fake information, then taking advantage of the trading patterns?
    (Would it be legal?)

    I assume the trading takes place so quickly most of the time, that there isn't any human oversight.

    • this type analysis by feeding it fake information, then taking advantage of the trading patterns?

      How would you inject false information into, say, Bloomberg data?

      (Would it be legal?)

      Absolutely not. But white-collar prisons are not so bad. Do you enjoy tennis?

      I assume the trading takes place so quickly most of the time, that there isn't any human oversight.

      Quants and HFTers are different but there is overlap.

      Much of what quants do is known as "picking up nickels in front of a steamroller."

      You almost always make a small profit. Until a black swan comes along, your model fails, and you get squashed.

    • this type analysis by feeding it fake information, then taking advantage of the trading patterns?
      (Would it be legal?)

      I assume the trading takes place so quickly most of the time, that there isn't any human oversight.

      How would you feed their trading program "fake information", unless you were one of the program's developers?

      If you *were* one of the developers, and you tried to make money based on your inside knowledge of what the trading program is about to do... yes, that would be quite illegal. I believe it's known as "front-running".

      • The answer is fairly obvious, they are looking at site traffic for businesses to indicate a positive effect on the stock. you could easily have a lot of robots load and browse a website.
    • of course you can manipulate it, you can also manipulate human driven ones in the same way, it happens all the time. And NO it is not legal.
  • You gotta love those dumb quants.

  • Comment removed based on user account deletion
    • Another great example was the collapse of Barings Bank.

      Nick Leeson had a sure thing, going long on Japanese derivatives ... until a black swan showed up: The 1995 Kobe earthquake. Oops.

  • When I saw the pandemic hit I switched into home entertainment providers (Netflix, Disney) and big box retailers that would remain open (Home Depot, Costco), plus I shorted oil stocks figuring that people would stop driving - I'm up something like 22% since February.
    • It is crazy, I bought my first stock ever around the end of March, figuring it was time to try things out with the market way down. It has to go up, right?

      But this is not the time to learn trading stocks because right now almost everything is going up. I can understand how an algorithm trained under normal conditions might go errant prone.

      With about 15 somewhat diverse stocks I'm up 15% overall, they are all up except one which I intentionally bought for the experience of selling, HTZ, lol.

      Tax time
  • by starless ( 60879 ) on Wednesday June 03, 2020 @06:29PM (#60142504)

    having realized the coronavirus pandemic had rendered its algorithm unreliable

    Is the SARS-CoV-2 virus then the equivalent of Asimov's "Mule" (from the Foundation series) in disrupting the predictability of socio-historical trends?

    https://en.wikipedia.org/wiki/... [wikipedia.org]

  • "Over in Sweden, Volt Capital Management AB (in charge of roughly $30 million in assets) has returned a loin-tickling 24% to investors this year. Volt's fund was also reportedly prepared for those plunging oil prices."

    Loin-tickling? WTF? I must remember to use it in conversation!

  • Coatue's $350 million data-driven 'quant fund' made a swift exit from the market in early April

    Shall we call it coatues interruptus?

  • by Aighearach ( 97333 ) on Wednesday June 03, 2020 @07:28PM (#60142692)

    So that's whose money I just made. Thanks, guys!

  • Early April was the bottom. Ever since then, the stock market has been doing what it had been doing up until March - nonstop rallying and ignoring any bad news whatsoever.
  • Stupid article (Score:5, Interesting)

    by ath1901 ( 1570281 ) on Thursday June 04, 2020 @06:08AM (#60144004)

    In other news: A roulette player returned 3500%. The player was reportedly prepared for the number 17 to turn up.

    I really hate this kind of reporting. It is cherry picking a few examples from a huge random sample and pretending there is some kind of clever or poor decisions that explains the result. Are any of the examples in the article statistically significant in any way?

  • "execs were reportedly concerned that data farmed in the midst of COVID-19 would confuse its in-house trading program."

    No, this is an opportunity to improve. This isn't confusing data, this is just more data on how markets behave. The stress illuminates flaws in their models, like the mentioned ignorance of physical store behavior, which they can correct in order to produce better predictions.

    Or maybe the real problem is that they didn't really have the kind of machine learning model they claimed to.

1 + 1 = 3, for large values of 1.

Working...