mechanical systems

Screen Shot 2015-11-18 at 4.55.47 PM

n an article for Bloomberg View last week titled “Why It’s Smart to Worry About ETFs”, Noah Smith wrote the following prescient truth: “No one knows the basic laws that govern asset markets, so there’s a tendency to use new technologies until they fail, then start over.” As we explored in WILTW June 1, 2017, algorithmic accountability has become a rising concern among technologists as we stand at the precipice of the machine-learning age. For more than a decade, blind faith in the impartiality of math has suppressed proper accounting for the inevitable biases and vulnerabilities baked into the algorithms that dominate the Digital Age. In no sector could this faith prove more costly than finance.

The rise of passive investing has been well-reported, yet the statistics remain staggering. According to Bloomberg, Vanguard saw net inflows of $2 billion per day during the first quarter of this year. According to The Wall Street Journal, quantitative hedge funds are now responsible for 27% of all U.S. stock trades by investors, up from 14% in 2013. Based on a recent Bernstein Research prediction, 50% of all assets under management in the U.S. will be passively managed by early 2018.

In these pages, we have time and again expressed concern about the potential distortions passive investing is creating. Today, evidence is everywhere in the U.S. economy — record low volatility despite a news cycle defined by turbulence; a stock market controlled by extreme top-heaviness; and many no-growth companies seeing ever-increasing valuation divergences. As always, the key questions are when will passive strategies backfire, what will prove the trigger, and how can we mitigate the damage to our portfolios? The better we understand the baked-in biases of algorithmic investing, the closer we can come to answers.

Over the last year, few have sounded the passive alarm as loudly as Steven Bregman, co-founder of investment advisor Horizon Kinetics. He believes record ETF inflows have generated “the greatest bubble ever” — “a massive systemic risk to which everyone who believes they are well-diversified in the conventional sense are now exposed.”

Bregman explained his rationale in a speech at a Grant’s conference in October:

“In the past two years, the most outstanding mutual fund and holding- company managers of the past couple of decades, each with different styles, with limited overlap in their portfolios, collectively and simultaneously underperformed the S&P 500…There is no precedent for this. It’s never happened before. It is important to understand why. Is it really because they invested poorly? In other words, were they the anomaly for underperforming — and is it reasonable to believe that they all lost their touch at the same time, they all got stupid together? Or was it the S&P 500 that was the anomaly for outperforming? One part of the answer we know… If active managers behave in a dysfunctional manner, it will eventually be reflected in underperformance relative to their benchmark, and they can be dismissed. If the passive investors behave dysfunctionally, by definition this cannot be reflected in underperformance, since the indices are the benchmark.”

At the heart of passive “dysfunction” are two key algorithmic biases: the marginalization of price discovery and the herd effect. Because shares are not bought individually, ETFs neglect company-by-company due diligence. This is not a problem when active managers can serve as a counterbalance. However, the more capital that floods into ETFs, the less power active managers possess to force algorithmic realignments. In fact, active managers are incentivized to join the herd—they underperform if they challenge ETF movements based on price discovery. This allows the herd to crowd assets and escalate their power without accountability to fundamentals.

With Exxon as his example, Bregman puts the crisis of price discovery in a real- world context:

“Aside from being 25% of the iShares U.S. Energy ETF, 22% of the Vanguard Energy ETF, and so forth, Exxon is simultaneously a Dividend Growth stock and a Deep Value stock. It is in the USA Quality Factor ETF and in the Weak Dollar U.S. Equity ETF. Get this: It’s both a Momentum Tilt stock and a Low Volatility stock. It sounds like a vaudeville act…Say in 2013, on a bench in a train station, you came upon a page torn from an ExxonMobil financial statement that a time traveler from 2016 had inadvertently left behind. There it is before you: detailed, factual knowledge of Exxon’s results three years into the future. You’d know everything except, like a morality fable, the stock price: oil prices down 50%, revenue down 46%, earnings down 75%, the dividend-payout ratio almost 3x earnings. If you shorted, you would have lost money…There is no factor in the algorithm for valuation. No analyst at the ETF organizer—or at the Pension Fund that might be investing—is concerned about it; it’s not in the job description. There is, really, no price discovery. And if there’s no price discovery, is there really a market?”

We see a similar dynamic at play with quants. Competitive advantage comes from finding data points and correlations that give an edge. However, incomplete or esoteric data can mislead algorithms. So the pool of valuable insights is self-limiting. Meaning, the more money quants manage, the more the same inputs and formulas are utilized, crowding certain assets. This dynamic is what caused the “quant meltdown” of 2007. Since, quants have become more sophisticated as they integrate machine learning, yet the risk of overusing algorithmic strategies remains.

Writing about the bubble-threat quants pose, Wolf Street’s Wolf Richter pinpoints the herd problem:

“It seems algos are programmed with a bias to buy. Individual stocks have risen to ludicrous levels that leave rational humans scratching their heads. But since everything always goes up, and even small dips are big buying opportunities for these algos, machine learning teaches algos precisely that, and it becomes a self-propagating machine, until something trips a limit somewhere.”

As Richter suggests, there’s a flip side to the self-propagating coin. If algorithms have a bias to buy, they can also have a bias to sell. As we explored in WILTW February 11, 2016, we are concerned about how passive strategies will react to a severe market shock. If a key sector failure, a geopolitical crisis, or even an unknown, “black box” bias pulls an algorithmic risk trigger, will the herd run all at once? With such a concentrated market, an increasing amount of assets in weak hands have the power to create a devastating “sell” cascade—a risk tech giant stocks demonstrated over the past week.

With leverage on the rise, the potential for a “sell” cascade appears particularly threatening. Quant algorithms are designed to read market tranquility as a buy-sign for risky assets—another bias of concern. Currently, this is pushing leverage higher. As reported by The Financial Times, Morgan Stanley calculates that equity exposure of risk parity funds is now at its highest level since its records began in 1999.

This risk is compounded by the ETF transparency-problem. Because assets are bundled, it may take dangerously long to identify a toxic asset. And once toxicity is identified, the average investor may not be able to differentiate between healthy and infected ETFs. (A similar problem exacerbated market volatility during the subprime mortgage crisis a decade ago.) As Noah Smith writes, this could create a liquidity crisis: “Liquidity in the ETF market might suddenly dry up, as everyone tries to figure out which ETFs have lots of junk and which ones don’t.”

J.P. Morgan estimated this week that passive and quantitative investors now account for 60% of equity assets, which compares to less than 30% a decade ago. Moreover, they estimate that only 10% of trading volumes now originate from fundamental discretionary traders. This unprecedented rate of change no doubt opens the door to unaccountability, miscalculation and in turn, unforeseen consequence. We will continue to track developments closely as we try and pinpoint tipping points and safe havens. As we’ve discussed time and again with algorithms, advancement and transparency are most-often opposing forces. If we don’t pry open the passive black box, we will miss the biases hidden within. And given the power passive strategies have rapidly accrued, perpetuating blind faith could prove devastating.

Screen Shot 2015-11-18 at 4.55.47 PM

There is a set of trading rules that only fails [true positive] 1/1000 times. However, when entering the trade, there is a 5% false positive rate in screening trades, which can take you out of the trade based on your risk management rules, thereby missing the return. You are only authorised by your partner/manager/firm to take a maximum of 10 trades each year. Trades are screened through this rule set randomly from all manner of securities, [stocks, futures, options]. A trade is screened and it is positive. What is the probability that the trade will fail? Will you take the trade?

Screen Shot 2015-11-18 at 4.55.47 PM

Suppose that over time traders have experimented with trading rules, drawn from a very wide universe of trading rules, perhaps tens of thousands of different iterations. As time progresses the rules that happen to perform well historically, attract more attention and are considered serious rules by the trading community, while unsuccessful rules gradually fall by the wayside.

If enough trading rules are considered over time, some rules, by pure luck, even in a large sample, will produce superior performance, even if they do not genuinely possess predictive power.

Screen Shot 2017-05-24 at 1.52.41 PM


Screen Shot 2015-11-18 at 4.55.47 PM

It was this [partial] post that got me thinking, as there are a number [lots?] of systems traders out there:

I don’t think I curve fit and use price action patterns as a basis for all my code , the majority of my indicators are custom written by myself ( why I will not divulge or discuss ) and represent tangible price action events . Median reversion is the basis of my systems and I filter what I see as important aspects of price action . The counts used in charts here are only a part of the process but they do measure trend and reversals and are integral part of what I do on every time frame .

So my question is: is every bar created equal?

The answer [to my mind] is clearly no.

Some bars contain nothing but trading noise, whereas other bars contain important information.

Will the market [the sum of all participants of a given bar] treat the same information in the same way as they did previously?

Again, my answer would be clearly no. The participants could, and likely would be, completely different, with the commensurate difference in subjective views. That is just one example, there are hundreds of reasons why the reaction could be different.

Therefore, building a trading system/methodology, based on historical data is prone to randomness, exactly what you are trying to eliminate.

Is this a futile undertaking?

Screen Shot 2015-11-18 at 4.55.47 PM

RICHARD CRAIB IS a 29-year-old South African who runs a hedge fund in San Francisco. Or rather, he doesn’t run it. He leaves that to an artificially intelligent system built by several thousand data scientists whose names he doesn’t know.

Under the banner of a startup called Numerai, Craib and his team have built technology that masks the fund’s trading data before sharing it with a vast community of anonymous data scientists. Using a method similar to homomorphic encryption, this tech works to ensure that the scientists can’t see the details of the company’s proprietary trades, but also organizes the data so that these scientists can build machine learning models that analyze it and, in theory, learn better ways of trading financial securities.

“We give away all our data,” says Craib, who studied mathematics at Cornell University in New York before going to work for an asset management firm in South Africa. “But we convert it into this abstract form where people can build machine learning models for the data without really knowing what they’re doing.”

He doesn’t know these data scientists because he recruits them online and pays them for their trouble in a digital currency that can preserve anonymity. “Anyone can submit predictions back to us,” he says. “If they work, we pay them in bitcoin.”

The company comes across as a Silicon Valley gag. All that’s missing is the virtual reality.

So, to sum up: They aren’t privy to his data. He isn’t privy to them. And because they work from encrypted data, they can’t use their machine learning models on other data—and neither can he. But Craib believes the blind can lead the blind to a better hedge fund.

Numerai’s fund has been trading stocks for a year. Though he declines to say just how successful it has been, due to government regulations around the release of such information, he does say it’s making money. And an increasingly large number of big-name investors have pumped money into the company, including the founder of Renaissances Technologies, an enormously successful “quant” hedge fund driven by data analysis. Craib and company have just completed their first round of venture funding, led by the New York venture capital firm Union Square Ventures. Union Square has invested $3 million in the round, with an additional $3 million coming from others.

Hedge funds have been exploring the use of machine learning algorithms for a while now, including established Wall Street names like Renaissance and Bridgewater Associates as well as tech startups like Sentient Technologies and Aidyia. But Craib’s venture represents new efforts to crowdsource the creation of these algorithms. Others are working on similar projects, including Two Sigma, a second data-centric New York hedge fund. But Numerai is attempting something far more extreme.

On the Edge

The company comes across as some sort of Silicon Valley gag: a tiny startup that seeks to reinvent the financial industry through artificial intelligence, encryption, crowdsourcing, and bitcoin. All that’s missing is the virtual reality. And to be sure, it’s still very early for Numerai. Even one of its investors, Union Square partner Andy Weissman, calls it an “experiment.”

But others are working on similar technology that can help build machine learning models more generally from encrypted data, including researchers at Microsoft. This can help companies like Microsoft better protect all the personal information they gather from customers. Oren Etzioni, the CEO of the Allen Institute for AI, says the approach could be particularly useful for Apple, which is pushing into machine learning while taking a hardline stance on data privacy. But such tech can also lead to the kind of AI crowdsourcing that Craib espouses

Craib dreamed up the idea while working for that financial firm in South Africa. He declines to name the firm, but says it runs an asset management fund spanning $15 billion in assets. He helped build machine learning algorithms that could help run this fund, but these weren’t all that complex. At one point, he wanted to share the company’s data with a friend who was doing more advanced machine learning work with neural networks, and the company forbade him. But its stance gave him an idea. “That’s when I started looking into these new ways of encrypting data—looking for a way of sharing the data with him without him being able to steal it and start his own hedge fund,” he says.

The result was Numerai. Craib put a million dollars of his own money in the fund, and in April, the company announced $1.5 million in funding from Howard Morgan, one of the founders of Renaissance Technologies. Morgan has invested again in the Series A round alongside Union Square and First Round Capital.

It’s an unorthodox play, to be sure. This is obvious just when you visit the company’s website, where Craib describes the company’s mission in a short video. He’s dressed in black-rimmed glasses and a silver racer jacket, and the video cuts him into a visual landscape reminiscent of The Matrix. “When we saw those videos, we thought: ‘this guy thinks differently,’” says Weissman.

As Weissman admits, the question is whether the scheme will work. The trouble with homomorphic encryption is that it can significantly slow down data analysis tasks. “Homomorphic encryption requires a tremendous about of computation time,” says Ameesh Divatia, the CEO of Baffle, a company that building encryption similar to what Craib describes. “How do you get it to run inside a business decision window?” Craib says that Numerai has solved the speed problem with its particular form of encryption, but Divatia warns that this may come at the expense of data privacy.

According to Raphael Bost, a visiting scientist at MIT’s Computer Science and Artificial Intelligence Laboratory who has explored the use of machine learning with encrypted data, Numerai is likely using a method similar to the one described by Microsoft, where the data is encrypted but not in a completely secure way. “You have to be very careful with side-channels on the algorithm that you are running,” he says of anyone who uses this method.

Turning Off the Sound at a Party

In any event, Numerai is ramping up its effort. Three months ago, about 4,500 data scientists had built about 250,000 machine learning models that drove about 7 billion predictions for the fund. Now, about 7,500 data scientists are involved, building a total of 500,000 models that drive about 28 billion predictions. As with the crowdsourced data science marketplace Kaggle, these data scientists compete to build the best models, and they can earn money in the process. For Numerai, part of the trick is that this is done at high volume. Through a statistics and machine learning technique called stacking or ensembling, Numerai can combine the best of myriad algorithms to create a more powerful whole.

Though most of these data scientists are anonymous, a small handful are not, including Phillip Culliton of Buffalo, New York, who also works for a data analysis company called Multimodel Research, which has a grant from the National Science Foundation. He has spent many years competing in data science competitions on Kaggle and sees Numerai as a more attractive option. “Kaggle is lovely and I enjoy competing, but only the top few competitors get paid, and only in some competitions,” he says. “The distribution of funds at Numerai among the top 100 or so competitors, in fairly large amounts at the top of the leaderboard, is quite nice.”

Each week, one hundred scientists earn bitcoin, with the company paying out over $150,000 in the digital currency so far. If the fund reaches a billion dollars under management, Craib says, it would pay out over $1 million each month to its data scientists.

Culliton says it’s more difficult to work with the encrypted data and draw his own conclusions from it, and another Numerai regular, Jim Fleming, who helps run a data science consultancy called the Fomoro Group, says much the same thing. But this isn’t necessarily a problem. After all, machine learning is more about the machine drawing the conclusions.

In many cases, even when working with unencrypted data, Culliton doesn’t know what it actually represents, but he can still use it to build machine learning models. “Encrypted data is like turning off the sound at the party,” Culliton says. “You’re no longer listening in on people’s private conversations, but you can still get very good signal on how close they feel to one other.”

If this works across Numerai’s larger community of data scientists, as Richard Craib hopes it will, Wall Street will be listening more closely, too.


Screen Shot 2015-11-18 at 4.55.47 PM

Shares of Valeant Pharmaceuticals International Inc. VRX, +12.33% shot up 9.7% in active premarket trade Wednesday, after Morgan Stanley analyst David Risinger turned bullish on the drug maker, citing the belief that major risks to the company have already been priced into the stock. Risinger raised his rating to overweight, after being at in line since October 2015. He raised his stock price target to $42, which is 58% above Tuesday’s closing price of $26.60, from $27. “Risk of severe financial stress should diminish as [debt] covenants are renegotiated and [Valeant] pays down debt, and deleveraging should drive equity value accretion,” Risinger wrote in a note to clients. Regarding risks of drug pricing resets, Risinger said Valeant has already experienced step downs in net pricing and access, and he his valuation estimates already account for generic competition for the company’s most controversial drugs–Isuprel and Nitropress–over the next six to 12 months. The stock, which was on course to open at a 2 1/2-month high, had plunged 74% year to date through Tuesday, while the SPDR Health Care ETF XLV, +0.08% had tacked on 3.1% and the S&P 500 SPX, +0.03%had gained 6.6%.

The increased volatility, sharp drop, sharp rise, are generating some nice profits in this stock.

Screen Shot 2015-11-18 at 4.55.47 PM

Seems to have turned bearish. The last few days have been red days in the markets and that seems to have turned much of the commentary very negative again.

Top of the list again seems to be oil, China, negative interest rates and poor guidance forward, European & Chinese banks with high default rates leading contenders for the why.

The question is still however: with so much surplus cash looking to earn a return, where does it go?

Likely, still into financial markets.

Any sign of a bottoming process?

I would say too early still. Longer term I am still a bull, after all in a fiat money environment, which won’t change until possibly it has to, stocks, as productive assets holding productive capital, will appreciate, if only to protect against inflation.

It is the volatility, or market fluctuations that push individuals out of markets, or encourage hyperactive trading.

Some form of mechanical system is the way to go. It keeps you engaged in the markets and keeps you in the market. The system must [obviously] have a bear component to it, other wise it suffers from the same flaw.

While it would be nice to be able to trade the absolute bottoms and tops, pretty unrealistic. The system has to be robust enough to ride out downturns and the volatility that is thereby engendered.


Screen Shot 2016-05-06 at 6.08.07 AM

Next Page »