#175 Inside Algorithmic Trading with Anthony Markham, VP, Quantitative Developer at Deutsche Bank

Things are going to go wrong, I think is the attitude to approach it with things are going to go wrong, you are going to be on the wrong side of a trade. At some point some sort of global event is going to happen that is a systemic risk to the market. Covid black swan events, things like that, that almost no one is going to predict. Right. You're going to be in a bad position. So, so the key is really to have proper risk management so that if these events happen, you can manage it yourself. You're not super exposed.

Hi Anthony, thank you for joining me on the show.

Hi Richie. No worries. Thanks for the invitation. I'm very excited.

Excellent.

Yeah.

So I think to begin with it's been really nice to have a bit of context on like what is algorithmic trading? Can you tell me what it involves?

Yeah. So it's very much a mystery to a lot of people. I guess the idea behind algorithmic trading, quantitative finance, all these sort of terms that get thrown around is essentially trading automatically. So trading with computers. So instead of a person executing a trade themselves, like me or you might, it's a computer doing all that. So the computer is going to do all the logic behind that on the back end and execute the trade itself without a human exactly telling you what to do. So, so it's all programmed. So these strategies are all programmed into the system, data driven strategies, all developed by people called quants. Okay.

And maybe can you just expand a bit on like which parts of this are done by computers and like where the actual human involvement is.

So humans usually design the strategies. So humans of course have complete oversight over this sort of stuff. I'm sure we'll talk later about the risks involved in this, but you can imagine letting a computer just go skynet and trade can be very, very risky. So that's not the case. So humans often design the strategies. They'll do all the analysis behind it, they'll be examining the data, they'll be testing strategies on historical data and all these sorts of things. Whereas the system itself will just execute the trade. So it really depends on the strategy. But the system might pick when to trade, what time to trade, it might pick a profitable time, it might ingest data from another source and infer something from that and infer that it's a great time to trade based on that. So yeah, it's very data driven. Okay.

So it sounds like the humans are involved in maybe the more creative part of things and the oversight and then the computers do everything else.

Very much so. I mean of Course, the computers are programmed. You sort of need software engineers and data engineers and all those sorts of typical roles to build out these models. And these models are very mathematical, very statistical. So maybe creative is the wrong word. But yeah, there is an aspect of creativity and that you need to be understanding financial markets as a human and you need to be sort of designing strategies around that. Computers, of course, can't do everything.

Okay, and can you tell me how this algorithmic trading is different from maybe more traditional or standard trading and investing?

I guess the speed at which it happens is probably the biggest difference. Algorithmic trading trading can be at any pace. You could execute one trade a day, for example. With algorithmic trading, the computer could decide, we'll execute one trade a day, or you could execute 1,000 a second. And so you can typically break that down into mid frequency trading or high frequency trading are sort of some of the terms you hear where that refers to the frequency of trades and at that sort of top level, the high frequency level. Yeah, it can be in the thousands of trades a second. So very, very high throughput.

Okay, so this is really something that.

A human couldn't do because no one's going that fast. Exactly. A human could never do this sort of stuff. And so it's a very different sort of investing. You think if you're me or you as a retail investor, you think we might buy a stock because we believe in the company. We might say the new iPhone comes out in a month and therefore the price of Apple stock is going to rise. Because I believe in Apple as a company, I believe the iPhone is going to be awesome. That's value investing. So we're investing in the value of the company and that product that they're releasing. Whereas algo trading strategies that people design, it's very hard to do that if you're trading a thousand times a second. You can't sort of trade based on value. So you're often on a much shorter time frame. So you might buy something and sell it 10 seconds later for a very small profit for a $0.01 sort of margin or a spread there. So very different sort of investing at its core.

So it sounds like there are these different strategies and maybe like this value investing or the sort of the let's make a little bit of money in 10 seconds are sort of different approaches. Can you talk me through what the analyses look like in order to accompany these strategies?

Yeah, and of course a lot is done there. So I suppose maybe the biggest goal obviously is to generate profit from this, from the Companies that do this. But these companies are also what I call market makers and the term is very literal. It means to make a market. So you're acting on both sides of the trade. So you might offer liquidity to the market, meaning offering volume to the market by being on both the buy and sell side. So in that example I gave of you buying some Apple stock, we would probably buy that in reality from a market maker, from a high frequency trading company. They'd be the ones on the side of the deal that we'd be buying from. Just because they're constantly trading back and forth, buying and selling, providing that sort of activity in the market. And so the activity certainly is there and the analysis is there in constantly being a part of the market. But then of course risk management is sort of a key part of the analysis as well on these strategies. And maybe we'll discuss it further along. But there has been cases in the past, very famous cases, that some listeners might remember the flash crash of 2010 where the US market lost about a trillion dollars in a 30 minute period and then rebounded straight away based on some quite shady high frequency trading, some bit stuffing and spoofing. So it's a very competitive space and so a lot of analysis is done. And that's why you'll see people in this space, typically with maths background, stats background, because that's what you really want. You want people who can sort of solve these problems and sort of figure out a solution.

$10 trillion in like 30 minutes. That's a number that I find very difficult to imagine.

That's a lot of cash there, 1 trillion. But yeah, yeah, listeners should look it up. The flash crash of 2010. Yeah, pretty incredible.

I find it interesting that it's not just, oh, let's see how much profit we can make or how much money we can make as a goal that you also have these ideas of like, well, oh, we need to do things like mess about with liquidity or mess about with like how do we minimize risk? So it does sound like there are a few different goals there involved.

That's right, yeah. I mean you certainly can't just focus on making money. You've seen all sorts of examples of companies that have had bad risk management practices. They're just throwing cash around and they've been very irresponsible with their investing. As a result, they've gone bust, they've caused instability in the market. So yeah, so risk management is a key aspect of it.

Okay, that's actually a really interesting idea. Like are There different skills needed for risk management versus some of the other goals. Like the straight up training.

Yeah. Generally speaking, people who are in a risk management area have more of a finance background, so they might have an undergraduate degree in finance or experience in risk management. Whereas people in my role for example, are mostly software engineers. So with a very technical background in maths, engineering, software engineering. But there is still a lot of work on model building on the risk management side as well. There's a lot of models that are built to sort of model exposure and model risk. So it's a combination. You'll have risk managers and analysts who will look at stuff from a finance point of view and you'll have software engineers like myself who will build out models that can analyze that data.

That's pretty interesting, the idea that you need both that sort of finance background and you need like the software engineering background. And also you need some kind of math or stat background as well. In order to make this work. I'd like to go into a bit more detail on what the data looks like and what sort of techniques you use. So I guess maybe let's start with the data. Yeah, yeah. What's involved?

A lot of data. So yeah, so certainly you can imagine if you're trading a thousand times a second as sort of an upper tier example there, that means you need data to get to the millisecond, essentially. Right. You need constant updated prices from exchanges and other people in the market. You need to know what's going on. You can't be entering a trade blind not knowing what the price of something is or what the volume is or what the market's doing. So you can sort of break it down into two types of data. You've got historical data which would go back decades, and you'd store that in a database somewhere, you know, in sort of cold storage. And that's often for back testing strategies. So you'll be running strategies that you might come up with on that previous data to see how it would have performed in the past. And you often hear that sort of financial advice. Past performance is no guarantee of future results. And it's of course true, but it's valuable data. You should be testing out as much as you can to get as much success as you can. And then you've also got the real time data. And that's where maybe some of the engineering challenges really come in, where you've got these data feeds and data pipelines that are feeding in pretty extreme amounts of data from a lot of different data sources. As Quickly as they can.

Okay, so having historical data on what the price of something is seems pretty important. What other data sources do you use in addition to that?

Yeah, I mean, so obviously having the price of the asset itself is pretty critical, but you'd also look at, you know, the interest in that. So the bid and ask spreads, you look at the volume historically, how much volume has been traded. If there's a sudden increase in volume, what does that mean for the price? You know, if a lot of people are buying, all of a sudden, the price might go up. You can sort of get the correlation. Efficiently with the lowest latency possible. That's a really big challenge.

And in terms of the techniques you use, it seems like time series analysis may be the obvious thing, but can you talk me through a bit about, like, what techniques used in these analyses?

Yeah, that's right. So time series data is what we deal with. We deal with, as I mentioned, a lot of data, often down to the millisecond, so broken down by millisecond timestamps. So a lot of data that you can plot up, you can visualize, and then you can do some sort of analysis on that that really comes into trading strategy. So, for example, you might use some NLP to look at natural language processing, and you can look at what's happening in the market as a result of that. You build all sorts of models off the back of this data. And generally speaking, there are so many models that you can build to pivot up this data in some way and to display something. You'll have all sorts of different views on the market from a risk point of view, from a trading point of view, from a wider perspective, there's a lot of analysis that can be done to look at that data.

I remember seeing a talk on financial analytics a few years ago where they were saying that actually, although supervised machine learning is an obvious thing to use for making predictions, it's very hard to make it work in the case of stocks. And I'm wondering whether that's something you've experienced or whether that's old hat now.

Yeah, that's very much the case. I haven't heard of any companies really leveraging machine learning for actually trading. So in other words, using a model that would say this time of the year, Bitcoin's going to double in price or something. It just doesn't take into account enough factors. Typically speaking, if you build out something like a deep neural network to look at this, it's often too hard to interpret. I think one of the Big issues with deep neural networks is that it's very much a black box. And I think when it comes to articulating strategy, you need to be able to understand what's going on there and you need to bring it back to the fundamentals. So market fundamentals. And I think with any price predictions using machine learning, you often see it online, you see the sort of tutorials, oh, we can predict the price of Bitcoin, it's going to double quickly. Buy it now in 12 months. Not the case. These models often just, yeah, you can't rely on them and you wouldn't want to use them in an active trading environment.

So I'd also like to talk a little bit about technology and I know you've been using Julia as part of your strategy, so can you talk a bit about what you're using for data analysis?

That's right, yes. I love Julia. I think it's a great language and I've always seen it as the best of both worlds in that, you know. I really hope there's more adoption of Julia in the next five to 10 years because it really hasn't see much adoption in industry in the past sort of few years, although it is growing. I see it and the reason I use it is I see it with the simplicity of Python. So relatively simple syntax, pretty easy to understand. It's not complex like C or something like that. It still maintains a lot of the concepts like data frames that you see in Python. Plotting packages are quite similar, it's very similar there, but it's significantly faster. Right. It's got type declarations, takes advantage of llvm, it's a just in time compiler. It's not, you know, the global interpreter lock isn't a thing in Julia like it is in Python. So it's a significantly faster language. So you're getting the speed of that low level language, but you've got the simplicity of something like Python. So that's obviously one language. I think one of the reasons is you have a lot of people from diverse backgrounds, so you of course got people like myself who are software engineers. Like once we know C, Python, you know, it's our job to write code. But there are also people in financial institutions who have finance backgrounds, Right. They will have done a finance degree, they will have done something like that and they won't know a programming language, they certainly won't know C and they shouldn't, it's not their job. So they need a language that's relatively simple to learn, that they can just plot up some data, visualize some data pretty quickly without taking them days or weeks. And Python, of course, is really the industry standard. There's, it's simple, the support is amazing. Lots of packages, you can kind of do anything quickly. But Julia, I think is a great alternative again. So for those reasons that I mentioned, I think it's great for an analyst who has no software engineering experience to pick up and quickly just work with big data sets.

That's pretty interesting. It certainly seems like, well, Python is the most popular language for data analysis right now and so the fact that you've got this mix of the two is really interesting.

That's right, yep.

Are there any particular areas or use cases where you feel that Julia has the advantage over Python?

It's definitely the speed that I mentioned. Yeah. So working with large data sets, which is of course sort of a key part of quant finance and Argo trading, the data sets are huge. We're looking at years of data, decades of data, and it could be broken down by millisecond, a huge amount of data. So that's where Julia definitely has a big advantage. I wouldn't say that. It's certainly better than Python. You know, I wouldn't want to sort of tell everyone, go, use Julia, get rid of Python. Python has its advantages. It really does. The support out there, for example, Python is a million times what you'll have for Julia. You know, Julia is still very much in its infancy in terms of the online support out there, in terms of tutorials, courses, resources, packages, that sort of stuff. Especially for a relatively niche field like quant finance, where Python has some great packages. Julia doesn't really have that. So Julia has that speed advantage, but it's not the be all and end all. And that's why you've seen most companies stick with Python.

Okay. So having that support and the broader ecosystem is very advantageous. And there's that trade off then between ecosystem and speed of your code running.

Exactly, yeah.

And you mentioned that you also have some finance people who have to get involved in analysis and they need something really simple. I'm curious as to what happens when you start using different languages and you've got teams with different levels of technical ability. How does that work?

It's usually pretty segmented. So for example, a development team like myself would be separated from a team of analysts, or you'd be working on very separate work. A developer would be building out models on the back end, you'd be writing C, something low level to actually work on the backend, whereas an analyst is very much working on pivoting of that data. So they're going to take the data from a database, they're going to do some analysis on it, make some plots, make some charts for a report, that sort of thing. So typically there isn't that sort of disconnect where a developer is talking to an analyst about something in C and they have no idea what's going on. You know, that's very much segmented in that everyone has a clear role there.

Okay, yeah, I can imagine you don't want someone without any technical background trying to comment on your C code. It's not going to work out.

That's right. Okay. All right.

So before you mentioned about risk management and risk analysis being an important part of the job and it does seem like there's a lot can go wrong with algorithmic trading. So can you just give me an overview of how you go about managing risks?

Yeah, that's right. I mean, I brought up the Silver flash crash of 2010 before. Things are going to go wrong, I think is the attitude to approach it with things are going to go wrong, you are going to be on the wrong side of a trade. At some point, some sort of global event is going to happen that is a systemic risk to the market. Covid black swan events, things like that, that almost no one is going to predict. Right. You're going to be in a bad position. So the key is really to have proper risk management so that if these events happen, you can manage it yourself. You're not super exposed. Funnily enough, Data Camp actually has a really good course on quant finance with a few risk manager models that are actually mentioned there. And one from that course is the VAR model. The VAR or the value of Risk model. And that's a pretty key model that is sort of the base of any risk management, risk management strategy. So yeah, it really is just very similar. Again, a lot of mathematical statistical models that you can use to constantly track your position. But I think where you see companies go wrong is often attitude. It's not a lack of technical skill, it's not a lack of the models, but it's attitude to their investment and that's where they go wrong, get themselves in trouble.

That's interesting that the problem isn't necessarily just the statistics behind things, it's a cultural thing. Could you maybe expand on that a bit?

Yeah, I mean, you think of some of these collapses that have happened previously. I suppose FTX is maybe a very global case that every listener will be aware of. FTX had incredibly smart people. They were incredibly smart developers. SBF worked at Jane Street, a well known sort of quant finance firm that sort of hires very intelligent people. He was a smart guy. But it was an attitude problem there. Right. Very clearly they were doing things that was illegal. It wasn't that they didn't have smart people, it wasn't they couldn't build out these models, it was that they didn't want to and they were just doing illegal things. So, yeah, so it's very much an attitude to risk management that is really the foundation rather than your model itself. That's the stepping stone.

Okay, so in that case, it seemed like because there was illegal activity, they were just like, okay, we believe what we're doing and we're just going to go for it regardless of anything else.

That's right. Yeah.

I think that's maybe not the case in a lot of organizations. Like, you're probably still doing legal things, but they're still vulnerable to treating risks badly. Are there any sort of cultural warning signs you can think of where you think, okay, maybe risk management isn't being taken seriously?

Yeah, I suppose most companies have a compliance department. You'll have compliance essentially reviewing trades and even trades themselves be reviewing trades. So a very common thing is post trade analysis where you will look at your trades and as I mentioned before, through machine learning, you should understand them. You should understand why those trades happened. If you lost money on a trade, that's okay, you're going to lose money on a trade every single day. But why did it happen? Why did you enter that position? Why did it come out badly? I think that constant attitude of self reflection is really important. And I think if you see that absence of that, if you see that absence of not really caring or that absence of not really wanting to reflect and debrief, then that's a concern.

So I find this very interesting. So at Data Camp we have this idea of intellectual honesty where if you do something silly, you got to acknowledge that you've done something wrong.

That's right. Yep, that's right.

So, yeah, that's absolutely fascinating that you do need some kind of review process and you do need to acknowledge where you've made mistakes in order to stop them in the future. Just on the subject of risk, how, how is risk dealt with by analysts? Is it a separate team that's involved in thinking about risk or is it something that analysts would have to worry about themselves?

Yeah, it's generally a separate team. You'll have a team of risk management Professionals who will do that and then developers who are geared towards building risk models as well. So people who are pretty professional at that. It's always something that should be front of mind to everyone. And all of the reporting is pretty open when it comes to risk management. So a lot of the reporting is seen by everyone, a lot of the metrics that attract risk management are seen by everyone. So that people can have a pretty transparent view of what's going on. But yeah, there are specialist teams for this sort of thing.

And on the technical side of things, can you maybe tell me a bit about like what the statistical techniques are to measure risk or are there any kind of analysis techniques there?

Yeah, of course, yeah. Again, quant finance is all about maths and stats. That's the core point I would give people. Right. It's all about maths, it's all about stats, it's all about a quantitative approach to this sort of stuff. So you think of all the sort of statistical measures that are out there, you can probably apply them all to a trade to a portfolio. The model I mentioned that was covered in the data gap course, the value at risk model of R will essentially look at a whole portfolio. So rather than looking at one trade that'll happen in a day of a million, it'll look at the entire portfolio and it'll try to estimate the maximum potential loss on that portfolio over a certain period of time with a certain confidence interval. So you could say, for example, a one day, 95% confidence into a VAR of a million dollars means there's a 5% chance that portfolio will lose more than a million dollars in that one day period. So it's trying to put a confidence interval on something. It's using statistics to say, okay, we're 95% confident that we're not going to lose more than this today. But then for individual trades there's all sorts of metrics as well. And of course that's taken into account before entering into a trade. You've got standard deviation and variance and pretty simple things like that. You've got the beta of a stock, so it means how sensitive the stock is to the returns of a benchmark. So a market benchmark, like a market index, you've got different ratios, like the Sharpe ratios, risk, risk adjusted returns, so it adjusts returns for risk and all sorts of methods. It's the sort of thing you could talk about forever, all sorts of analysis you can do there.

Okay, so it just seemed like there's a pretty comprehensive suite of possible statistics and possible Techniques to use.

That's right, yeah. There's never any sort of shortage of that.

And I know like there's always people who want to know about how to get a job. So I'd like to talk a little bit about how you get into this field of algorithmic trading. So maybe can you talk me through like what are the different data roles that are available?

Yeah, definitely. I think the structure is very similar to a normal tech firm. Some of the titles might be a bit different. My title is quantitative developer. It really just means software engineer. But you know, different sort of titles but yeah, very similar to a normal tech firm. So you have data engineers who build their data pipelines, manage the back end databases, things like that. The data scientists, just like you will at any other tech company who work on machine learning models and build models and things like that. Data analysts. From on the analyst sort of side of things, you'll have developers like myself, quant developers, software engineers, same sort of thing who are just traditional software engineers. There's a lot of roles. It's really very much the same as any sort of standard tech company. It's. Yeah, everyone sort of has a pretty diverse skill set but you know, even the recruitment process is quite similar to tech companies these days. It's very hard to find a difference there.

That's interesting. So you really do have quite a wide range of choices, it seems, of background. So you can come either from a finance background or from that sort of STEM background or maybe from a. Yeah, software development.

Yeah, yeah. Okay.

Are there any particular qualifications you need to get into the field?

Yeah, I think the one thing that sets sort of these sorts of firms, quant trading firms, apart from maybe a traditional bank or an investment bank, is that there is a big focus on those quant methods, of course. So any quantitative degree is advantageous. So a maths degree, engineering degree, a stats degree, physics degree. The vast majority of people I know came from an engineering or maths background, physics background, science backgrounds, things like that. Just because they understand those statistical methods, they understand the maths straight away. The finance can be taught and it will be taught internally at any quant trading firm. They will teach you the finance. You don't need a finance background. And I think most people that often fall into these roles are people who don't really know what exists. It's quite a niche area. People don't think algorithmic trading, what is this? Not many companies do this, they're not really talked about much. But they will particularly target math and math students on university Campuses because they know math students are really sort of the best problem solvers and they have the best mind for this sort of stuff and they understand, they can teach them the finance, the finance can be taught, but the maths, that's much harder to teach.

That is interesting that it's easier to teach finance to a math person than math to a finance person.

Very much so, yeah.

And from a data point of view, what are the most important data skills that you need in the field?

I think it's quite similar to any other sort of data engineering role, data scientist role, whatever you're applying for at other big tech companies, I think the backend infrastructure is going to be the same. You should be understanding relational databases, NoSQL databases, you should be confident in data pipelines. I really think that sort of recruitment process is very, very similar to any other data engineering role. Obviously being comfortable with big data is important, but I think every company these days is working with huge, huge data sets, whether it's a finance company or not. So I think yeah, that's quite standard on that sort of side of things.

Actually you mentioned things like data pipelines and some of the more sort of back end roles there. We talked a little bit about the analysis tools you use. So Julia and Python on the backend, like what are the sort of tools used there?

Yeah, so maybe languages first. Python of course, industry standard these days, some companies use Julia or whatever. But typically Python is a pretty standard sort of thing for everyone to know. So not even a developer. But if you're coming in as an analyst, learn some Python, learn a bit of Python, you know, learn how to work with data frames, pick it up some data, make a plot, something like that. Extremely beneficial to have for everyone then if you're in a role as a developer, as a software engineer, of course, low level language is the key. We're talking low latency, we're talking high throughput, lots of data. So C, quite common, C Java, some companies take advantage of Go. Languages like that that are incredibly fast are pretty common. And then I mentioned KDB as well. KDBQ is something that some companies leverage as well for the time series database there and that sort of homogeneous environment. And then on the database side, a bit more standard again, KDB is a database, but you know, both relational databases. So Postgres Oracle is quite common but then also no SQL databases as well. So MongoDB and particularly Cassandra Apache. Cassandra is quite commonly seen.

Okay, so quite a wide variety of tools then and particularly with databases. So there was the KDB for time series and then you got the NoSQL databases as well, etc.

And you'll often see in one company there's certainly more than one. Yeah, okay.

And you said that the companies will typically teach you financial skills, but are there any particular sort of core financial concepts that are useful to know?

Yeah, I think just understanding trading is the key thing. So, you know, understanding what are futures, what are options, how do options work, you know, what's a call option, what's a put option, what's the payoff of a call option? Sort of pretty key fundamental concepts are the most important thing. What is an order book, what is buying and selling bids and asks, what's a spread? And I think a lot of that comes down to knowing what the company does. If you apply to one of these companies, you should know what they do. Are they a market maker, are they doing some high frequency trading? What's the tech stack? That's probably information you can find publicly as well. It's probably public information somewhere. Probably lots of talks on YouTube, on the website, things like that. So I think understanding what the company does and then learning a little bit about that, you know, do they trade bonds, fixed income securities, do they trade commodities, what do they trade? And then learning a little bit about that is a way to stand out.

Okay, interesting. So all of it's really about understanding, like what does the jargon mean and what are the core concepts there?

That's right. Like any role, there's a lot of jargon that's going to confuse you. But if you can get a head start on that, it's going to be beneficial.

And it seems like there's a bit of overlap between what you're doing with algorithmic trading and sort of more general data and technology skills. So how much of what you do is sort of finance specific and how much of it do you think is sort of transferable or usable in other industries?

Yeah, I think very transferable. I mean, I think anyone, if we start at a software engineering background, anyone who works in quant finance could also go work somewhere else and vice versa. If you know C, you can go work at any tech company. Right. It's a widely adopted language and very similar for anyone in a data role. Data scientist, Data engineer. Yes. There are key things you'll need to learn about finance, about financial markets that may not be applicable to a typical tech company or to some other role, but the tech is going to be the same or very similar. You're going to be using SQL, you're going to be dealing with data pipelines, you're going to deal with relational and non relational databases. No SQL databases, in memory databases. All those core concepts are going to be the same wherever you go. So a lot of transferable skill in the tech industry, maybe for analysts, not as much. If you're someone with a finance background, a finance degree, you probably need to work in finance. Unless you're sort of making some sort of change. But that's not even necessarily true these days either.

All right, super. And do you have any final advice for anyone who's interested in algorithmic trading?

I think just learning a bit about it, like I said before, is the key thing. You know, if you're someone who's currently studying, focus on maths and stats, try getting good mathematical background, statistical background. Even if you aren't doing that as a degree, try to learn a little bit. The core concepts aren't hard. You don't need to be a maths genius. I think there's often an image that you need to be a maths genius to work at these places. You really don't. And I think even thinking about sort of the interview process, the questions often aren't incredibly hard, but they're designed to see how you think. So you're not going to be asked some incredibly complex mathematical problem. The problem is probably actually quite simple, but it will just take time to think through it. And what we want to see is how you solve that problem and how you break down that problem and how you articulate it. It's more important to think things through correctly than it is to even get it right at the end. We want to see your thought process. So focusing on things like that, getting a good background in maths and stats. If you're a finance student, learn some Python. You know, I think it's not a hard language, it's really beneficial wherever you go. Every company these days is using Python somewhere. If you're doing analytical work, you're working with data, Python will help you a lot. So I think a little bit of upskilling there. If you're a math student, maybe learn some finance. If you're a finance student, maybe learn some maths, learn some sort of software engineering and that'll really, really help you a lot.

I can see how thinking things through before you start is incredibly important with algorithmic trading. Like you don't want to lose the money after the trade, you want to see it's going to go wrong beforehand.

Yeah, very much correct.

All right, fantastic. Okay, thank you very much for your time, Anthony.

Thanks a lot, Richie. Thank you.