MCAT CARS Skills—Machine Learning in the Finance Sector

Session 19

Computers and machines are heavily used and studied in society. While although they may increase ease, are they really always better?

With me is Jack Westin, the premier MCAT CARS tutor online. If you are struggling with the CARS section of the MCAT, or any section of the MCAT, the MCAT CARS Course that Jack Westin offers will get you to where you need to go.

Link to article:

https://www.bloomberg.com/news/articles/2018-10-09/the-big-problem-with-machine-learning-algorithms

Machine learning is enabling investors to tap huge data sets such as social media postings in ways that no mere human could. Yet, despite the enormous potential, its record remains mixed. The Eurekahedge AI Hedge Fund Index, which tracks the returns of 13 hedge funds that use machine learning, has gained only 7 percent a year for the past five years, while the S&P 500 returned 13 percent annually. This year the Eurekahedge benchmark dropped 5 percent through September.

One of the potential pitfalls for machine learning strategies is the extremely low signal-to-noise ratio in financial markets, says Marcos López de Prado, who joined AQR Capital Management as head of machine learning in September and is the author of the 2018 book Advances in Financial Machine Learning. “Machine learning algorithms will always identify a pattern, even if there is none,” he says. In other words, the algorithms can view flukes as patterns and hence are likely to identify false strategies. “It takes a deep knowledge of the markets to apply machine learning successfully to financial series,” López de Prado says.

Nigol Koulajian echoes that view. The founder and chief investment officer at Quest Partners, a New York-based systematic macro hedge fund that manages $1.7 billion, says that quants coming out of finance programs and high-tech companies often expect to create optimizations at a much higher level of precision than is warranted in finance. “They’re coming with a mindset that we’re going to conquer the world with big data,” Koulajian says. In finance, though, the market regime is not static, and markets aren’t closed systems like a chess game. “You can have one little pin drop that can basically make you lose over 20 years of returns,” he says.

Consider one risk-on strategy that had worked well in the decade since the financial crisis: bottom-fishing equity indexes. “Everyone’s buying the dips,” Koulajian says. “There are all these people who have learned to basically suppress the vol,” or volatility. If you use machine learning, you can implement dozens of versions of this strategy. The risk, according to Koulajian, is that the relentless bull market that made such strategies work so well was driven by central bank liquidity—and that’s being pulled away now. Meanwhile, skew, a measure of tail risk, is implying the S&P 500 could fall 30 percent, Koulajian says: In August, the CBOE Skew Index, which tracks out-of-the-money index options, reached a record high. If you’re buying the dips with machine learning, Koulajian says, it’s easy to congratulate yourself on using much more complex model optimization and lose sight of the larger risks.

Machine learning isn’t really new, says Robert Frey. Frey in the late 1980s started a hedge fund that was absorbed a few years later into Renaissance Technologies, where it became the nucleus of the statistical arbitrage strategy in the enormously successful Medallion Fund. “You hear all this stuff about machine learning and AI,” Frey says. “Most of those techniques, however, have been around for decades—and we, in fact, used a lot of them at Renaissance,” he says. “The fundamental processes that we’re talking about here are a combination of advanced statistics—computationally intensive statistical analysis—and then the neural-network-type branch where you’re looking at these models, which are basically classifiers.”

[02:54] Paragraph 1, Sentence 1

Machine learning is enabling investors to tap huge data sets such as social media postings in ways that no mere human could.

Jack says:

“Investors” is a keyword they mentioned here, meaning people who are trying to make money.

[03:38] Paragraph 1, Sentence 2

Yet, despite the enormous potential, its record remains mixed.

Jack says:

They word “yet” here is a contrast word. So machine learning is probably not helpful to investors.

[04:25] Paragraph 1, Sentence 3

The Eurekahedge AI Hedge Fund Index, which tracks the returns of 13 hedge funds that use machine learning, has gained only 7 percent a year for the past five years, while the S&P 500 returned 13 percent annually.

Jack says:

You don’t need to know the normal annual percentage return here but they’re telling you there’s some kind of difference. You’re expected to be sharp enough to realize that this is significantly bad. Compared to just regular investing. So now you can go to your friends and say, “Hey you know what? It only returns 7%. I should get 13%” or whatever.

[05:25] Paragraph 1, Sentence 4

This year the Eurekahedge benchmark dropped 5 percent through September.

Jack says:

We don’t have to know the exact percentages that make it bad, but you should know that this is an example of how maybe using machines isn’t so helpful. Machine learning isn’t so helpful after all.

[05:54] Paragraph 2, Sentence 1

One of the potential pitfalls for machine learning strategies is the extremely low signal-to-noise ratio in financial markets, says Marcos López de Prado, who joined AQR Capital Management as head of machine learning in September and is the author of the 2018 book Advances in Financial Machine Learning.

Jack says:

You don’t have to know anything about low signal-to-noise. Be confident at this point. Don’t give up on yourself. Look at MCAT like a puzzle. It’s like you’re going in a maze, and you’re going in one direction, and all of a sudden you see a wall. You can either stop at that wall, look at that wall for eternity, or you can turn your body, go the other direction, and maybe you’ll learn something as you keep going that new direction. It’s a new path and it can be scary, but that’s life. You have to take new paths, take new directions.

[07:41] Paragraph 2, Sentence 2

“Machine learning algorithms will always identify a pattern, even if there is none,” he says.

Jack says:

So this has something to do with looking at patterns. Something to do with signal-to-noise probably is looking for patterns and understanding things. Signals meaning patterns.

[08:18] Paragraph 2, Sentence 3

In other words, the algorithms can view flukes as patterns and hence are likely to identify false strategies.

Jack says:

These strategies, if there isn’t a pattern, the algorithms may see something and identify a strategy, even though there isn’t anything there.

[08:40] Paragraph 2, Sentence 4

“It takes a deep knowledge of the markets to apply machine learning successfully to financial series,” López de Prado says.

Jack says:

This is another dilemma many students have is how deep do I need to understand this stuff? NO, try to look at MCAT CARS as a camera, like an iPhone or Android camera. You can zoom in and zoom out with your camera. If you stay too zoomed in, you’re going to miss the bigger picture of this paragraph.

This last sentence doesn’t make a big difference here. It’s important to know that you need to apply deep knowledge to markets to be successful, but that’s beyond the point. The point of this paragraph is really about how machines are not necessarily great at finding patterns. Or if they do find patterns, there really wasn’t one to begin with. It’s causing issues that way.

So this has something to do with low signal-to-noise, understanding that machines are messing up with pattern recognition. That’s what really matters. You have to zoom out, you have to go back to get that, and if you’re too zoomed in on every little sentence you’re reading, you’re going to miss that big picture. The big picture here is this idea of how machine learning finds patterns when there really aren’t any, and that’s what Marcos Lopez de Prado believes.

[10:33] Paragraph 3, Sentences 1-2

Nigol Koulajian echoes that view. The founder and chief investment officer at Quest Partners, a New York-based systematic macro hedge fund that manages $1.7 billion, says that quants coming out of finance programs and high-tech companies often expect to create optimizations at a much higher level of precision than is warranted in finance.

Jack says:

This is another name here that seems to be agreeing with Marcos. It brings up a name, and you probably have to know what this person is saying, but what he’s saying is so tough to understand unless you’re really into this stuff.

Quants coming out of finance programs, and high tech companies often expect to create optimizations at a much higher level of precision than is warranted in finance. So again, something to do with precision. This could pertain to accuracy, which is kind of what we were talking about in the previous paragraph.

What you’ve got to know is not exactly what Nigol says, but knowing that Nigol echoes that view, which means that Nigol agrees with Marcos. So regardless of what the details are, we know that Nigol and Marcos agree that maybe machine learning has some problems. They identify things that aren’t there.

[12:20] Paragraph 3, Sentence 3

“They’re coming with a mindset that we’re going to conquer the world with big data,” Koulajian says.

Jack says:

The word “they’re” here refers to the public at large – most people. Whenever something goes wrong, we always want to kind of blame the public. Why does life have to be so hard?’ So ‘hard’ relative to what? Relative to other people, other systems, the system that we live in. “They are” is referring to the system. The people that run the system. So in this case, “they are” is referring to how what people mostly believe.

[13:10] Paragraph 3, Sentence 4

In finance, though, the market regime is not static, and markets aren’t closed systems like a chess game.

Jack says:

So it’s saying that it’s not a closed system so maybe it’s an open system.

[13:54] Paragraph 3, Sentence 5

“You can have one little pin drop that can basically make you lose over 20 years of returns,” he says.

Jack says:

This is going back with the idea of “not accurate,” but there are issues with it. It also brings in this idea that what if some element that machine learning hasn’t calculated occurs? Then the machine learning won’t know exactly what to do. What if some event happens? Can it erase twenty years of returns – twenty years of work or investment? But again, those are details not as important as knowing the big picture that Koulajian agrees with Marcos that there are some problems with machine learning.

[14:53] Paragraph 4, Sentence 1

Consider one risk-on strategy that had worked well in the decade since the financial crisis: bottom-fishing equity indexes.

Jack says:

We don’t need to know what bottom fishing equity indexes are here. You don’t have to know these words to be confident in understanding what’s going on. So as you read this, it’s just saying that something is working well since the financial crisis and that would be bottom fishing equity indexes. We don’t know what that is, but we know it’s doing well.

[16:00] Paragraph 4, Sentence 2

“Everyone’s buying the dips,” Koulajian says.

Jack says:

So maybe the strategy is buying the dips. Dips could mean bottom, but that we don’t know what the bottom is. Maybe it’s buying a stock at a very low price. Again, you don’t have to know what that is. You’ve been able to know that there is a strategy we’re talking about, that strategy includes buying the dips, whatever that is.

[16:34] Paragraph 4, Sentence 3

“There are all these people who have learned to basically suppress the vol,” or volatility.

Jack says:

That’s the strategy. It’s been working for them, but again, we don’t really know what the volatility is, how they suppress it, and that “not knowing” does bother students.

But we don’t know what is suppressing and what is volatility. And you have to be comfortable being uncomfortable. You have to be comfortable not knowing things. It’s kind of the opposite of a university class. Usually students go into a university class and they try to learn everything from every angle. For that week, they try to read the book on the topic, they look at their notes, they ask the professor. But here, you don’t have that time. You don’t have that luxury on this CARS passage, so you have to go with whatever they give you and have faith that it’s enough to answer the questions

[17:37] Paragraph 4, Sentence 4-5

If you use machine learning, you can implement dozens of versions of this strategy. The risk, according to Koulajian, is that the relentless bull market that made such strategies work so well was driven by central bank liquidity—and that’s being pulled away now.

Jack says:

This is going back to what Koulajian was saying, that unknown factors – things that the machine doesn’t account for, open system environments – can affect things dramatically. And it seems like it’s affecting the bottom fishing equity strategy, which may work well. It might work well for a while with machines, but then you have this issue where the central bank decides to be pulled away, and that messes things up.

Again, this is that zooming in. If you zoom in too much, you’re going to understand this stuff and that’s great, but they’re not going to ask questions about the central bank. They’re not going to ask questions about bottom fishing equity. They’re going to ask questions about how Koulajian believes that the system, this machine learning, does not work well – the big picture. So don’t be too zoomed in because if you’re too zoomed in, you’re going to lose track of that big picture.

[20:19] Paragraph 4, Sentence 6

Meanwhile, skew, a measure of tail risk, is implying the S&P 500 could fall 30 percent, Koulajian says: In August, the CBOE Skew Index, which tracks out-of-the-money index options, reached a record high.

Jack says:

No idea what any of these words mean other than it’s saying this CBOE reached a high in August.

[20:09] Paragraph 4, Sentence 7

If you’re buying the dips with machine learning, Koulajian says, it’s easy to congratulate yourself on using much more complex model optimization and lose sight of the larger risks.

Jack says:

Now, you could get frustrated reading at this point here as what students probable would be too. It’s painful to read this and students blame themselves. This is the difference between a high scoring student and someone who typically just kind of gives up halfway. How willing are you to persevere? How willing are you to just keep going when times are tough? That’s what they’re trying to do. The MCAT is an intimidation test. They want to intimidate you. They want to see if you really want this bad enough. Because if you really want this, you’re willing to go through this hell. You’re willing to read this boring passage about something you could care less about, and that’s what makes it so difficult for many students. They don’t find it relatable or it’s not something they like to read, something very convoluted. But in actuality, you can still get the big picture, and that’s all you really need.

You’re distracting yourself with things that you don’t need to tell yourself. And this happens for many students, and it’s kind of like going into a tournament, and while you’re competing, whatever that tournament is, or if you’re part of an event of some musical group, all of a sudden, you start doubting yourself, thinking it’s not going to go well. So you’ve got to believe in yourself at least until the end of this passage – at the very least.

[22:43] Paragraph 5, Sentence 1

Machine learning isn’t really new, says Robert Frey.

Jack says:

Another name is mentioned here and he’s saying that it’s not something new.

[22:53] Paragraph 5, Sentence 2

Frey in the late 1980s started a hedge fund that was absorbed a few years later into Renaissance Technologies, where it became the nucleus of the statistical arbitrage strategy in the enormously successful Medallion Fund.

Jack says:

Takeaway here is Frey had a hedge fund that was absorbed into this other company, Renaissance Technologies, and that became a central to a successful medallion fund.

[23:26] Paragraph 5, Sentence 3-4

“You hear all this stuff about machine learning and AI,” Frey says. “Most of those techniques, however, have been around for decades—and we, in fact, used a lot of them at Renaissance,” he says.

Jack says:

So Frey here is saying that this is not something new, that it’s not cool, and that we’ve been doing this forever.

[23:52] Paragraph 5, Sentence 5

“The fundamental processes that we’re talking about here are a combination of advanced statistics—computationally intensive statistical analysis—and then the neural-network-type branch where you’re looking at these models, which are basically classifiers.”

Jack says:

So the author is saying here that it’s just a bunch of math and that it’s just statistics – a neural network type branch looking at models.

[24:25] The Big Picture

The big picture here is that this machine learning and everything that we’ve been learning about for the previous paragraphs isn’t new. It’s been around for a while. Machine learning can have problems even though it’s been around. There are issues with it that we don’t really notice, and that we have to be cautious with machine learning in general.

This is a great example of a passage that you might not know what’s going on, but you could still understand the big points, and that’s what MCAT is all about.

So if you’re struggling with these kinds of passages, preparation is key. Prepare yourself. Practice these kinds of passages, practice what’s challenging for you.

And then another key is that everyone’s going to complain at some degree when they read something like this, but don’t let that fester. Don’t let that build up to the point where you burn out, to the point where you just give up halfway through the passage.

Instead, go balance yourself, and go exercise, and eat well, and get ready to attack this one more day – one more day, one more week, until you get it down. Besides, not everyone would be perfect at all passages they read. That’s a myth. Not everyone is perfect at reading everything they read. But if you’re open-minded and you’re willing to embrace whatever the passage is saying, you have a better chance of succeeding.