Is AI Ready to Optimize Your Portfolio?

0:00Your favorite neighborhood spot grows with Square. Indeed, my favorite neighborhood spot has quickly become Todd Snyder in Williamsburg. Todd Snyder is one of my favorite menswear shops and has supplied me with all the clothes I have needed this quite hot summer. Every business has different goals, but Square is the business platform that supports them all. From opening a new location, selling something new, or just expanding their reach. Indeed, I've seen it with Todd Snyder. In Square, also, you can get real-time insights, so don't wait for end-of-day reports. Go to square.com forward slash go forward slash realvision to learn more about how your business can grow with Square.

0:43That's S-Q-U-A-R-E dot com slash G-O slash R-E-A-L-B-I-S-I-O-N.

0:57Hey everyone, if you like this podcast, go behind the paywall to get privileged access to the smartest minds in finance. Join the Real Vision community and learn how to become a better investor. Visit realvision.com slash rvpod and use the promo code podcast10, that's podcast10, to get 10 % off our essential membership for the first year. Now, to the top analysis of today's crypto markets.

1:30Stefan Klauser, welcome to Real Vision. Hey, Ash. Great to be here. It's great to have you. So much to talk about. We were talking a little bit off camera about your company, iSot, that does AI, this convergence between AI and crypto, one of the most interesting stories that's happening in the space. But by the way, you have a background in Ethereum. We were talking about this a little bit off camera. Obviously, last night, the Chapelle upgrade. We can see if we bring up the chart of Ethereum price, Ethereum cutting above the$2 ,000 mark a little bit earlier this morning. I think it's trading a little below 2 ,000 right now.

2:06Looks like on my screen around$19.99, spot 9.5. Boy, we're right at that threshold. Stefan, let's talk a little bit about what happened last night about the Chapelle upgrade. Even the name is confusing. Let's talk a little bit about what it all means. Yeah, Ash. I mean, it's great to talk about this. I think it's a major milestone, again, in the Ethereum network. Clearly, the move to proof of stake was one of the first important milestones in this respect. And now it was about being able to kind of retrieve the staked coins. and this is another important aspect. And, you know, there was a lot of talking about the fact or the question whether this is going to influence the price of Ethereum and in which kind of phase this is going to happen.

3:06But as we've seen, there's a lot of effects that can play a role here. On the one hand, the transition was where it moves. So I think that was certainly a positive. on the price, that there were no disruptions, no issues. And on the other hand, there was the question whether everybody is going to just retrieve the stake points and going to sell them and whether this is going to put some pressure on the price. But this is obviously not what would happen, at least not in the short run. There was also some mechanisms that prevent this from happening, right? There is a maximum that can be withdrawn.

3:46And at the same time, I think what people forget sometimes, the people that were staking on Ethereum in the first place are not the ones that were in for a quick buck, right? And this might be even a more important aspect of this. Stephen, you have an interesting background. Let's talk a little bit about how you got involved in the space of cryptocurrency more generally and then about the transition that you made into AI. and of course the union of these two just red hot topics right now yeah i mean uh it's actually really interesting my background originally is in political science political economy um that's what i what i studied at university um and then i was doing a lot in innovation management especially also in the international level um and then it was back in 2014 or so when a colleague approached me with the topic of fintech.

4:43And this was really something that was just popping up in Zurich, Switzerland, where I'm based. And we said, oh, this is really something that is ideal to kind of foster a bit in terms of innovation because we have a strong financial market, but we're not so good at service innovation. So let's get engaged there a bit. And then it was very quickly that I stumbled up on the topic of blockchain back then. And this was in 2015. And I got so interested that I even signed up for the first DEFCON 1 of Ethereum in London. That was back in 2015 as well. It was still rather a small group. and some of the topics that are still discussed today were already imminent back then, but the topics were also a bit different, right?

5:40It was a lot about decentralization, about governance, about the code is law and similar things. And at the same time, staking was already the kind of first idea that was around. And I got really stoked. So I then went to work for ETH Zurich, which is not Ethereum. It's a technical university. It's one of the most prestigious universities in Switzerland or in continental Europe. And at ETH Zurich, I led a blockchain project that was based on ETH Ethereum. So it's quite funny. And this is how I have a really good hook with the topic. A lot of Ethereum. Yeah, a lot of ETH. And there was some confusion in the beginning.

6:29I think sometimes I got invited to panels because people thought I work for the Ethereum Foundation, but I actually was working for the university. So it worked out perfectly. Absolutely. And then we did this project on distributed incentivization mechanisms for sustainability that we built an MVP on Ethereum. And that was kind of a publicly funded project. So as often these projects come to an end at one point, and I also met some extremely talented machine learning researchers at that time at ETH. And they were doing some papers around novel machine learning algorithms that they invented. And these algorithms made it more simple to combine different data sets with structured and unstructured data.

7:28You imagine market data on the one end, more text-related data on the other end, combine that and then forecast Bitcoin volatility. And they could show that they can bring down forecasting errors by up to 50%. And this was a huge thing, right? There was a lot of, at that time, arbitrage traders in Bitcoin. It was 2019. People that came to us and said, can you please tell us how this is done? And this was kind of the first moment, the first foundation for the company, ISO. Say that one more time. So, Kenneth, this was the foundation for ISO. So from this paper and the reactions we had, we somehow realized, okay, there is a big demand for machine learning driven approaches.

8:22And we decided that you want to expand this into a platform. And so iSOT was born. So let's talk about where we are right now with iSOT. What are the products and services that you offer? What marketplaces are you in? And what's the competitive advantage that you guys feel you have in terms of product market fit? yeah i mean we we have a couple of products currently most of them are still related to to the crypto market um so you see and usually when you do machine learning or quant techniques in finance most people these days would would do that on equity markets some very little on the bond side or fixed income and then some of it some of the activities in is are going on commodities and then you have a lot of talented people in crypto but they usually use use these insights in a prop trading environment right there is very little or no as a service component there and what we do is kind of we provide advanced machine learning and quant insights as a service So mostly what that means, if you have a running portfolio or a product, a fund, an ETP or anything, you can come to us.

9:47We have the engine to kind of say, put this into the engine, the machine, and tell you what a completely systematic view would mean on this portfolio. And we can also create full strategies out of this. So we have the full spectrum there. And so the competitive advantage in that sense is clearly we provide this as a service to companies that don't have their own quantum machine learning team in-house. And there's many of those, right? And we are working with some of the most talented researchers around machine learning in finance worldwide. And this means we can always provide the latest technology to customers.

10:35Give us some examples of that application specifically in terms of what you guys are producing and how it's used. Yes. So we will go into a live demo a bit later even. For example, the audience here together with us can create the portfolio. But I have a very easy, clear example for you. So you have a fund, right, the crypto fund. They have their fundamental views as much as you can create the fundamental view on crypto, right? This is already a bit difficult sometimes. There is not so much data and information that you can base it on. In terms of evaluating crypto in the terms of traditional cash flow portfolio allocation structure, it is a little bit more challenging.

11:24Although we're moving in that direction, obviously, with staking on ETH. Absolutely. What I wanted to say here is mostly that it's very important that people have these fundamental views, also that they're kind of checking the different blockchain layers and protocols, whether they're reliable, how much code is created, etc. But on top of that, what is often missing still is a machine-driven view that takes all the relevant data around prices, around macroeconomic events, around news, etc. And then with all the information, more or less in real time, creates a machine learning and quant-optimized view on that portfolio.

12:11So you can come with your portfolio or product, you give it to us, you give all the restrictions, We let this run and at rebalancing date, we give you the allocation that the machine would do at this point in time. Hey, everyone, we're going to take a quick pause and hear a word from our partners. We'll be right back.

12:34So what would be an example of that with a simplified sort of toy portfolio that we might be able to understand in light of, let's take, for example, CPI data coming out yesterday?

12:48I mean, I think it makes most sense if we do it in live and show the audience. Should we jump in, Stefan? Should we jump right in? We could do it right now and then take questions or take it from there. Yes, absolutely. That sounds fantastic. I love doing live demos. It's something that we do here at Real Vision that I think is just extremely cool because it gives you a sense of what it's like to actually apply the technology. So whenever you're ready, let's get that spun up and we'll take a look. All right, there we are. So just some disclaimers up front, right? First of all, it's not UI optimized yet.

13:28So it's mostly about functionality. Currently, we generate these views in-house, but the next step is that people would get access to the beta platform directly. and then you can generate your own views. So what you see here is, first of all, the backtesting period. You see that the regular backtesting period starts on 1st of January 2021. It has a simple reason. First of all, most of the coins that we cover, they did not really exist for much longer. So if you go further back, you would have what's called a survivorship bias in the outcome. and that's what we want to emit. And also we have many different, let's say, market phases since January 2021.

14:15You had very bullish phases and then also very, very bearish phases. So it's an ideal, ideal environment to check the robustness of such a portfolio. This is the first step. Then we have the optimization parameters, right? We start usually with the cash allocation, So a risk-free component to the portfolio. And we could select anything from zero to 100. As I would like to have this as interactive as possible. Maybe I can see the chat here, but you can tell me or ask the audience if somebody has a view or you have a view, Ash, how much cash would you allow maximum in the portfolio? Let's cap it at 20%.

15:01Okay, let's put 20 % for this. Then we usually can allocate a fee. So customers usually in crypto, sometimes it differs dramatically, right? Some customers have very low fees in trading, more the crypto banks here in Switzerland that are more regulated, have much higher fees. So this is then automatically deducted from the P &L. And then what we put here is a target risk. volatility. You can imagine that like this, we can steer a bit to whether we want to have less volatility or a more aggressive high volatility portfolio. It's only the downside volatility that we covered here. So that means that the resulting volatility will be around two times the amount that you see here.

15:55But in general, more volatility, more profit, or less volatility, what would you vote for? Well, let's keep this relatively conservative. What are the plug numbers you generally use in this for your ring? Actually, I think what works very well is if we try to restrict volatility a little bit, but not too much. So if I choose something around 37 or so, that's what I would think would work really well. Okay. Then maybe for something for the audience in the chat, if they feel like selecting something, otherwise we will just randomly do that. Well, we've got some coins from the audience. You tell me, Stefan, how many can we plug in?

16:47I would say 8 to 10 easily we can also choose more but I don't know whether we cover all of them we have around 50 or so in the database currently. Okay that sounds good I think we've got about 7 or 8 that I'm looking at right now. First one of course no surprise is Bitcoin Yes I think we should probably put in Ethereum as well given the events of the day we've got Filecoin

17:19Bitcoin Cash don't yell at me guys this is from the audience

17:30Horizons

17:33I don't think we have that okay we skip that one let's just get the last two because I know they're both popular things we've talked about here on Real Vision Chainlink yeah

17:48and finally zcash for the privacy coins

17:56yes okay um cool how many do you have two four six let's put two more um something fun right that always everybody likes uh from the other audience that i spoke to Doge Doge Doge Is Doge off Twitter now? Are they no longer in the Twitter sidebar? I think they're no longer there I don't know I don't think so, yeah What else? We have Solana, Uni Maker, Matic, Ocean Anna Let's do Matic Matic, okay Cool So then what we can do is to assign minimum and maximum allocations. So we can either say let's keep it to the machine to decide, or we can say we have some preferences. What I would usually do just for these demos is to just put it to, I don't know, 0 to 40 % each.

19:07so so basically no floor for um um and this and the uh and the minimum is uh there's no minimum no minimum and the max is 40 percent yeah exactly so but if you have sometimes professional customers right the funds they want that minimum allocation in bitcoin and ether usually um so then we can easily deal with that so we can also do a 10 let's say a minimum yes minimum minimum 10 each so they go they can go a bit higher as well maybe and the rest to 40 maybe we say doge we don't want to be 40 percent in doge maybe only 20 max and then right let's do let's do it like this right and then what we can do here is that customers can upload own data files so if they have a quant or research department it can be uploaded here also if they have any kind of text files or so that we can work with so it can all be taken in.

20:07You have to imagine what happens in the background in principle afterwards is that we do the machine learning based forecasts and these are done pairwise between all the coins tokens that we have in the system. Then you create the covariance metrics and we created the risk forecast that is during the cash allocation and then all this information are taken into a so-called black litterman optimization that then afterwards creates the portfolios but let's start this what's now happening on the fly and the whole optimization is taking process and taking part and then the actual training of the models has already happened.

20:59This is happening dynamically. This is really being run. This isn't just a demo. This is actually being run. This is real numbers, real processes. It's not a demo. Okay. Then we have the results and you see our portfolio, or the customer portfolio here was quite a good one. So we would currently be the two green lines or two variants of our strategy. One is the volatility is trying to minimize volatility a little bit more than the blended, but you see they have been very similarly performing here between 612 and 682 % plus since January 21. If you compare this to an Ether tracker as a baseline, Ether would be a plus 156.

21:53And then you have some benchmarks like Bitcoin, which is a plus four, and some market cap weighted indexes, which are in similar areas or even below Bitcoin. So this is quite a good outcome. Then we produce the statistics around it. You would see Sharps around 1.4, 1.5, Sortinos around 2.4, 2.6, which is a huge difference to Bitcoin and Ether. And we would even avoid some of the drawdown, right? As we only selected 20 % max cash, we were still in a similar area, but we had less drawdown, similar volatility like Ether and clearly better returns. and now you can. Let's just break this down a little bit because I'm sure that folks in our audience who are professional investors immediately understand what you're talking about here when we're talking about Sharpe ratios.

22:51For those who have not had the benefit of Mr. Sharpe and Mr. Markowitz's portfolio theory, let's talk a little bit about what this means for someone who may be a retail investor who may be a little bit puzzled by what some of these data represent. Absolutely, yeah. So sharp is usually when you take, in simple words, when you take the return, but you control it through the volatility. So that means it's return, that outperformance in terms of returns as compared to a benchmark. So in other words, risk adjusted returns. Exactly, to be very, very quick and simple risk adjusted returns. And Sortino is doing the same, but it is limiting it to the downward volatility.

23:44So actually in principle, that's usually what we want to maximize because you don't want to divide through the full volatility usually, because volatility we say is not necessarily a bad thing, it's only a bad thing if it goes down. So Sortino is the measure that looks at the risk adjusted in terms of downwards risks. And then you have... Max drawdown is a calculation that shows the sort of return to worst from the highest point to the lowest peak to trough. Exactly. And you see, this is very quite similar. It's even a bit better than with Bitcoin. naturally this comes from the dynamic cash allocation in the portfolio.

24:33And then we have the volatility. You have to imagine that Bitcoin from all the coins and tokens that are usually in these portfolios has the lowest volatility except if you compare it to a stablecoin naturally but otherwise Bitcoin has quite low. So voila, all the other coins have quite much higher volatilities. So what you want to achieve is that you get a portfolio that has volatility that is not exceeding the Bitcoin volatility by a huge margin. And this is what we even can achieve if we put the volatility a bit more down and put a little bit more cash. we can easily also achieve volatilities in the same area as Bitcoin.

25:24Hey, everyone. We're going to take another quick break and hear a word from our partners. We'll be right back to the Real Vision Crypto Daily Briefing.

25:36So let's talk a little bit about where we are right now in terms of deployment of this technology. Is this currently live in the wild? What does it look like in terms of your customer base? Talk a little bit about the go-to-market strategy. Yes, so where are we now? We use this for some products. Mostly are these structured products that we issue. We like to do that with partners in a white labeling manner so that we can profit from the client base and distribution power of partners. we like to do the strategy and they do the rest but we also start to issue an ISO product that will be available in the first place it will be available mostly on the Swiss market it's a bit tricky in terms of regulation to do that internationally but we can always also apply this to client portfolios directly so this is ready to go live.

26:49The next step then is what also is a bit easier from a regulatory perspective is that we give customers access to the platform directly. So that's what we did here now. If you run your own fund, if you're a so-called qualified investor, meaning a professional or institutional client, you can already today apply for access to this beta platform. We've got some questions coming in. Perhaps we could just take some of these because they are interesting ones. Here's a very sophisticated question from Ralph on the Real Vision website. Does iSOT's tool do CVAR that's conditional value at risk? Is that one of the portfolio parameters that you are allowed to modify?

27:39Currently not, but it's on our list, right? The features that we want to implement. So what you see here today is what the features are that are implemented already. What the idea is of now getting better access is then that better users get access to the platform as it is today and then they can request new features and new things to be built in and we try to deliver on them very quickly. Very cool. So, you know, I know you guys are working on this and obviously developing it as the product begins to mature and you build more features. Are there any other potential applications that you guys see right now for AI in this space?

28:24One of the things that we've been hearing a lot about here at Real Vision is the idea of the security component of artificial intelligence AI, essentially doing automated code audits and some other interesting things. We should say this technology obviously is embryonic right now There are lots of applications that are coming online. Anything else that you guys see that you think is interesting or might think is interesting at the future? I mean, it potentially is huge, right? We have to make sure that we focus on a few things in the beginning and we don't get lost in that space. I mean, what is certainly for us interesting at the moment is anything around strategy and portfolio optimization.