Ronnie Sadka: Thanks for that kind introduction. Good afternoon everyone. Very nice to be here. I see a lot of, friendly and, familiar faces. It's great. So I'm going to talk to you about past, present and future. Let's start from the past. For the past. I thought, I'll just spend a few minutes telling you a little bit about the genesis of MQTT media stats. You guys remember what happened in August 2007? First two weeks of August 2007. Remember that something happened in the market, right? The market itself. Average. You didn't see anything. Market was up in August. But anyone is doing any cross-sectional strategies. Knew that something was going on because value moment. They all flipped very aggressively. Remember that and talk to you a little bit more about value moment later. These two factors. It's very interesting. Like what's the narrative. What happened there for Gideon and myself. Gideon is my co-founder is sitting somewhere right here. You know, we were thinking, wow. I mean, no wonder everyone is getting hurt. Everyone is doing the same thing. The all red farm and friends are doing the same thing. One gets hurt, everyone gets hurt, or one big enough gets hurt. Everyone's going to feel it right. They're going to trickle the trades because it's going to cause some risk management reactions, and everyone is going to feel it, and there's going to be some contagion and all of that. So our idea was, well, let's maybe try to do something else.

Ronnie Sadka: If we rely on different type of information, maybe we'll get different signals, okay. And where we will be able to differentiate our strategies and really obtain alpha that is over and beyond what is spanned by the current factors. Okay. Fast forward a couple of years, I think this was August 2009. 2009, I think. And Gideon and I were starting to think, well, let's maybe look at media, let's look at the internet, let's look online and try to gather information online. Both of us are engineers. We thought, oh, this is simple. We can do it. So start writing code. And this was, if, you know, in Europe in August especially, let's say Paris, no one's around, no one's working. But Gideon was in the office 8:00 pm. He tried to run a code that automatically retrieves information from the web. Within a minute he got a call from the IT office saying, hey, what you're doing? No one in in the bank was working at could access Google anymore because it shut down the entire bank because they realized there's a bot they thought someone was trying to automate, retrieval of information. They didn't want to do that. That's when we realized maybe we have something here. It's not that easy to do. Something's not easy to do. That means there might be rewards for actually the people who are willing to spend the time and effort to do so.

Ronnie Sadka: So fast forward now you know more, almost 15 years later, and media that have almost 25 people in this group, half of them are full time engineers that are just dealing with data retrieval. And the rest of us are analysts and PhDs in finance statistics. And we try to show insight from this data. And we have this wonderful partner, State Street, that helps us show, showcase all the different technologies and signals that we're able to retrieve. So what I want to talk to you about today. I want to show you a few things that we've done, that we're currently doing, and how we think the industry is going to shape up in the next several years. So in terms of, you know, past, I'm going to talk to you a little bit about some theoretical background, because a lot of what we do, we're going to try I'm going to try to show you use cases, but all of them are grounded in deep economic theories. So I'm going to show you a few articles. Try to review them really quickly just to convince you that there's, you know, there's some economic research on this. Then I'm going to go more into what we're doing now. How do we retrieve information? What are the type of signals we use, and go over a few use cases. And then I'm going to show you a little bit more about how Llms and AIS can help us even more, become more efficient in retrieving more insight.

Ronnie Sadka: So I have a slide here. This is really that's the only slide that I'm going to talk about. Theory and academic work. There's two slides to this slide okay. And it's not a balance sheet. One side is market efficiency right. That's going to be on the left hand side and your right hand side. Behavioral finance right. Everyone heard these terms before right. So let's just review this. And we have you know marketing efficiency Fama 1970 and his idea. Well the market you know the price should be right. Right. For information is embedded in the price. The markets are efficient. Right. That's how we think about market efficiency. There's forces in the market that would force this or reinforce this. So that's so that's what would happen if people the price is wrong and enough people know about it. They will trade to the direction it should be. And in equilibri the market should be efficient. And then there's a few a few different degrees of efficiency that he talks about. Right? A semi-strong strong and weak for efficiency. One interesting article is Grossman and Stiglitz. I'm sure you've heard all these people that you see or all these notes, all these papers, each of them. I just picked them because each of them was a Nobel laureate.

Ronnie Sadka: So you got the Fama and Stiglitz and Merton Thaler, Shiller all received Nobel Prize for these ideas that are really the backbone of what we're trying to get at with our signals. The interesting thing about Grossman, Stiglitz, they talk about the cost of obtaining information. So it could be an equilibri result where it's costly to obtain information. Those who pay the cost are able to retrieve returns or alpha. And what I the way I think about our effort is that we are trying to gap to bridge that information gap, and we're basically doing the work for you. We're investing in understanding. We're obtaining information, therefore, for everyone, helping everyone better off if you become a state suite client. An interesting result in Merton is that assets. This is where we start to think more about the media. When you look at companies that are that have media coverage versus those who don't have media coverage, you can think about getting some intuition from Merton 87, saying that some firms are known. Some firms are not known, though those are more well known are going to have higher price, therefore lower expected returns in the future. He shows. This is an interesting equilibri result, and we're going to use media coverage to proxy for familiarity, for attention, for people they know particular asset or they know particular theme or narrative. I'm going to talk about that soon. On the other hand, you have behavioral finance that are relying more on some kind of bounded rationality, some friction in the market or in han behavior.

Ronnie Sadka: That would create some predictability. So the price is not right. People. Some people understand it, some people don't. Those who understand it might get, be able to extract more return. So the different biases people talk about are overreaction, overconfidence. There's all sorts of different, behavioral biases. Bob Shiller was here in this conference, I think four years ago, 3 or 4 years ago, talking about narrative economics. And in his view, he drew some, motivation from Samuelson's dict saying that it seems that when you look at individual firms, the market is very efficient. A lot of analysts are looking at firms what their financial statements. There's a lot of coverage on firms. But when you look at the macro economy, a lot of people miss the global or, you know, long run trends. So that's his idea about micro efficiency versus macro efficiency and what we're going to be doing today. What I'm going to show you is we're going to be able to quantify these narratives that might seem intangible or elusive. We're going to be able to give you some tools to actually quantify that. All right. So now let's talk a little bit about what we actually do. We obtain media information. Point in time. We have about 150,000 different sources.

Ronnie Sadka: That set grows all the time. And we collect everything. It's like saying you have another copy of the internet. It's about 5 to 7 million articles every week about relevant, you know, relevant articles relevant to the to the assets that we're trying to, understand to the topics, etc.. So you have here, you know, sources, we obtain all this information, we organize them into reservoirs. What does that mean? A reservoir a reservoir is a set of information pertaining to a particular asset or theme. So the ones that we typically look at is we have country equity news. We have firm specific news, we have ethics related news and we have just general press okay. So these are not they're not mutually exclusive, sets. But we organize all our media information into these sets. One of the reasons we do that is because we understand that there's a lot of biases in the media coverage. So the journalist decides to write about a particular firm at a particular time. So we need to try to correct for these biases. We spend a full session just on that. So I'm just going to say that by organizing into reservoirs we're able to correct for some of these biases. And then we create indicators at the single stock level macro indicators central bank and thematic. So I'm going to go over just a few indicators today specifically single stock and and central bank.

Ronnie Sadka: And then I'm going to show you some narratives. And so here's one thing we did early on is trying to understand the media coverage of particular firms. So we have a suite of products that is just at the firm level. What you see here, this is the S&P 500. And it's going to be really large cap firms. But we create these indicators for each firm. What is an intensity indicator. What is that. That's looking at the percent of articles. Of all the set of articles that we have in a given day, the percent of articles that talk about a particular asset, okay, the stock in this case the stock, but we can do it on a currency and we could do it about a country equity, any asset that we are interested in. So we look at intensity is how much attention is given to a particular asset. Sentiment is conditional having coverage. What's the overall sentiment? We'll look at an article. We'll look at the words and try to understand what's the sentiment for this particular, suite of products. This is this is a second generation of a you know, the first generation. The second generation now uses Elm. So we actually use a large language model to understand sentiment. The first type of product we had just looked at, counting words and looking at the bag of words approach. It does seem to be that LMS provides some additional, additional alpha.

Ronnie Sadka: And you're going to see it in these results in the results later. This agreement is another type of indicator looking at how dispersed the sentiment is across the different articles, the coverage, the coverage, a particular stock. That's this agreement. Hard content ratio is looking at the articles and looking at how many articles talk about actual nbers versus more soft information. So that also seems to have some predictive power. And there's a HDR dispersion which also looks at the dispersion in, in the in the amount of positive amount of quantitative versus soft information. Long story short, when you put all of these. Indicators or signals together. Our research shows that you have some outperformance. Okay. We can talk, you know, maybe later in Q&A or you can reach out to a representative, explain exactly what we did here. But this is a strategy that rebalance rebalances every day. It's a five year, five day, holding period. It's equally weighted. There's a lot of details in it, but the point is that there seems to be information in this media coverage. And when you look a little bit more specifically into the reservoirs, remember I said there's different reservoirs, different types of information. I think it's pretty interesting that when you create these signals on a subset of articles, you see that there's difference in performance. So when you, for example, when you look at, let's say, look at general or corporate, what are these general is articles that are mentioned in the general press versus corporate.

Ronnie Sadka: These are articles that are these are really, just press releases of companies. And when you run these strategies on general and corporate, you see the returns are not as significant as looking at financial. Financial are magazines that are, let's say, specialized magazines that talk about finance. They seem to be more predictive than just the general general press. So I think that's interesting. And we also did some results with in region and out of region. This is again, without going into too much detail, we can look at the amount of traffic that goes into different websites. And so what we do is we either over we can overweight sources that have more traffic in them. Okay. And we can do it by region okay globally. So that's again just a lot of things we can do that the amount of granular granularity could be really significant. Again I'm going to go over this just fast and show and just say this seems to be that the data is very informative. Let's talk about central banks. So stage three asks us to look at central banks and whether we can say anything about, about whether the the bank is hawkish or dovish or what can we say about yields. And the first thing, we take a lot of pride in our ability to work with data.

Ronnie Sadka: But the first indicator we looked at was just the central banker height that actually ended up performing pretty well out-of-sample. The problem of course, that the frequency is not that high. But anyway. But more seriously, what's the issue with central bank if you just look at central bank communication? Because meetings only happen eight times a year on average, okay. Every six weeks you don't have really information in between meetings. When you look at media coverage, you could because they talk about it all the time. We all love talking about what's happening in rates. Do we have the information more continuously? The other issue is when you look at, communicate action from central banks, they're so heavily scrutinized it almost becomes uninformative. Anything they say, they really look at every word. Right. So then that really takes away all the juice. It's like they can't really say anything. But the interpretation of what they're saying that ends up being important. That's what we're looking at okay. So we created some indicators and we we published a few papers. Let me, let me let me share with you a few of the results. So we look at, what you see on the bottom. The bottom left is, is just a smary table of, you know, eight year period. We had about 500,000, different articles that talked about the fed.

Ronnie Sadka: And what we did is we we tried to understand whether an article is more hawkish versus more dovish. So we did it in different ways. and the first way we did is just looking at the nber of times people talk about increasing rates versus decreasing rates. Okay. That was our first hawkish dovish indicator there. That ended up performing well. But I'm going to show you what happens when we include some other information. The other thing you see here is just an example of a dashboard that we have that we created. And what I wanted to show you here is that we actually look at the communication, the sorry, the coverage of each central banker. So we can look at each person, whether they're hawkish or dovish, and what the different colors that you see here, they they differentiate between the voting members versus non voting members. Basically we did that for the fed. We did it for the ECB and the Bank of England. And in the bottom what you see here there are different graphs. We aggregate that information. So you can look at overall you know the voting members are the more hawkish or more dovish than the board or than the the better the the or the chairman. So these are different you know we different indicators that we that we publish. And and you might have seen a few articles that that states with analysts have written on this.

Ronnie Sadka: Let me show you some of our new research that compares. The Gen one is what I just mentioned to using a large language models to understand how Dovishness. So in this in this slide here, what I'm showing you is two things. One is we're trying to understand relevancy and the other is yield prediction. So first thing is what what we're trying to do is look at an article and understand whether it's hawkish dovish and see whether that is contemporaneously moving with yields okay. If it does it means that probably we're we're finding you know our model works in the sense that it captures instantaneous contemporaneous movements. The second thing is yield prediction. Can can it actually be used to predict yields. And what we have here is gen one that's this first coln. And then you have Roberta, which is fine tuned as well and fine tuned to to looking at central banks, communication. And we have the ChatGPT that we all know of and Lama two and we, we looked at understanding hawkish governess with these four different models. And we try to look at the correlation contemporaneously across the yield curve. So you see for the one month, the three months, six months, etc. etc.. So these are contemporaneously weekly changes in yield and what you can see. And I and I put there you can see it in red that ChatGPT seems to be actually doing pretty well here.

Ronnie Sadka: Okay. The Gen one is significant. Chatgpt gives it another boost 67% boost. Okay, you can see that coefficient. I can't point with this, but you see the 163 versus the 268. Okay. I plotted these results on the top right. You can see that all of them seem to be working, but it does seem to be the case that LMS could give you a little boost when you look at prediction. That's the bottom graph on the right. It does seem to be the case that again, all these models work well in terms of understanding future changes in yield. And this is for the two year rate. But the, the LMS seem to be. Increasing the rate of prediction by 28%. Okay. So again, the point of this is to show that yes, we can include LMS. We're working on that. We've been doing it. They can give you and they can give you enhanced measures both contemporaneously and in terms of prediction. All right. Let's now talk about narrative analytics. So I just want to explain again the difference. What I've learned so far is you look at a particular asset or you're looking at a central bank and you're trying to understand what they're talking about. What's the tone? Can that help you predict returns? What I'm going to do now is I'm going to switch this to talking about narratives.

Ronnie Sadka: I'm not going to focus on the specific asset. I'm first going to just try to understand what are people talking about, okay. What is the narrative has nothing to do yet with any asset prices. How do we quantify a narrative? We look at we have this vast information from the web, and we code the percent of articles that talk about the narrative. What I have here for you are some examples. When you look at the top graph it has three graphs there. There's the Covid graph. So this is a graph that looks at the percent of articles that talk about Covid 19. You can see that, you know early 2000 you see it went up to 0.75. So that's 75% of all articles in our reservoirs talked about Covid. I think that makes sense. Right. And when you over time it dropped there's two more graphs there. One is inflation and the other is military escalation. So you can identify again the military escalation. One is the one that in 2022 remember February 24th the Russia invaded Ukraine. And you can see that goes up that jps up and then it falls after. And the third graph is inflation. I think the inflation graph is very interesting because you can see around 2017 2018 there's a couple of bps there. You see that. It's interesting because people were talking about inflation. It didn't materialize, but people were talking about it.

Ronnie Sadka: This approach allows you to understand the time variation in attention to different themes, topics, narratives. Once we quantify them, okay, we can do a lot of things with it. I'm going to show you a few applications, but the point it gives you a score, it gives you a score. Today military escalation is five. Tomorrow is two. Okay. Inflation discussion today is one, tomorrow two maybe after that it's one and a half. That's what you see here in the data. We're going to take it a step forward. We're going to say okay how do stocks or assets are how do they respond to this. Like what's their beta with respect to the narratives? That's where we're going before that. Just to show you in the bottom graph we looked at, we kind of zone in on on Covid and we're saying, well, when you look at Covid discussion, not all discussion was as fast. When you look at the graph that really goes up, that's reservoir. So if reservoir is a reservoir that it's really traders talking about what's happening. Okay. These are kind of trade magazines. And normally they talk about anything that happens in the world that happens really quickly. And there's immediate coverage in these magazines and these venues. So you can see that immediately jped even before it was about to mid-January. There's already discussion on Covid. It wasn't called Covid pandemic in mid-January already in trade magazines in effects after that, corporate and media that discussed corporate started picking up on that, then general and then politics.

Ronnie Sadka: And you see what happened after a month or two months, politics shot up. Remember, the US was very divisive, right? Mask, no mask. Right. And that took up the entire discussion. We look forward after that a few months later around the presidential debates. You see again a jp in politics with respect to Covid. Okay. So that's just to show you again, the power of understanding who's talking about what, who's providing attention. Seems that the trade magazines here effects were were the first to the game to explain it to to talk about this. And that's why for a lot of our application, like the central bank did show you before, we rely on the reservoir to create these indicators. Okay. So what do we have? We done the first thing we've done with narratives. And you can see these are graphs that I took from insights. So from the portal from insight you can access these. And this is what we call the narrative map. What does the narrative map do. It shows you on a two by two. It shows you the amount of discussion on a narrative. That's the vertical axis and the horizontal axis. It's like an r squared measure. You take the time series. In this case it's the S&P 500.

Ronnie Sadka: You take the time series during the S&P and you regress it on the changes in the narrative discussion. Okay. And we did that for many, many narratives. So you can look at everything in a two by two, where you have the amount of discussion versus the importance of the discussion, like the impact of the discussion on the market. All right. And so we differentiate between these four quadrants. And you see that the top I actually like to start from the third quadrant. If you look at the bottom left that's a quadrant that there's no discussion. And it doesn't impact the market. Right. When you move to the right that's a quadrant. Quadrant four in the Cartesian sense, that quadrant, you see there's not a lot of discussion but the r squared is high. These are narratives that are just starting to pick up. But when there's some discussion they really move the market a lot. You go up to the first quadrant. These are narratives, a lot of discussion and the impact the market. And then you go continue counterclockwise. You're going to go going to go to the quadrant nber two, the top left where there's a lot of discussion, but it doesn't really impact the market anymore. When you look at this particular graph. This was taken a week ago. You can see that us, US growth, political elections and gold seem to be important and there's there's high discussion and they move the market.

Ronnie Sadka: When you look at other narratives, let's say fiscal sustainability or international conflicts, they still move the market, but there's not so much discussion about them. If you compare it to the, you know, before the previous three months. If you look at the time series of this, you're going to look at something like this. We don't need to go into details, but what you can see is you can take each narrative and see what's the overall discussion over time, what's the R squared, the impact over time. And it could give you an idea in the relative sense where you stand relative to the previous year, previous couple of years. If you look at political elections, you know, the top left. You see it's a situation recently where there's an increased discussion and increase in R squared. Okay, that makes sense. That's why you had a few discussions today about politics and what's the situation in the US. Indeed, people seem to be interested in it and it affects market returns. All right. Let me continue. So here's a few more use cases. Let's talk about again the concept of a narrative beta. What's a narrative beta. We've talked about quantifying narrative. So now we have a narrative. We know the time series of this narrative. And we can calculate therefore betas for a particular. In this case I'm going to show you stocks or sectors, commodities, you can take any asset and regress it on the narrative and understand its narrative beta.

Ronnie Sadka: Okay. What does it mean? A positive coefficient means when there's more discussion about the narrative, the asset moves. You need to worry a little bit about the directionality. That's like a detail we do like we look at the changes in the in the amount of negative discussion. So we take care of directionality if you're worried about that. But again the point is some assets are positively exposed to some narratives and some are negatively or not exposed. So the first thing it could help you with risk management to understand which of your assets are exposed to the different narratives. But the second thing, you can also create narrative mimicking portfolios. You can try to say, okay, here's a narrative trade war. I can build a portfolio that would mimic this, this narrative, how I'm going to long stocks that have a high beta to this narrative, and I'm going to short stocks, have a low bid at this narrative. That spread should follow the changes in the narrative. So the use cases here that I have, it's a few strategies that are based on something very similar, which we call rotation. What's a sector rotation? A sector rotation is you have these, let's say 11 sectors or 12 sectors. And you look at you look at the sector and you say how much discussion was on this sector in the last three months compared to the previous three months? Okay.

Ronnie Sadka: Sectors that have more positive discussion tend to outperform the future. Those have more negative discussion, tend to underperform the future. And if you look at the spread within the top four sectors versus bottom sectors for top three, bottom three, you'd see the graph that you see here on the on the top left. If you do something very similar to Mega Trends, Mega Trends, we looked at a group of about 24 Mega Trends. So these are, different, I guess technologies that that we base this based off of MSCI Mega Trends and they have like Smart City and they have a aging economy and they have, they have, a future education, all sorts of mega trends. And when there's more discussion on these mega trends in the these three months versus the previous three months, they seem to be outperformance in the future. So the whole point of this is that we're trying to understand what people pay attention to and the narratives that people pay attention to. There seems to be a continuation in performance in the next 3 to 6 months. So this entire rotation strategy is based on you go along those that have a positive discussion and short those have negative discussion, and you rebalance every period. We did the same thing for commodities.

Ronnie Sadka: The results there are also very impressive. And we also did a strategy which I presented here last year that looked at narrative rotation in a general sense. We took a universe of, I think it's about 600 different narratives for each narrative. We created a long, short spread, high beta versus low beta, and we invested in these baskets that exhibited positive growth in discussion shorter. Those have negative growth in discussion. And these as you know, you see the results right here. So the point is we can track narratives by understanding what people pay attention to. You can place positions and it's going to help you understand the future returns. Okay. So where do we where do we think this is going? So I have a few comments that I want to share with you. One is there's all this discussion about LMS, GPT, all that stuff. Here are a few things. We've been toying with this for a while. We actually hired a couple people that are experts in in this field, and we think there's at least two ways where we can fuse. We can fuse LM technology with economic variables. We can build an LM that is particular like fine tuned to the particular, needs of stated clients. So for one thing we can do is we can instead of looking at LMS right now, they just look at, you know, hawkish doves, whether, you know, whether a communications hawk is versus doves, but they don't take into account whether that communication happened before or after the meetings.

Ronnie Sadka: It didn't take into account the state variables in a given economic state. Variables at a given point of time are the rates high or are the rates low when this communication happened? Okay, so one thing that we think is going to be very interesting is to take economic state variables as an input to the process of training the LM. To understand yields. Okay. So that's one thing that we can we think could be useful. Another way is to just to take an LM and train it to predict yields to begin with. And the results I show you here is a first attempt in that. So this is really fresh out of like last night. So basically you take an LM and here we use Roberta. And what we did is we took the contemporaneous five day change in the two year yield. And we trained took four years. And we, we trained it to say, okay, can you predict the yield with Roberta? Then the following year, we ran the model to, you know, what's the prediction of the yield? We ran the regression of the actual yield on the predicted yield. And what you see is that t stat of 2.9. So that just suggests that you could use machine learning to predict yield. Okay.

Ronnie Sadka: If you do sorry. That's for contemporaneous. If you do the same thing for the future, the second row is future five day change in yield. There also you get these stats close to three okay. So if you try to begin with take four years, predict the yield and then use that same model to predict the yield in the next year. And you kind of do that every year. You can see that that could give you some results. So again, this is just A taste of what this could do. It takes a lot of it's very expensive time wise to run these things, but this gives us hope that this would be this is a good place to invest our time. Second thing I want to talk about is basket creation. We think that using the technologies that we developed and understanding narratives could help clients better create baskets tailored to these narratives. In general, we think that people don't trade factors. It goes back to the discussion of moment and value early on. People trade narratives. They wake up in the morning and say, oh, elections are coming up. How should I build my portfolio? Or something is happening in the Middle East. How is that going to affect my, you know, my portfolio? And what should I do to protect it or to invest if I think that something, you know, maybe I want to take advantage of this situation.

Ronnie Sadka: Okay. Or maybe I think, you know, maybe an AI itself, I want to invest in that sector. So how do I build portfolios that are really tracking better the narrative? So the way there's a few slides here that I'm going to show you, just an example of how that could be used. First thing, the way that we can use AI or in this case, you know, LMS is first understanding. If you have we've had someone come to us. Well, can you help me understand cybersecurity? If I wanted to build a cybersecurity basket, how do I do that? First thing we did is we went to our machine and we created what we call subthemes or sub narratives. What cybersecurity. Well there's sub sub there's cybersecurity insurance. There's I listed here a few there's security analytics. There's other things. So it came with a list of ten sub narratives. But in each sub narratives what are the different keywords that we should be looking at okay. So we use the AI or LMS to or ChatGPT to understand those. And then you get a list of keywords for each sub narrative. Then you can go into SEC filings and you can identify each sub narrative. You know which firms are talking about exactly that particular sub narrative. So you can look at the web descriptions of the firm. You can look at SEC filings that you have in the bottom, in the bottom left, and that chart that you have there with the circle and hold that network connection.

Ronnie Sadka: This is an example of a Palo Alto Networks and the firms that are connected to this firm. So we looked at the entire universe of SEC filings and firms that pointed to Palo Alto Networks from their SEC filings are noted in this chart. So, in other words, the way to say it is, if other firms point to you that you're in that area and you're competing with them, you're likely to be related to that narrative. So again, these are different ways to identify a firm to a narrative. And then at the end result is that every year or six months, any time you want to run this, you can come with up with a list of firms that are associated with the narrative, and you can do it. It's a multi-pillar approach. You can look at SEC fines, you could look at the web, you can look at narrative betas. As I explained before, we create the narratives and calculate the beta. And you put all of that all scores together, and you pick the firms that score high for each of these, for each of these pillars. And the last note that I want to say is, I think that you want to look at the industry. And I've been in this industry almost 20 years now.

Ronnie Sadka: Many people are talking in terms of factors. This goes back to what I started with. People talk about moment. They talk about value. They talk about credit. And I think what they really mean is trade war. What they really mean is, you know, bank credit. What they really mean is, is inflation maybe. But they have a hard time translating their thoughts to factors. The existing factors and what we need to come up with is a new paradigm where you can you can actually build a factor based on the underlying narrative. The narrative is what matters. That's the underlying. It's not the factor. It shouldn't be moment. It should be. What's the actual narrative, the economic narrative that in play. And by using the technologies that have showed you so far today, you can you can actually address that, that problem and solve that. So let me let me smarize. I try to give you here a little bit of a glimpse of, you know, what's the economic foundations. And there's theories behind everything that we're trying to do here. I think our effort helps understanding or bridging the gap between finance and technology. We're able to quantify narratives. Again, it's a very important, I think, aspect of just trying to quantify narratives that are otherwise they seem intangible. We're able to quantify them, and once you can quantify them, you can run betas, you can look at association and you can build baskets, factors that are going to help clients better their portfolios. Thank you very much.

Lee Ferridge: Say thank you. Ronnie, I think we've got time for one, maybe two questions. So let's try and take it from the room. We've got a roving mic. Has anyone got a question for Ronnie? No one at all. So I did have one on the iPad. Was the, the chart you showed about the rotations. They all seem to take off in 2020. What was behind that? Is there anything specific that was behind that?

Ronnie Sadka: Well, we noticed that I think we're comfortable that it's it's not the opposite that the other way around. But we do think that over time we've just been able to better capture information. And that's that's what you see. So so I think they just improve in technology. Okay.

Lee Ferridge: Last chance. Anyone got a question in the room.

Ronnie Sadka: All right. Thank you everyone Rodney thanks very much.