HOWARD WALPER, CEO AMERICAS, COMMODITIES PEOPLE:
Hello everyone, and welcome to today's webinar, “Fueling Success: Empowering Energy Traders with Automation Keys to Seamlessly Integrate Algotrading into Your Trade Flow.” My name is Howard Walper. I'm the CEO of Americas for Commodities People, and I'd like to thank everybody for being here this morning, afternoon, or evening, depending on where you are. We're really delighted to see how much attention this webinar has attracted in the global trading community. We've got more than 600 registrants from all across the globe, so we're expecting a really exciting discussion today. Over the next 90 minutes, we're bringing together a fantastic panel of experts to discuss some of the key issues involved with algorithmic trading, including how to manage this on the trade floor, how algo trading can integrate into your current structure, what sort of setup you need to do that, how to set up a trade flow and ensure execution, safety, and efficiency, and what are some of the challenges around risk management with algo trading and how to mitigate some of the very unique risks in this area. So I'm absolutely delighted and thrilled to be joined here by an exceptionally knowledgeable panel bringing a wide range of experiences to the session. But before we get started, today's webinar will start with a series of short presentations, followed by a panel discussion and Q&A from the audience. So, throughout the webinar, please post any of your questions in the Q&A box. And if you see any that come through that you particularly want to hear answered, be sure to upvote those questions. We're also going to be doing a little poll at one point during this webinar, so please do contribute your thoughts. We'd love to get some of your feedback. And on that note, I'm delighted to pass the floor over to Sebastian Mayer, managing partner for EStar. Sebastian, the floor is yours.
SEBASTIAN MEYER, MANAGING PARTNER, E*STAR:
Thank you, Howard, and welcome everyone who's joining us for today's webinar. I'm Sebastian Meyer from EStar. At EStar, we stand for a full range of energy trading, empowering energy traders with comprehensive SaaS solutions. Our solutions not only enhance reach and liquidity but also support your trading success in this evolving landscape. Today, I think we would like to learn more about the complexities and explore further opportunities to empower energy traders. And therefore, I would love to introduce you to our experts who will be sharing their insights with us. And I'm quite proud of it. So first in line will be Dr. Stefan van Aaken from Uniper. He will confront the challenges of distributing trading desks and integrating algo desks into their larger trading structure. And his insights will pave the way for us to understand the dynamics of smooth desk integration, I guess. Next in line will be Michael Ostendorf from E.ON, who will explain the path from seamless integration of market access towards the implementation of trading strategies. Michael's presentation promises to be a deep dive into the evolution of market engagement through advanced trading technologies. We're also happy to have…
MICHAEL OSTENDORF, HEAD OF PORTFOLIO ANALYSIS & DEVELOPMENT & HEAD OF EEM TRADING, PORTFOLIO STRATEGY, E.ON:
William.
SEBASTIAN MEYER, MANAGING PARTNER, E*STAR:
…Wilmot and his colleague Oliver Heming from ICE Global Network as part of the presenters. And together they will break down what it takes to achieve the fastest possible market access and what goes into setting up everything close to where the trading happens. Their talk will shine a light on the importance of speed and proximity in today's trading environment. Concluding our expert presentations will be my colleague and Managing Director at EStar, Dr. Andreas Campa. He will address the technical complexity of exchange connectivity. His presentation will center on setting up trading desks and algorithms across various exchanges to meet individual client requirements. And I think a topic of immense relevance, if you ask me personally, in today's fragmented trading landscape. And last but not least, thanks to Howard from Commodities People, who will orchestrate today's session. We anticipate having a good presentation session, followed by a really good panel discussion that is both engaging and enlightening for all of us. So thanks again to the speakers and all of you for joining us today and taking the time. And I think now, Howard or Stefan, are you taking over?
HOWARD WALPER, CEO AMERICAS, COMMODITIES PEOPLE:
Well, Sebastian, thank you very much for that introduction. It's a pleasure now to turn the floor over to Stefan to discuss a bit about some of the challenges of the distribution of trading desks and how to get the algo desk integrated into your overall structure. So, Stefan, why don't I turn the floor over to you?
STEPHAN VAN AAKEN, SVP DIGITAL TRADING DEVELOPMENT AND OPERATIONS, UNIPER:
Thank you very much, Howard, for the kind words and for the introduction. So, my topic today is to shortly talk about our company, Uniper, our challenges that we have seen and what we have been facing with our geographically separated trading desks, and on how to actually integrate our algo trading activities into it. Now, I thought I might start with a very simple slide. And please don't see these as complete. Geographically separated trading desks—what does it mean? For some, this is clear, but how do you actually end up with such a situation? There can be many reasons: strategy, risk tolerance, and the like, each with its own price to pay but also with opportunities. When you're in different locations, you face different regulatory environments, time zones, taxations, compliance, and reporting requirements, but you also have access to different talent pools and proximity to particular venues.
Long story short, there are good reasons to have different trading locations. And when you look at the technology used in these trading locations and the processes, it can either facilitate your trading or make your life miserable. That's where EStar comes in, which is a product we are using, and probably some of you as well. It's our single access to the market, to the venues, a major part. And the colleagues from EStar can explain that a lot better than I can. For us, it's very helpful to communicate with one system rather than multiple systems, and let someone else take care of communicating with the different venues. We see our competitive edge in making smart trading decisions and programming smart solutions that make those decisions.
Now, moving on to augmented trading—how we see the role of our traders developing. The future trader orchestrates computer programs, algos, adjusting parameters, and these algos interact with the market. Meanwhile, we still have traders who interact with the market manually, and you need to integrate these so they don't conflict. We introduced an internal market, a place where all trades from programs and humans are collected and sent to the right venues. We use EStar for the forward part and a solution from Volume for short-term needs. Our algos, the clever stuff, are developed in Python for non-time-critical tasks and in a suite called Deltex, using C#, for others. We use EStar and Volume's connectors to access the market, simplifying our setup and integration across trading locations. And with that, I think I've covered the main parts I wanted to say. Howard, is that time-wise okay?
HOWARD WALPER, CEO AMERICAS, COMMODITIES PEOPLE:
Yes, Stefan, that's perfect. Thank you for your insights.
HOWARD WALPER, CEO AMERICAS, COMMODITIES PEOPLE:
Sure, this is perfect. First of all, thank you very much for that presentation. It's a pleasure now to turn the floor over to Michael Ostendorf from E.ON, who will be discussing the bundling of market access to auto Algo trading. So, Michael, the floor is yours.
MICHAEL OSTENDORF, HEAD OF PORTFOLIO ANALYSIS & DEVELOPMENT & HEAD OF EEM TRADING, PORTFOLIO STRATEGY, E.ON:
Yes, thanks. I think Stefan needs to unshare his screen, otherwise, I can't share.
STEPHAN VAN AAKEN, SVP DIGITAL TRADING DEVELOPMENT AND OPERATIONS, UNIPER:
I'm just trying to find that.
MICHAEL OSTENDORF, HEAD OF PORTFOLIO ANALYSIS & DEVELOPMENT & HEAD OF EEM TRADING, PORTFOLIO STRATEGY, E.ON:
Thank you. Yes, that worked.
STEPHAN VAN AAKEN, SVP DIGITAL TRADING DEVELOPMENT AND OPERATIONS, UNIPER:
Thank you.
MICHAEL OSTENDORF, HEAD OF PORTFOLIO ANALYSIS & DEVELOPMENT & HEAD OF EEM TRADING, PORTFOLIO STRATEGY, E.ON:
Thanks a lot as well. So I hope you know it works. You can also all see my screen. Happy to have you all here and also happy to be introduced by Howard and have the chance to speak also on that topic here. So, what I want to talk about is what E.ON has been facing in the past and what we are looking into. So, from a seamless bundling of market access to an auto routing and algo trading world. And first of all, where is E.ON energy markets coming from? So, E.ON Energy Markets has been founded three years ago after the merger with Innogy. So we decided that all the different countries which we have in general, need to be bundled via central market access. We have started with that journey. So, energy markets function is responsible for one of the largest customer portfolios in Europe, with 400 terawatt hours of supply, mainly to B2C and smaller B2B customers, E.ON Energy Markets GmbH. So the unit which I'm working for and where I'm heading currently the front office providing market access to the regions is covering Germany, Netherlands, and the UK, which is our biggest portfolios overall. So if we look into that from a challenge perspective, so our customers usually say we don't really care where we get our energy from, it just comes out of the socket. If it gets colder, I don't care, just use more. So we are facing different risks as a supplier. So mainly commodity risk, shape risk, and volume risk. But the main risk which we had to cover with the integration of Innogy and after the spinoff of Uniper was that we do not have any assets anymore. So they are all with Uniper, so with Stefan's team, and we have just a very big customer portfolio. So the credit and counterparty risk is one of our major challenges. So we decided that you have all seen that it also became quite critical last year when prices have risen quite a lot. We have a natural short position in the end, because we need to supply. Our customers don't have our own assets, so that's why we need to take care of that. And with Exceda or eStar, we enabled a seamless market access for all the processes in the regional front offices, because our challenge was to bundle the credit and margining risk within one legal entity, to provide a central market access and also increase the diversification of supply on the other side, because previously we had a lot of supply from our colleagues from the past, from UNIPA global commodities or back then, E.ON global commodities. Same was on the Innogy side with RWE supply and trading. So we need to enlarge first of all and diversify our supply. The other challenge was that we still wanted to keep our decentral front offices. As Stefan has mentioned, there is a reason why we have different desks everywhere, because we want to have them closely to our retail businesses to ensure that we have very close collaboration. And on the other side, we also want to avoid cross-border tax implications like fixed or permanent establishments. So with eStar, we were able to integrate these different trading venues from all the forward and spot markets into one platform. And we've also been able to provide an internal market functionality, including an auto routing for really seamless process for these decentral front offices, and to avoid these cross-border tax implications. We have also pretrade checks implemented which are automated so that there are volume limits, value limits, authorized products, sanity checks, etc., that nothing is going on screen without a decision from our side. So taking that, where are we heading to as E.ON energy markets? So we are looking into three main types of auto and algo hedging use cases. So one is to automate the trade execution. So means really automated market price monitoring and applying a predefined execution strategy. And we are already running that on short term and we are under development and close to run that on forward as well. The second one is that we provide decision support. So we have machine learning running which is generating signals to provide decision supports for the traders. And that is based on supervised learning. And these forecasts are also predicting short-term and forward price developments. So that we know is it now a good time to buy or not for our retail business. The third part is that we are working on a benchmark trading functionality. So that really AI tool is analyzing the market and proposes trades to cover the portfolio positions, which is self-learning over time. We do that based on reinforcement learning and also working there together with some bigger companies like Microsoft, for example. So if we look now into the different angles of the auto and algo hedging activities, which we do, so we have a lot of stuff on the short-term side where we do short-term forecasting which is using algorithms. We support the intraday hedging with auto trading tools and we have a short-term optimization with also auto trading tools running on the forward side. We have that hedge recommendation engine. So as described where we give automatically generated signals to the traders telling them if it's now a good time to buy or hold or even sell back, then we have a tool which we are close to implement which is an automated sales hedging. So we have an automated market price monitoring including hedge execution for specific scenarios which we run via Exceda or eStar. And the third bucket which we're working on is that benchmark trader which then finally possibly can trade on its own. So we may be a bit behind Uniper on these topics, but at least we are deeply looking into that. And just here one example on that automated sales hedging. What we are also doing, we have developed an algo trading framework in the end where we also use the latest technology stack which is then in between what the front end or let's say that all to sales hedging app which is used by the traders and how it translates stuff then further to the market access venue, and that's mainly it. What I wanted to talk about note was a bit shorter than expected, I guess. But I think we will have a lot of time to discuss on the panel to go a little bit deeper.
HOWARD WALPER, CEO AMERICAS, COMMODITIES PEOPLE:
Absolutely. On that note, a reminder to everyone in the audience that this is being done for you all. So please do feel free to contribute some questions for the Q&A a little bit later on in the present, a little bit on this webinar. We'd love to get your questions, your ideas, any way you'd like to interact with the panel. So, Michael, thank you so much for that presentation. At this point, it's a pleasure to turn the floor over to William Wilmot and Oliver Heming from ICE Global Network and ICE, who will be discussing how to support ultrafast access and colocation. So, gentlemen, the floor is yours.
WILLIAM WILLMOTT, SALES EXECUTIVE, ICE GLOBAL NETWORK:
Thanks for the introduction there, Howard, and thanks, EStar and Commodities People, for organizing this webinar and event. It's really great to participate in it. Oliver and I, as Howard has touched upon, are from the ICE Global Network and the wider ICE group. ICE has got a raw business, as I'm pretty sure everyone in this webinar will be aware of. Of course, the exchange, which everyone will be familiar with. The ICE data services business, which I hope most of you at least, will be familiar with. Then you've got the clearing and mortgage technology. We're not going to touch on that today. Oliver and I, specifically from the data services business. And in there specifically, is where the ICE global network lies. So we'll go into detail about how that network, Netronet, disseminates data in a variety of ways, whether it's low latency or sometimes a high latency mechanism. I've been in the team for four years now, and previous to that, I was in the commodity space at another data company, which I'll refrain from disclosing today. But they've changed their name three or four times in the last five or six years, so maybe you can figure it out. Thanks for having me. And then, Ollie.
OLIVER HEMMING, HEAD OF CONNECTIVITY SALES ENGINEERING – EMEA & APAC, ICE:
Yeah, thanks, Will. So, I'm Oliver Heming. I'm a sales engineering manager for EMEA and APAC. I've been with ICE for almost ten years now. I'm responsible for designing connectivity and hosting solutions for our clients globally. This is a quick overview of ICE Global Network, which is a financial market extranet, which provides connectivity to over 150 different exchange venues and 600 content sources. It's a highly resilient private and secure network, which serves hundreds of our clients globally. Our services cover fiber, wireless, connectivity and hosting, timing services, and historical market data, which we'll come on to later.
WILLIAM WILLMOTT, SALES EXECUTIVE, ICE GLOBAL NETWORK:
To start this off, we wanted to give a bit of detail between the collaboration between EStaR and Ice, and then the Ice Global network. So, estar and Ice have been partners, long established partners, for a very long time now. EStar, as everyone should know, on the webinar, help bring flow to the exchange, whichever the trading participant is active in. So in this case, ice, of course, but other venues as well. And that's the real value in eStar's platform, outside of the trading tool and the ALGo tools they have as well, being able to aggregate these liquidity pools into one place. So, how does the Ice global network interact with Estar? Well, we're actually the extranet, the underlying connectivity, as Oliver touched on just before, in the previous slide, as I touched on the intro, that provides the connectivity to ice, but not just ICE. This is like a massive task that we're here on the team and with Oliver charged with trying to share with people. We don't just provide connectivity to ice. We have over 150 venues on net, specifically in the commodity space, NASDAQ, Nordics and Ex. So we provide the connectivity to these venues in physically redundant sites. So you have two sites that estar access this here. So if one site goes down, they're able to still service their clients, providing the market data feeds, which they handle, and then the execution services, which they provide to you. One side goes down, fails over to the other. That's the relationship there. You've got the front end of Estar, and then beneath that, the plumbing, if you will, of the ice global network accessing these sites. So this is an image I hope most of you might be aware of. This image is the traditional, original trading floor. So this is the naive. This chap here is a man called Peter Truman. Maybe you've seen the image, maybe you haven't. But essentially, as we know, the trading landscape has changed a great deal since then. Since the original pit trading, open outcry form of trading, everything's electronic now. Everything's digitalized. So if you want to take it to the next slide, with the exception of a few notable exchanges, the NAIsI that was just there, maybe the most notable, based in the USA. And of course, the LME, which is based right near this office here. Actually, stones go away. They, of course, still have a pit and a trading floor for exchanges where buyers and sellers meet. But there has been an evolution. And one of the main pioneers of that evolution was Ice founder and CEO Jeff Spracker, especially in the commodity space. He procured CPEX, the Continental Tower Exchange, which is an OTC market, with the objective, the goals of changing that from an analog form to a digital form, so electronifying that market, moving it to screens, making process more standardized, little bit of history there, and basically providing the platforms of giving birth to the platforms like EStar and Webex and the like. There are other platforms out there I'm sure many of you will be aware of. Then this evolution continued, moving to the semi automated space where you even have this on retail platforms. Now, whether limit orders stop losses. I have an anecdote from a client virtue trading, massive systematic trading firm. And they were telling me in their infancy, when they were talking or bringing in algorithms, and this could be counted semiautomatic, they were programming their computer so it could just keep clicking and accepting bid offers for their market making strategy. Not sure how successful that strategy was, but that's what they were telling me is quite a fun story. And then of course, the evolution continues. The birth or advent of flashboys colocation hosting. So you're actually putting these algorithms as close to the matching engines of the exchange as possible, reducing the latency to generate alpha in this way. Rather than trading from the other side of the world, perhaps you're putting it directly next to it to take advantage of arbitrage opportunities, whether that be on the venue itself or between venues. With that, it was a massive investment as well for these firms. These firms, like maybe your citadels, your Hanas, your jumps, there's a massive investment in actually acquiring the space by the the matching engine, but also the hardware that would occupy that space for the operations that would happen there. We're going to touch on during this presentation a little bit how the barriers to entry for that have immensely come down, and that firms who are not just looking to generate alpha, but are also looking to mitigate risk potentially being hit off, or whatever term you'd like to use, can leverage low latency trading environments that don't require entire dedicated team to do that. So traders, like traders on this call, can focus on the trading that they need to do. So to add a bit more context to what I just spoke about there. So there below there's a busy diagram. It's not that pleasant on the eye, arguably a bit of an eyesore, but what you have there is the ICE global network topology map. Maybe not the best piece of marketing by there. Admittedly, this is a capital markets purpose built network.
ANDREAS KAMPER, MANAGING DIRECTOR, E*STAR
So.
WILLIAM WILLMOTT, SALES EXECUTIVE, ICE GLOBAL NETWORK
Each link between each data center is ultra low latency. It's point to point, built with performance, availability and latency in mind. Now, why am I showing this map? Well, look on the left side, you'll have the Chicago area. You've got Sirmap where the ice is matched. The energy products that many people in here will be trading are matched in that data center. And Aurora is the CME. In Europe, you have a couple of major data centers, LD four, FR two. There's vast ecosystems here where you'll have banks, systematic internalizers, brokerage exchanges, FX venues. They're matched here. And a lot of clients or extranets will be connecting into this venue. So you now refer to the caption top of the screen, forgetting to Cermac and Aurora. This is a best case scenario, round trip time for latency. Now, the likelihood of that latency being like that would depend if you're actually in that data center and using our network, other extranets are going to be not as high performant as ours is. And if you're using the Internet, the latency again is going to be much higher, the points deterministic. And if you're in a cloud environment, sorry, if you're in a cloud environment too, you're going to have added latency. So just trying to contextualize the move to more performance based trading and the infrastructure and technology behind it. And now Oliver is going to speak about the solutions that I support, which again, lower the barriers to extra massively.
OLIVER HEMMING, HEAD OF CONNECTIVITY SALES ENGINEERING – EMEA & APAC, ICE
Yeah, thank you, Will. So as he touched on there, we've got the three main colocations for probably.
WILLIAM WILLMOTT, SALES EXECUTIVE, ICE GLOBAL NETWORK
The three most latency sensitive commodity markets, being ICME and DEX URX.
OLIVER HEMMING, HEAD OF CONNECTIVITY SALES ENGINEERING – EMEA & APAC, ICE
So we've designed a low cost entry solution for smaller trading firms, which are looking to get into algorithmic trading, which we actually call our premium shared hosting. So this is for clients that don't necessarily want the large cost overhead of taking a full cabinet and direct exchange connectivity, and instead want essentially the next best thing, which is for a service provider such as ICE global network to offer them a smaller footprint within the colocation space. But still with the latest networking technology, that being layer one switches to achieve the lowest latency access to the market at an affordable price. So we're able to offer that infrastructure in these main commodity colocations as well as others. And really from the client side, one of their next key decisions is their application level and how they're going to interpret the exchange's proprietary market data feed and their order entry system. So EStar does have an offering for that, and offer this in those colocations that we mentioned. There's other third party or ISB software applications out there. Or another option is for the trading firm to write to the exchange's native format themselves. That has quite a bit of development effort and does require what's called conformance testing from the exchange just for their software or algorithm to be validated. The final key element of this setup really is the time synchronization, where once you're moving more into the algorithmic low latency space, being able to measure your execution times relative to the exchanges time is really accurate. It's something that we offer as well. One final advantage of this solution is that we take care of all of the networking complexities, so the network layer is all managed by ICE, and we can assist the client with the hardware for the trading server side. And then we've already touched on the software piece as well. And then finally just moving on to once a client has sort of developed their algorithm, how can they actually validate that before they go to market and begin trading in a live manner? So the sort of industry proven way of doing this is with historical data. So there's various different flavors of historical data. I'm sorry, what's happened there?
WILLIAM WILLMOTT, SALES EXECUTIVE, ICE GLOBAL NETWORK
Apologies.
OLIVER HEMMING, HEAD OF CONNECTIVITY SALES ENGINEERING – EMEA & APAC, ICE
Yes, there are various flavors of historical data. We offer a raw native exchange format historical data service, which allows clients to replay scenarios based on their algorithm of historic events. It's not uncommon for a client to take six months or a year of historical data and then run simulation environments to see how their algorithm would have performed in real-life previous events. This process allows them to make tweaks and adjustments to their algorithm, replay the scenario in that test environment, and analyze the profit and loss of those different scenarios. There's a little diagram there on the right. I won't delve into too much technical detail, but one of the key points is that we insert a receive timestamp. This tells the client when any high-frequency trader or other colocation participant would have received that market update message, allowing them to measure their own reaction time. From that, they can determine if they would have been able to exploit a perceived opportunity in the market. Knowing their reaction time and the time they received the update, they can start to play through those different scenarios. The final point I would add is that for greater flexibility in running these algorithm simulations and accessing the data, we operate in a cloud environment. This approach removes the storage overhead, greatly improves performance, and makes it an agile solution for clients to use.
WILLIAM WILLMOTT, SALES EXECUTIVE, ICE GLOBAL NETWORK
Yeah, thanks for that, Ollie. To summarize, we have the capability to assist with back testing. The barrier to entry for trading firms and commodity traders, who are more performance-conscious and looking to lower their latency, has never been lower. You can focus on your trading; the environment, collaboration with the ICE Global network, and a firm like E*Star take care of the application level, network level, and hardware. This setup allows you to be closer to the matching engines, whether you're looking to generate Alpha or mitigate risk. I believe that covers everything we wanted to discuss today.
MICHAEL OSTENDORF, HEAD OF PORTFOLIO ANALYSIS & DEVELOPMENT & HEAD OF EEM TRADING, PORTFOLIO STRATEGY, E.ON
Thank you.
HOWARD WALPER, CEO AMERICAS, COMMODITIES PEOPLE
Wonderful, gentlemen, thank you very much for that. Really enlightening. So, finally, last but not least, let's turn the floor over to Andreas Camper, managing director of E*Star. Andreas, it's over to you, sir.
ANDREAS KAMPER, MANAGING DIRECTOR, E*STAR
Thanks a lot. I hope you can see my slides now.
HOWARD WALPER, CEO AMERICAS, COMMODITIES PEOPLE
Yes, they are up.
ANDREAS KAMPER, MANAGING DIRECTOR, E*STAR
Great. So, yeah, what I will be covering now are some insights about the problems and obstacles that customers I've been working with for a couple of years now, and pretty much every bigger customer in this space, will face when they're starting to do algo trading. When we talk about algo trading, at least from our perspective at E*Star, we need to differentiate between three different types, which behave a bit differently depending on the kind of algos they are and how you need to treat them.
The first one is what we call assisted trading, which I will go into more detail about in the following slides. The differentiation between these types is based on the integration into a trading front end. If it's working autonomously over a longer time, taking specific decisions, the flexibility, which means how much can I configure the algo versus individuality, what can I do to adjust it to my needs if the basic functionality, the standard functionality, is not sufficient?
Well, the assisted trading part includes things directly built into the view of the traders. Examples are iceberg orders, hidden orders, yes, all very simple, but it also extends to trailing orders, stop orders, stop loss orders, especially if they are not available natively by the different exchanges. Some are, some are not. The main thing is it needs to be integrated into the order entry box. The order entry box needs to adjust to what is needed here, now the plus order, so intuitive usage, deterministic. Normally, we or together with the exchange do the conformity checks. The method documentation is available for internal usage, no real configurability needed for the system itself. But on the other hand, it offers limited customization options and is more or less a standard function of all these functions.
The automation and all these categories that I was just mentioning are overlapping. Automation means that you have standard processes that you want automated, which are not yet acting on themselves other than rather discrete functions. So, as a simple example for these kinds of things, any kind of automations where a back-end system says, “I have a position, please trade exactly that,” or “place that order,” without a real position trading algo, but also things like non-native spread trading, complex spreads trading.
In our solution, I created virtual order books. I can create them in a virtual price feed with a little algo, taking all the legs with the factors into account and presenting. And that is the main advantage, directly in front of the trader where the current market is, he can place orders and then the system just trades the corresponding legs out, which would also work with dark spreads, spark spreads, or clean darks and spark threads and so on. Why is it not assisted trading? Because there's a bit more complexity. How do quantity mismatches are handled? You might have additional factors, let's say, for currency conversion, FX trading, and so on, or additional platforms that you want to trade out. And how do you handle the edge cases, or the problematic cases where you have slippage and so on, but you are still not in the system where one system works completely autonomously, which is the third part, which is real algorithmic trading.
The algorithm automatically, you don't have the need that it's directly integrated into your trading front end. Other than that, you want to see that. But it's not a problem if the algorithm is started and stopped by a third-party system, your own development, a different platform, and so on. But for the algorithmic trading, you obviously, or oftentimes need to do the MiFID documentation and the conformity checks, or some additional conformity checks depending on the market on your own. But you have the full flexibility, you can obviously do whatever you like. And the IP is then with the customer, especially in the discussion that I had over the years with customers. A strong part that all IP for all bespoke solutions is with the customers. Examples are prop strategies, complex market strategies, and so on. And the complexity obviously is with testing, conformity checks, and so on.
That being said, in addition to what the previous speakers already mentioned, at least from our perspective, you have already, if you have more than one trader, some problems that you have to solve. You have portfolio managers, market access traders, asset optimizers, and maybe a sales team. They all need either direct or indirect access to the different markets, to all the different exchanges that were already named there. And then you have the problems of safe trade prevention, Chinese walls, maybe you need a compliance view, control, visibility, and so on. And now, taking Algos into account, and maybe even further out, sub-customers, or even algos for sub-customers, depending on how you're structured, you will run into the problems that the colleagues from Uniper, E.ON, and so on already mentioned. You need a platform to structure your access to the external market. There is no way around it. Otherwise, you run into the problems of self-trades, right management, Chinese walls, emergency controls, and so on, and MiFID registration.
From our perspective, it's a marketplace that you need in between, that you can completely control and use in order to do all of that. And just in differentiation, a marketplace or an internal market, does not only mean a netting functionality between two orders which might fit from an Algo and a human, it goes way beyond that and allows real structuring. Your access, sleeve trading, and all these kinds of things might come into play. And the algo, at least from our perspective, is just one additional player to the market, which adds to the complexity. People get on their feet, step on each other's feet. The Algo placed an order.
The human needs to place an opposing order, but cannot, because the algo is in its way, and vice versa. All of these problems need solving. Otherwise, you lose quite a lot of flexibility and opportunity, and sooner or later, maybe not in the beginning when you start algo trading, but definitely in the long run, if you have a couple of traders and so on, you will run into these problems, especially in distributed environments.
What was already mentioned, all the different markets and all the different platforms, you want normally to connect natively to these different markets, because they offer additional functionality, maybe even additional order types. You have the latency advantages if you connect natively to the exchanges, but especially if you would try to implement them on your own. That comes with complexity. All the different markets have different interfaces. They all behave slightly differently, even though in general they behave similarly. But the details are important in that regard. The liquidity for most of the markets is split between different platforms, and therefore you want to cover them all. But not lose either latency or functionality from these different exchanges. And if you implement them, we have implemented more than ten. Pretty much every month one of them has an update which you would need to implement.
That being said, the advantages not only or the functionality that you need is not only the market phases and then leaving the customers alone with the problems that arise from the native interfaces, but also providing convenience functions like the internal netting, but now making the example with Epex, EDPA, and Norput spot three intraday exchanges. Yes, the prices in the XBIT phase are the same, but in the local phase, for example, they are different or could be different. So nobody is paying you to build this little algo snippet which switches between the different markets depending on where the best prices are. And definitely everybody will be sooner or later be able to achieve that. But it's slightly more difficult than you would assume. And these kinds of convenient functions we can offer to the customers to say, “I don't care where I get it from, the system should just pick the best price and I don't want to care about if it's an EpeX output spot or EDPA or ICEX or NASDAQ OMX in that regard.” And on top of that, I need internal netting functionality, the market functionality, so that I can write algos which are only specific to the tasks that they have. I don't want to overload one Algo, which is maybe a position trading Algo, to cope with internal netting functionalities. That is not the task of the position trading Algo and it should not be bothered with it.
So from a requirements perspective, once you start really looking into it and doing it, you have the problem with the different programming interfaces and the different tools that you want to use. So we got in the past a lot of requests. One customer wants to use Python, one customer wants to use C sharp, one customer wants to use you name it. And also we tried in the beginning to force every customer to one language which didn't turn out and had to go, which is sooner or later for bigger customers is a need. Different programming languages need to be used. Some solutions, and especially based on the requirement on the libraries that are available for the different languages drive. Oftentimes which language is used for a specific market or for a specific algorithm. So the availability and the permission or the possibility to use all the systems that you have, like Stefan said in the first meeting, they are using Deltix and connect that to our solution. It's how you are structured internally that should determine what tools you need and not the other way around and saying this is what you need to use, because this is the holy grail of energy trading.
Coming back to what Will and Oliver from ICE said, the latency gets sooner or later one of the most critical parts of your Algos. If you have a strategy, you want to quickly roll it out to more and more locations, to more and more products. And that forced us as well. When we started our ZAS solutions in 2018, we only offered Frankfurt. In the meantime, we are offering colocations in London and in Chicago, which obviously comes in a way you need to decide where you want the Algo, and if you have something which is trading ICE and ex, you can't win them all. But that is often not the case that you can better choose where to place the Algos, either in Frankfurt, in Chicago, or in London, depending where the exchanges are. And that gives you quite, or can give you quite an advantage against other players in the market. As Oliver and Will have shown, the latency between the pure geographical latency between Europe and the US is over 80 milliseconds, and that is significant if you can't get around that.
So obviously also what the requirements were and often came to us, which we had to fulfill, is the decision which is not as obvious as you might think. Do we want live data or historic data? Where do we get it from? Do we want snapshots or complete order books, or do we process tick data? And is it really worth the effort for all of the Algos, or do we have slower trading Algos? Do we want aggregated order books across all the different venues, or do we want separate order books with implied without implied? That is also something along the discussion with that and how the developers then with the customers try to solve or touch on the system is quite a task which can take, in the beginning quite some time too far.
So what I wanted to present here is pretty much, at least from my experience, all the problems that you will, sooner or later, at least in bigger companies, face when you're starting Algo trading, not necessarily in that order. We have customers who started first with Algo trading and then expanded to desks and the other way around. But the internal structure of your own company will be something that needs to be organized. The market specifics, depending on which markets do you take into account and in which way, the languages and libraries you need, the technical setup and the latency. And a big task, especially on the long term markets, will be risk and compliance, and the discussion how MiFID and so on is ensured in the current times. And lastly, the data requirements that you have. And that is pretty much what E*Star and before Exeter is now doing for over 15 years, helping customers with out-of-the-box applications, but the possibility to deeply integrate into the needs and from standard products even move to bespoke solution as an extent to what is really needed by the different customers. Thanks a lot.
HOWARD WALPER, CEO AMERICAS, COMMODITIES PEOPLE
Wonderful. Well, thank you, Andreas. I really appreciate your perspectives. We're going to launch into our panel discussion portion of this program, but before we do, first of all, we've seen quite a few questions come in, so please do continue to ask any questions you have either about the material that was just presented or on some of the topics we're raising in the panel discussion. We also have a short poll that we'll be pulling up here. Feel free to start filling it in. We'll launch into our panel discussion concurrently. But yeah, please contribute your feedback into these polls and we'll share the results in just a little bit. So launching into this, by now you should all see a pop-up box that came up with some of the first poll questions. I believe if you fill those out, it just continues to scroll through to the next one. But anyhow, let's launch into the panel discussion. We have a few questions here for the group. This is one primarily for Michael and Andreas. What distinction do you make between assisted trading, automated trading, and algorithmic trading? And this is for Michael and Andreas.
MICHAEL OSTENDORF, HEAD OF PORTFOLIO ANALYSIS & DEVELOPMENT & HEAD OF EEM TRADING, PORTFOLIO STRATEGY, E.ON
Yeah, so maybe I'll go first, because Andreas has shared already during his slides the distinctions they do at their end. E.ON is a big corporation, so what does a big corporation first do? We write a policy. So we made a policy and we differentiate between three types of automated and algorithmic trading technologies. Automated is, for us, everything where a trader has, in fact, predetermined all the relevant parameters. It's a fixed formula: if-then-else. So there is no leeway for the tool itself to make a decision on its own algorithmic trading. There, the trader does not have full control of all parameters, so that means it's more sophisticated. There is a computer algorithm behind which has the freedom and autonomy to determine at least some of the parameters. It doesn't mean all, but at least some. And the third bucket, which we have, is high-frequency algorithmic trading. And the main reason is that any high-frequency algorithmic trading in financial instruments will automatically trigger MiFID II licensing requirements. That's why that is, let's say, the highest hurdle to take. But that is how we differentiate between these different types. Assisted trading – that's something where we're just using an iceberg order or something on the screen. That's nothing which we really take as an algorithmic topic.
HOWARD WALPER, CEO AMERICAS, COMMODITIES PEOPLE
Got you. Thank you for that. Another question for Andreas and Michael. How has the role of human traders sort of evolved with the advent of algorithmic trading? I guess when it comes to upskilling, what sort of skills are now essential for traders in this environment?
MICHAEL OSTENDORF, HEAD OF PORTFOLIO ANALYSIS & DEVELOPMENT & HEAD OF EEM TRADING, PORTFOLIO STRATEGY, E.ON
Yeah, so maybe I take that one also first and then Andreas can contribute. So from my perspective, currently a trader is a brain and executor in the end. They look into portfolio and market data, they analyze the energy markets, have the hedging framework at hand, and then they see also their performance, so they execute it all. But they are developing more into a quant modeler and supervisor because all these tasks – monitoring these activities and, let's say, the rules and framework around it – is something which an algo can do. But let's say there is still a human brain needed to feed that algo. In the end, the trader will be more of an optimizer, a supervisor, and also, let's say, a quant-trader developer of that algorithm.
ANDREAS KAMPER, MANAGING DIRECTOR, E*STAR
At least I give the floor to Stefan.
STEPHAN VAN AAKEN, SVP DIGITAL TRADING DEVELOPMENT AND OPERATIONS, UNIPER
Yes, I gathered as much. So I can completely see it the same way as Michael says, we are actually seeing that very concretely in our areas already, that the actual traders develop more in the direction of developers. But then we see in particular, with what Michael, what you said earlier, with all the regulation and the rules around it and making sure that you're compliant, we are moving in the direction where we only pretty much allow our traders to prototype at most Algos and then have our real professional developers develop, test, and so on, the actual algos, because to really get them through our risk department, which I, of course, very much love, and many other parts, that would be too much of a hassle. So the traders should come up with the idea, test it a little bit, and then before we actually let it loose, we let the pros look over it to make real software out of it.
ANDREAS KAMPER, MANAGING DIRECTOR, E*STAR
But that is pretty much also where I'm coming back a little bit to the last question, but also where the human traders come in; they have the ideas, they want to execute these ideas, and sooner or later they need assistance, which is more and more technical, requires more data and so on. Yes, a simple iceberg order is not, or the market has decided that it's not MIFID relevant, but it starts with stop orders, stop loss orders, trailing orders and so on. It can get more difficult. But regardless of how simple the Algos are, sooner or later you will come to a point where a normal trader is just overwhelmed with too many markets, or the other way around, that you can give the trader more tools to do the same good work on more markets at the same time, without them spending the whole day putting 200 orders manually into the different markets.
HOWARD WALPER, CEO AMERICAS, COMMODITIES PEOPLE
Excellent. Any other comments on that? Well, moving right along. So the world, as we know, has become a very interesting place over the last few years. What are some of the implications of geopolitical events and market volatility on algorithmic trading, and what should firms be doing to adjust their strategies? And I guess anyone can pick up on this.
MICHAEL OSTENDORF, HEAD OF PORTFOLIO ANALYSIS & DEVELOPMENT & HEAD OF EEM TRADING, PORTFOLIO STRATEGY, E.ON
Yeah, starting off. So, from our perspective, in 2021, we decided to have an automated and algorithmic trading policy, not only to have a policy but especially because we see that with increased speed and also the complexity of algorithmic trading and all these market implications, that was even before the last year. We need to be able to manage and control potential errors. So this policy is mainly there to outline how do we stop, for example, the algorithms from doing something which would not be, let's say, what it should do, because it's still a model which is running, right? So you can't have everything coded in, so you should have, let's say, some trigger levels, stop points, controls around it, and monitoring. And we even have a decision on a case by case basis for all algorithms and automated tools going online, if they need to be supervised 24/7 if they're running for 24/7 or not, what the specific controls around them need to be. So there is a very close interaction, for example, with our risk team on this, that we don't run into a situation where market volatility is just, let's say, killing us because our auto trader was just blindly trading, as we programmed it before.
HOWARD WALPER, CEO AMERICAS, COMMODITIES PEOPLE
Anyone else want to weigh in on that?
MICHAEL OSTENDORF, HEAD OF PORTFOLIO ANALYSIS & DEVELOPMENT & HEAD OF EEM TRADING, PORTFOLIO STRATEGY, E.ON
Stefan raised his hand as well.
STEPHAN VAN AAKEN, SVP DIGITAL TRADING DEVELOPMENT AND OPERATIONS, UNIPER
Not to anyone's surprise, I can say it's pretty similar with us. When you think particularly about the recent year or months we've seen, it's just super important that the traders are really involved in the development, fully involved in the development of the Algos, and they really understand what the Algo does and what the shortcomings are. Because let's face it, in the end, each algorithm, what you program is making some simplifications, some assumptions, and so forth, and if they break in the market and you don't understand that your algo is no longer applicable, then you've got a problem. That's where the safeguards that you, Michael, were just mentioning, come into play; the risk guards are the most important anyway. They're the kind of last line of defense. They're the ones cutting in and saying, ‘That's it.' But it's even better if the trader upfront knows, ‘Okay, there is something coming up. I know that the Algo can't handle this.' And one other thing that, obviously, in this particularly interesting market situation is what was mentioned before. We all do that with our algos. We test them against historic data. Now, that doesn't suffice anymore, right? If you see these crazy markets, which you haven't seen before, that scenario is not reflected in the historic data, which presents a double problem, in the end, it's in the past not reflected. But now, looking forward, we collected all this historic data with these high volatility values. What does this mean for the future? Because that's also not likely to continue. So it's an interesting challenge.
HOWARD WALPER, CEO AMERICAS, COMMODITIES PEOPLE
Before we move on, would anyone else like to comment on that question? All right, let's move to the next one. And that's about compliance. How do you communicate with your compliance team on algo trading matters? How do you ensure that the regulatory framework is being followed?
MICHAEL OSTENDORF, HEAD OF PORTFOLIO ANALYSIS & DEVELOPMENT & HEAD OF EEM TRADING, PORTFOLIO STRATEGY, E.ON
On our side, for every single new automated or algorithmic process, we run a so-called new authorization process. So every department, including compliance, is asked to give their feedback, sign off that this can be executed. So we are in very close collaboration with compliance, but also all the other functions, which in the end need to say that it's ready.
MICHAEL OSTENDORF, HEAD OF PORTFOLIO ANALYSIS & DEVELOPMENT & HEAD OF EEM TRADING, PORTFOLIO STRATEGY, E.ON (continued)
…to go live and that they have all their business processes ready so that we can run that. And as I said, on financial instruments, if we trade them via an algorithmic or automated trading tool, there are even more topics which we discuss very closely with our compliance teams, because MiFID II is then affected or could be affected. So that is where we are in very close cooperation with them.
HOWARD WALPER, CEO AMERICAS, COMMODITIES PEOPLE
Any other comments on that?
ANDREAS KAMPER, MANAGING DIRECTOR, E*STAR
Stefan, I always give you the…
STEPHAN VAN AAKEN, SVP DIGITAL TRADING DEVELOPMENT AND OPERATIONS, UNIPER
Yeah, as expected, Michael, we are having the same rules, and I think that's probably what many companies that are on the more careful side are doing. When we initiated our Algo activity, we went that way because we had the tendency to make no compromises on compliance. We wanted to make sure that, particularly in the early stages when we started with algo trading, we didn't have any compliance issues arising. And it was a bit new to everyone. So that's why we put this special emphasis there, and we actually appointed a person in my teams to specifically take care of all the compliance and rule adherence, ensuring that each individual algo is properly vetted through the organization. To assign this responsibility to a trader could risk that it's done with varying quality standards. The last thing we thought we could need is to have any significant compliance issue very early in our algo trading journey, because people were already skeptical enough. It's like, ‘Oh, this algo trading is going to do something…' That's why we decided to have a dedicated person who was then working very closely with the various departments like risk, legal, and so on, ensuring that the same standards are applied. That approach actually worked quite well.
HOWARD WALPER, CEO AMERICAS, COMMODITIES PEOPLE
Go ahead.
MICHAEL OSTENDORF, HEAD OF PORTFOLIO ANALYSIS & DEVELOPMENT & HEAD OF EEM TRADING, PORTFOLIO STRATEGY, E.ON
Maybe just one final thought. As you can see, it's only been seven years since Uniper and E.ON split off, so we seem to have some shared DNA still.
ANDREAS KAMPER, MANAGING DIRECTOR, E*STAR
Maybe to add from my side, from a pure technical perspective, and also considering the new requirements we've got, transparency about all order actions for all the different markets is crucial. And having a feed that shows not only algo activities but all activity which is…
Mark Copley, CEO, EFET – European Federation of Energy Traders 
Marc Merrill, President & CEO, Uniper North America 
Nicki Dalgaard, CEO, Disruptive Biotrading 
Andrew Singleton, Partner & Expert in Energy & Commodities Business & Technology Transformation, Baringa 
Catherine Newman, CEO, Limejump 
Eddie Tofpik, Head of Technical Analysis & Senior Markets Analyst, ADM Investor Services International Ltd. 
Latif Faiyaz, Head of Energy Trading & Strategy, Northern Gas and Power 
Paul Monk, UK Power Specialist, Catalyst Commodities 
Nick Speechley, Strategic Projects Manager, Octopus Energy Generation 
Jeremy Plester, Commercial Manager, MetDesk 
Trevor Neil, Managing Director, BETA Financial Limited 
Clive Lambert, Chief Technical Analyst, FuturesTechs 
Nicole Elliott, Technical Analyst & Private Investor 
Mark Randall, Trader Performance Coach / MD & Co-Founder, The AlphaMind Project 
Steven Goldstein, Trader Performance Coach / MD & Co-Founder, The AlphaMind Project 
Mark Meyrick, Head of Smart Grids & PPAs, Ecotricity 
Robert Carver, Systematic futures trader and author, Independent 
Chris Sass, Founder & Host, Insider's Guide to Energy 
Kelly Vaughan, Director of Digital Strategy, Mercuria 
Peter Styles, Executive Vice Chair of the Board, EFET 
Harry Huang, Managing Director Gas Trading, Uniper Global Commodities 
Alejandro Navarro, General Counsel, Energopole 
Post comments (0)