Blog
Podcast
-
May 29, 2024

Exploring Quantitative Investment Strategies: From Smart Beta to AI Integration

Stefan (00:00:45) - I'm here with Sean Adler, CEO of GSI and founder of GSI empowers finance investors with alternative financial data engineering. I particularly would like to talk to you a little bit today about quantitative investment strategies and other related topics. but I think it might be worthwhile to start initially to understand if you key terms that we're probably going to be using in our conversation. So is it possible maybe for you to define and distinguish between smart beta factor investing and what I already mentioned, quantitative investment strategies?

Sean (00:01:23) - Okay.

Sean (00:01:24) - So smart beta you know obviously refers to portfolio volatility in terms of beta weighing and variance. and so there's dynamically updating that factor investing. You know you have micro macro. This is more on the econ and business and fundamental analysis and so on and so forth. and then on quantitative finance, which is probably more where we specialize, there's no shortage of things. And so there's everything from analysis of econometric time series to, you know, jump diffusion or, you know, various algo trading strategies, ie so on and so forth. And they're all sort of enmeshed at this point in time.

Stefan (00:02:10) - You mention it, a lot of words, big words for some people. is this sort of a common myth about that kind of investing in our industry that you believe needs to be debunking?

Sean (00:02:25) - Yeah. So I firmly believe that there is no one true way when it comes to investments. And so this is coming from, you know, somebody on the quantitative and AI side. And so what you hear from a lot of people who are experienced quants is not to wholeheartedly trust your back tests.

Sean (00:02:42) - And everybody who who's not on the technical end is like, yeah, of course. But when you deal with people who, you know, are strongly based in computer programming, like they they trust them wholeheartedly, and that creates a huge number of issues. So I think that's one of the biggest myths. And there are plenty of successful managers who, you know, don't do anything with quant finance. And then of course, there are legends like Jim Simons who are straight on finance, and there's a good balance in between where everything is enmeshed. And I think that that's probably the biggest myth out there, and that's should be dispelled.

Stefan (00:03:18) - Yeah. And I think I always joke, you know, show me a big test that doesn't look good. Basically. You know, it's there's a. A negative selection process. Basically I would say, you know, you mentioned investing and there's no right way pure and only one way to invest right now. There's also, I think, different perceptions of risk. And what is your risk perception and how do you define it.

Sean (00:03:50) - so I honestly probably fall more into the heavily balanced, high risk category. And so, I'm a huge fan of derivatives baskets and, you know, more precise strategies. And also, you know, like a dabbled in, you know, the venture side. And of course, everybody in, in private investment banking is interested in these obscenely high risk deals with, you know, large payouts.

Stefan (00:04:15) - But these are sort of more asymmetric payouts in a sense.

Sean (00:04:19) - Yes. So, so like anything in the private side is much more asymmetric. you know, when you start dealing with derivatives trading, it's much easier to, to model it dynamically and come up with symmetric returns. I think that, definitely more in, in the balanced high risk category. But the thing is, is that when the strategies and the models are right, a lot of the high risk is, is mitigated. And so it becomes more of a balanced risk. And when you, you know, start doing, you know, applications of smart data to, you know, basket and rainbow options and everything else, you factor them into a model that's based on equilibrium.

Sean (00:04:58) - You know, a lot of that gets mitigated. And it's about scaling things onto themselves.

Stefan (00:05:04) - And how do you measure risk in your investments when you look at it?

Sean (00:05:09) - And so a lot a lot of data weighing obviously everybody checks you know profit and loss beta delta dollars. You know various volatility ratios for options. I use a lot of more standard metrics for basic risk gauging on and more non-standard things for, you know, actual modeling.

Stefan (00:05:32) - I mean, some people look at, you know, loss of capital, Sharpe ratios, you know, volatility by itself.

Sean (00:05:42) - I mean, yeah. So Sharpe ratios are pretty much everybody's go to you know, so with volatility ratios a lot of that can be gauged from various portfolio betas. and whether it's, you know, one relating to the instrument or the overall portfolio or, you know, an individual basket of derivatives, I think that that's important, obviously, like if if a model is bleeding money, like that's a giant red flag. But before that happens, you know, a lot of metrics usually start going haywire.

Stefan (00:06:15) - Understood. Now, you mentioned AI earlier. how is this sort of being integrated now in investment?

Sean (00:06:25) - So AI is is everywhere. And I think that, there are a lot of ways that's being applied in finance. And so there are things in natural language processing where people will, you know, essentially screen pools of SEC filings or they'll go over social media data on Twitter to see if, you know, the next GameStop jump happens and so on and so forth. And so there's the language processing side. on the quantitative side, there is a lot of, you know, modeling in terms of, you know, how you would select securities or how you would place options and when. and so I think that that's, that's more heavily rooted in modeling and time series, you know, things like Kalman filters. So that's a different aspect. And I think somewhere in between there is alternative data. And so that's anything from measuring traffic from satellites to, you know, the number of cars in a target parking lot, to, you know, like R&D reports from biotech companies.

Sean (00:07:29) - And so it's a little more new. I think it's also a little more iffy. but the number of funds are using it.

Stefan (00:07:38) - Cynics could say, you know, for example, I what you see sort of in GGP right now is like it's just a pattern recognition tool. If if you feed it with enough data, it will recognize the pattern.

Sean (00:07:50) - Yeah. And so that's also why I becomes dangerous in finance. Is that everybody everybody who's, you know, spent enough time trading knows that things happen like period. You know, something unexpected happens. There's always some random jump. And the models don't always account for that. Even if you've got something that's, you know, screening every single news update and is sort of gauging risk sentiment from that, there's always an X factor in the financial markets. And so that's why I think the, you know, the balanced approach is important because strictly relying on AI for trading is is somewhat dangerous. But when you start dealing with more of the traditional aspects, you know, like, you know, key financial ratios and so on and so forth, when you understand what you're trading, you can mitigate any kind of risk from, you know, overfitting models or just, you know, computerized prediction wise.

Stefan (00:08:52) - I mean, you mentioned already overfitting models. The other challenges I have found often with that one is sort of how good is the data quality or and how much quantity you have, or even you might not even have enough and, or some data is missing. And that poses a certain risk of making the wrong conclusions.

Sean (00:09:11) - Yeah. And I think, so a lot of like the financial data is readily available. And so, there are different ways to sort of mix things in. And so if you are a developer, you can essentially build, you know, AI models from scratch by just running an equation and then feeding the outputs into a model. that's probably overkill for most people who, you know, aren't in development, but there's everything from web scraping to, you know, additional API feeds. And there's really no shortage of data out there that can be used.

Stefan (00:09:47) - But how do you make sure that that data is actually meaningful or for better word correct?

Sean (00:09:55) - Test it and get results. And so my, my honest opinion about about any kind of data is that.

Sean (00:10:04) - It's as good as the results. And so I think everybody likes to talk about how phenomenal the data is. And I think that if you can demonstrate reliance from the data, that's how you really gauge the results. Because at the end of the day, that's what people are really about in this industry.

Stefan (00:10:23) - And reliance. How would you measure that one? Call it the hit ratio or your return side the return.

Sean (00:10:33) - I mean, like, you know, you could do everything from hit ratios to, you know, things from standard ROC curves like, you know, sensitivity or specificity, etc., etc.. you know, if you you want to get more into the esoteric side of AI, you know, you could look at entropy and everything else. But I think that a safe bet is is generally hit ratios, profit and loss and, and some of the, the basic statistics and like sensitivity and specificity because it gets really easy to overanalyze some of the things. And you honestly just want to simplify things and see what works.

Stefan (00:11:11) - My field was always a little bit with AI. I mean, if AI is just rational for a better world, it will at one point find a way of manipulate how it can manipulate the market. Like spoof the market fake is certain liquidity. How do you prevent AI from doing that?

Sean (00:11:31) - So at this point I is is almost just as good as the, you know, people who create the models. And so the accuracy and the design of the model is really going to be determined by the people building it at this stage. You know, obviously you can start getting into these, these theories where eventually, you know, there'll be like the Terminator and, you know, like it will just operate on its own. And yeah, in, you know, 20 years maybe. But I think that for actual market manipulation to happen, people have to sort of allow it within the models. And there are some things where, you know, like ChatGPT has, you know, engaged in insider trading, but, you know, to the extent at which that affects the market is largely dependent on, you know, like frequency and volume and everything else.

Sean (00:12:21) - And so there's definitely potential for market manipulation. But a lot of the regulations within trading will just remain intact regardless of what the AI does.

Stefan (00:12:35) - Now, when you go about in designing, say, an investment strategy using AI or something, what are sort of the most significant challenges that you face when designing and implementing it?

Sean (00:12:50) - I think probably updating at this point. It reaches a point where people become very comfortable with, you know, like the design, the efficacy of a strategy. And then it becomes about maintaining that as you continuously update and you add new things to it. And I think that the continuous optimization side is probably the most challenging.

Stefan (00:13:14) - And to know when to do it. Or is it just a constant? You just you do it even before something happens. Just to stay on top of it.

Sean (00:13:24) - So usually I think before any trades get placed, like there has to be significant analysis beforehand. But I think that everybody in this industry is constantly chasing, you know, like more accurate or better returns.

Sean (00:13:37) - And that's kind of just the nature of the beast and how everything gets sized up. And so the process of continuous optimization, in order to, you know, sort of survive and compete in the industry is somewhat of a necessity. But, it's the balance between continuing to scale and pushing the risk too high that everybody is sort of walking at this point.

Stefan (00:14:01) - For some people who already have a certain, let's say, certain reputation in the business, they can probably raise money for strategies like this. But the challenge I often I've seen in a great idea, it looks good. It might even have a real track record. Try it out in a managed account. Everything else. But how crucial is it if you want to raise and go up above money? The transparency in trying to be able to explain what the strategy actually does to your clients. And so, you know, how do you have to balance complexity versus comprehensibility?

Sean (00:14:37) - Decrease complexity when speaking to investors and in due diligence and focus on comprehensibility. At the end of the day, like what people really want in terms of initial meetings is just an overview of things.

Sean (00:14:52) - And so like once you know, you're actually in due diligence and you know, everything from background checks to, you know, trading logs, you know, like a company returns everything else that's in question. At that point, you're dealing much more with the. The complexity of things. but it it takes a lot of, you know, comprehension before people are really interested in getting to that end, due to the degree of detail people have to kind of go over before any of these transactions, you know, happen or even clear.

Stefan (00:15:28) - And I mean, that was a challenge I always found or still find when it comes, for example, with derivative strategies. And where do investors put this in their portfolio? You know, often they just stick it into what they call the alternative budget and ignore any effect it might actually have on the overall portfolio. Is it risk reducing, risk increasing, giving more exposure to equity or not? Is there a way how to how you can approach this with strategies like this.

Stefan (00:15:55) - Because they they can quickly change their exposure for a better world.

Sean (00:16:00) - Yeah, I think that if you're managing your own capital, you know, I think that you could probably do something that's, you know, maybe up to 20% derivatives. I think if you do that when you're managing other people's capital, I think a number of people would probably be somewhat uneasy about that. And so, you know, all these funds talk about like, you know, maybe put like 1% or like 0.5% and something like cryptocurrency futures. And so, you know, you can balance it that way. I honestly think that in terms of incorporating derivatives, a lot of that depends on the overall composition of the portfolio, because everything is relative to itself. And so you can you can sort of dynamically adjust the ratios. But in general, like if you're dealing with other people's money, I wouldn't push it above. Above 10% unless everybody's comfortable with it.

Stefan (00:16:57) - And when it comes to these let's say AI driven strategies. Same thing a little bit applies there as well that they.

Stefan (00:17:05) - Where do you put them in your portfolio or the exposure to them.

Sean (00:17:09) - And so I think there's, there's a little bit of AI in everything. You need to be careful. You can you can use the AI for econometric time series. You can use it to screen financial ratios or, you know, coupon rates on bonds and everything else. And I think it becomes useful in that, you know, regard. But AI is is never perfect. If the AI was truly perfect on the financial markets, you know, everybody on Wall Street in Silicon Valley would just essentially be sitting back and kicking their feet up.

Stefan (00:17:38) - Correct? Yes, I agree. On the other hand, you could say it's just an arms race, basically who can who's faster and has a new idea of how looking at something else or analyzing potential investments in a different way.

Sean (00:17:52) - I think that's fair. And so I mean, like in, in the corporate world and, you know, hedge fund startups, like there's always sort of this, this arms race.

Sean (00:18:01) - And I think that the biggest issue with the arms race is that the people who actually win the race are usually something slightly different. And so you've got people like Jim Simons and David Shore and like, you know, Jim Simons had a PhD and, you know, like mathematical geometry. David Shor was a physicist who, you know, ran a quant fund and does research and computational bioinformatics. And so the people who sort of who win the race are not usually, you know, just a run of the mill MBA, although I think that's more of what the the corporate world attracts.

Stefan (00:18:40) - If somebody you know would like to contact you and engage with you, interested parties is their is their best way to contact you.

Sean (00:18:49) - The absolutely emails probably easiest on LinkedIn is is helpful if you send a message a corporate emails. Just Sean Adler at refinance if anyone's inclined. My personal as Sean Xu, SCA and Xu at ProtonMail. Com probably go for corporate first and then personal, but on you know generally to happy to speak to anyone about this.

Stefan (00:19:12) - And my last question, which I always like to ask everybody, what's your favorite song right now that you listen to or music?

Sean (00:19:25) - I want to say, actually, Paul is dead by scooter. And so people, everybody in Germany probably knows who scooter is. I'm not so sure about Switzerland. I've been a fan of scooter for almost a decade now.

Stefan (00:19:39) - Excellent. Thank you very much, Sean. Very much appreciate you took the time to speak with me.

Sean (00:19:44) - Anytime. Thanks for featuring me.

the engine
behind active management

Schedule a call