Quant funds in trouble this time. Any insights into what went wrong?

Discussion in 'Professional Trading' started by helpme_please, May 4, 2019.

  1. Maverick1

    Maverick1

    I don't think so. Those who use quant work as a tool to assist fundamental analysis rather than the whole enchilada itself will do better.

    But if people keep thinking that pure quant is the way of the future, they are sorely mistaken. The results are starting to bear that out...
     
    #21     May 9, 2019
    lovethetrade likes this.
  2. lovethetrade

    lovethetrade Guest

    Is this the same person that thinks the marginal cost of electricity to run his trading system is less than a free alternative that delivers a host of other benefits that can increase one's competitive advantage?

    Must be the same person that thinks he has more control running his hardware and system locally than in the cloud using SSH and FTP.

    Welcome to the 1950s.
     
    Last edited by a moderator: May 12, 2019
    #22     May 12, 2019
  3. This is terrible. Utterly bad timing. Having highest level of equity exposure near market peaks.

    https://city.wsj.com/articles/6acdd08c-8f2a-4fdc-89fa-61a5719fa8ec

    Key Facts
    • Computer-driven volatility-targeting funds generally scoop up riskier assets like stocks during calmer periods.
    • When volatility hits, it sends them scrambling to sell their stocks.
    • At the same time they move into safer assets like Treasurys.
    • Asset managers like Vanguard Group and insurance companies run some of the bigger strategies of this type.
    • Volatility-targeting funds had an estimated 44% equity exposure last Tuesday.
    • That was their highest level of equity exposure since early October, said Wells Fargo Securities.
     
    #23     May 13, 2019
  4. Quants Think Like Amateurs in the World’s Wildest Stock Market
    Bloomberg News
    May 15, 2019, 12:00 PM EDT

    • Modeling the behavior of individual investors can be tricky

    • BlackRock monitors about 100,000 online chat posts a day

    Tom Zhou runs a $500 million hedge fund and has degrees in civil engineering, economics and finance.



    Like other quantitative trading whizzes attempting to make sense of China’s $6.6 trillion stock market, Zhou spends much of his time trying to think like a novice investor.

    As quants adapt their models to a Chinese market where mom-and-pop investors drive more than 80% of trades, they’re coming up with clever ways to predict where the nation’s so-called dumb money is headed. Getting it right isn’t easy in a country with the world’s highest stock-market volatility and price swings that often appear to defy logic.
    To anticipate how China’s 147 million retail investors will behave, quants are combing through social-media posts and using artificial intelligence to predict when popular technical indicators will spur waves of buying and selling. They’re buying troves of data from the likes of Tencent Holdings Ltd. to gauge investor sentiment, and weeding out factors that work well in the West but fail to outperform in China.

    The efforts underscore how international investors will have to think differently as they increase exposure to Chinese stocks in the wake of the country’s entry into MSCI Inc.’s global indexes.

    “In the U.S., quants are trying to make money off other institutional investors with complex models or automated transactions at lightening speed, but in China many strategies don’t work well and quants’ arch rivals are retail investors,’’ said Zheng Xu, a former portfolio manager at Millennium Partners who now teaches finance at Shanghai Jiao Tong University. “Understanding retail investors’ behavior and sentiment is extremely valuable here.’’

    While quants are typically loathe to give away details of their approach, a few were willing to share the broad outlines of how they model the impact of China’s individual investors on the nation’s stock market.

    Zhou, a former quantitative analyst at MSCI Barra who’s now chief investment officer at Shanghai River East Asset Management, said one phenomenon that stands out is Chinese traders’ tendency to lock-in profits more quickly than their peers in the U.S. That means short-term price reversal factors tend to perform better in China than momentum factors, he said.

    At High-Flyer Quant, which oversees more than 6 billion yuan ($870 million) in Hangzhou, traders use AI to anticipate as many as two days in advance when widely followed indicators including moving average convergence divergence, or MACD, will prompt individual investors to buy or sell, according to Simon Lu, High-Flyer’s deputy head of research.

    BlackRock Inc., the global investing behemoth that has big plans to increase its presence in China, tracks changes in retail-investor sentiment by analyzing social media data, including about 100,000 chat-room posts a day on websites like Eastmoney.com and Xueqiu.com. The firm buys stocks that attract growing attention from investors and sells those that show waning interest.


    Such factors aren’t always reliable. Wang Pei, chief executive officer of Shenzhen-based Focus Technology Ltd., said his fund gained an extra 7% to 8% annually in 2013 and 2014 by incorporating data on the most-viewed stocks in trading apps operated by Tencent and Hithink RoyalFlush Information Network Co. But the factor stopped working in 2015 as competitors copied the approach and China’s market crashed, Wang said.

    The country’s unique market structure also poses challenges for quants. The biggest hurdles include a dearth of liquid hedging tools and a rule that prevents investors from buying and selling the same shares in a single day, which makes some high-frequency trading tactics unfeasible.

    China’s censorship of social media -- and the constantly evolving online slang that netizens use to evade official monitors -- can also present challenges for firms using AI tools like natural language processing to monitor investor sentiment.

    The median return among funds linked to China’s CSI 300 Index beat the benchmark by 3.63% in 2018, according to Citic Securities Co. In the first quarter of this year, the funds underperformed the benchmark by 3.22%.

    Prospects for outsized returns are enticing enough that international quant shops including Boston-based PanAgora Asset Management Inc. are increasingly experimenting with new models to trade the Chinese market.

    parse the slang Chinese retail investors use to discuss stocks on online message boards, which the government censors. He became interested in the challenge in part because of some quirky moves in Chinese stocks in the wake of the 2016 U.S. presidential election. When the result became clear, a listed Chinese company whose name sounds like “Trump Wins Big’’ in Mandarin surged, while a firm that sounds like “Aunt Hillary’’ slumped.


    “There are a lot of inefficiencies, and many of these are driven by retail investors,’’ Chen said in an interview. PanAgora hasn’t yet used the model for trading. “You want to understand what they think, but you can’t do it through the traditional financial statements.’’

    — With assistance by Jun Luo, Evelyn Yu, Dingmin Zhang, Saijel Kishan, and Jeff Kearns
     
    #24     May 15, 2019
  5. "Quant" is just an outdated idea.
    There is no one running any money that is not using some form of programming and modeling in 2019.
    To make the narrative that "quant" is a single variable that everyone is doing the same thing in 2019 is utterly ridiculous.
     
    #25     Jun 4, 2019
    qlai likes this.
  6. There is a ton of hate on HFT. Many on the forum love to blame their losses on it. When I watch the markets (and I have a real edge) I can see a lot of predator prey behavior in the quotes. It seems like the guys that have the real edge (large buyer) are VERY dominant when they want to be. Its almost like they just prey on the seller like its a helpless animal. I think understanding HFT cash and carry arbitrage is one of the best edges in equities now.

    Oh and to the point above. The assumption that a price process is stationary is joke. The theory of parameter estimation assumes parameters do not vary during sampling. This is required for almost all of the convergence theorems to apply. Even a doubly stochastic process is a nightmare to model. Price processes are much more like predator prey system, but not any of the deterministic ones. Maybe I meant to say compounded distributions instead of doubly stochastic, but you get the idea, i.e. non stationary. The point is, the complexity increases (hugely) when you can't make that assumption.
     
    Last edited: Aug 18, 2019
    #26     Aug 18, 2019
    Van_der_Voort_4 likes this.
  7. qlai

    qlai

    Sorry for my ignorance but is it just an analogy or a scientific term? What does this mean in practice? If it's not deterministic, that would mean it cannot be automated well enough to get an edge? Thanks!
     
    #27     Aug 18, 2019
    murray t turtle likes this.
  8. Hey qlai I will try to explain what I mean.

    I used to develop processes that would do things that you see in markets. The model was 'self referential' in that it was a stochastic process whose variance was a function of the realized process trend.

    In other words the variance of the process tended to increase when the process became 'trendy' which would cause the trend to be broken which then reduced the variance. The model worked pretty well to simulate markets because you see this happen all the time. Markets rally, then break, then drift, then rally etc.

    My point here is that a process with self referential parameters is not what statistics is designed to deal with. It creates randomness that is itself (seemingly) random. Hope that makes sense. This stuff is hard.
     
    #28     Aug 18, 2019
    qlai likes this.
  9. Overnight

    Overnight

    So it means you were spoofing?
     
    #29     Aug 18, 2019
  10. TommyR

    TommyR

    i agree. often its cheapest to smash a globalist into randomness to make other areas more trendy roi wise. no spoofing involved
     
    #30     Aug 24, 2019