MM075: Do You Need Coding Skills For Trading?

Okay, hello and welcome to episode 75, The Market Maker podcast. And a little bit different than our usual routine because Piers Curran is on holiday again. And Piers, I know you're listening, so enjoy yourself, but come on, you got to do some work at some point. But, uh, instead, in his place, we're going to change things up a little bit, and I'm joined by Milinddeep Bassi. And hey, Milan. And Milan is part of our technology team. And what we're going to do is I'm going to quickly wrap up all of the things that have happened in markets as quick as I can, because I'm just conscious of there's a lot that has been happening. I don't want anyone to go without. And then we're going to bring Milan into the conversation, and we're going to talk a little bit about his background and his expertise. Where he helps design and deliver some of our most advanced quant-related training to the likes of Citadel, for example. And so really want to pick his brains and get into that subject matter because I know it's a real growth area for a lot of students and future applications. So first off, what's been happening in markets this week? So mega-cap tech earnings have really dominated Actually, I guess summary, they weren't as bad as feared. Can't be said for all of them. Meta, Facebook continuing to fall on rough times. But let me just go through a few of them. Amazon, last night, we're recording this on Friday 29th, so on Thursday night, their shares went up 12% post-market. We are talking about one of the biggest companies on planet Earth going up 12%. So that's a That's a huge move. Strong sales and a strong outlook lifted their shares. Apple were also up 4% last night, beating on revenue and profit. Earlier in the week, we had Microsoft. They actually turned in their slowest revenue growth since 2020 at just 12% year-over-year growth in the quarter, which is quite insane. On the conference call, more importantly, so although their revenue growth slowed, On their outlook, they said they expect revenue, operating income to increase at a double-digit pace for fiscal 2023, and the shares rocketed after that. Alphabet missed on most numbers. I think when you were watching those coming out earlier in the week, you were thinking, "Oof, doesn't look too good." But their advertising revenues came in at $56.29 billion, and that was above expectations. There was a lot of anxiety in the market about these online advertising stocks because of the precedent that have been set by Snap and Twitter, who had disastrous figures because of marketing budgets generally just being slashed with the looming recession. But Google's stranglehold on that digital spend being— well, is ultimately, I guess, a necessity for businesses, is keeping that ticking over nicely. But Meta, as I said, reported their first year-over-year revenue decline in Q2, and it's going to get worse. And so they expect a range of $26 to $28.5 billion US dollars in Q3 revenue, and that was against the Street estimate of over $30 billion. So they've downgraded their future outlook. They cited continuation of weak advertising demand environment we experienced through the second quarter, which they believe is being driven by broader macroeconomic uncertainty. I think I read on their, their press statement and then listening to a conference call, I think I heard the word TikTok more than I heard the word Facebook. So that kind of is a summation of their fears, I guess, at the moment. Bringing earnings a little bit closer to home, Shell, the oil major, there to accelerate their share buyback program. They reported record profits for the second consecutive quarter. Obviously they're benefiting from surging oil and gas prices. France's TotalEnergies, which is their other, I guess, competing big company within that space, reported this week their second quarter profits that had almost tripled. Whilst the majority are suffering at this point, these energy stocks have really benefited from that energy squeeze. On the macro side, The European gas crisis is going from bad to worse. That is influencing then and really destabilizing consumer confidence in the eurozone, which is now at a 17-month low. The key factor there is flow on the Nord Stream 1 pipeline, that critical pipeline servicing then gas flows coming out of Russia, particularly to Germany. That's now been cut to just a fifth of normal capacity, and there's a lot of pointing fingers between whose blame, uh, that lies on in that sense. So other things, Q2 advanced GDP in the US fell 0.9% on an annualized rate, and that came after 1.6% decline in the first 3 months of the year, meaning 2 back-to-back periods of contraction now for the US. So all the different terminology aside, this is a technical recession. And so a lot of pressure now for what are the Fed going to do next. And we had the Fed decision, they delivered back-to-back 75 basis point rate hikes, but they surprised with the dovish hint that the Fed could slow the pace of rate hikes. And actually, if you were looking at your, your US equity charts, I think we're up over 7% in the NASDAQ 100 in the last 48 hours or so. So kind of, um, the worst case averted for any fears that might have been present for these big mega-cap tech earnings, in combination with the looming slowdown with the commentary from the Fed in the US, has meant that then rate hikes have been pared back. And in fact, European stocks are now looking at their best monthly gains since November 2020, and the NASDAQ 100's just hit a 7-week high. Um, talking of risk appetite, if stocks are rallying So is crypto. So Milan, you'll be happy. Bitcoin, Ether heading towards their best month since 2021. Ether is up about 72% on the month alone. So seeing a bit of movement there as well. But undoubtedly the biggest news of the week, of course, was McDonald's raising the price of their cheeseburger. Did you see that, Milan?

Yeah, I saw it.

How do you feel about that?

Well, I'm at the point where I've noticed near me anyways, I have 3 McDonald's on the same road. I go to each one, they have different prices. Now there's one in the middle which is the most expensive, and that's now going to be like over £5 for a meal. And I don't want to go there.

Okay, we've got an arbitrage opportunity here. I reckon we should buy some at the cheap place. We'll park out the expensive one and sell them on discount. So look, we've got a business idea there. But yeah, the cheeseburger in the UK, has gone up 20%, and, and it's the first time in 14 years. So yeah, the cost of living crisis is real, folks. Um, but look, let's delve into a conversation then with, with Milan. Really happy that he's taken out some time to, to talk to me. And really, it came about because at the moment we've got our summer analyst training program happening. Uh, Milan runs some really cool simulation sessions with them that particularly around Python and programming and quant trading. And I jumped in, and it's the complete other side of the area of markets I work in. And I was in the office the other day looking at Milan's screen, and it made me feel— it made me feel sick, if I'm quite honest, because it was just full of code. Um, but Milan, perhaps you can tell us a little bit about yourself and your background, your role at Amplify, and then we can go into perhaps a bit more of a conversation about um, the kind of tips and tricks about the quant space?

Yeah, of course. I think, um, well, I found Amplify, I think, back in 2018. Now, um, my dad showed me and he was like, you should be a trader. Traders, you know, they have nice cars, they, they live the lifestyle, you know, uh, the stuff that used to kind of get sold around back there. So I was like, okay, you know what, this sounds good. I'm gonna, I'm gonna check these guys out. Maybe I'll have a Ferrari or a Lambo or something by like 25, right? I was hooked on the dream at 18. And I was studying computer science at the time at Royal Holloway. And I was like, I remember telling people, I was like, I'm gonna shift, you know, I don't want to work in these FAANG companies. I want to be a trader. That didn't go very well. But the point was, I guess, is my dad got me hooked on finance. And then from there, obviously, I came and did the program with you guys for 3 weeks and back then it was in person, and oh, that was— it was an unreal experience. You know, I still talk about it with people today. Um, and then like things clicked in place, and I started out as an intern, and I used to do a lot of the grunt work, I guess. No programming back in the days. I used to— little things like fix this Excel sheet, add a formula or something here. Um, and then as I did more of my degree, I guess I found out where my passions lie with like Java and Python and things like that. And then I've kind of transitioned across many roles. So I think my, my current title right now is like a, a full-stack DevOps engineer, um, which basically means I can do, you know, your front end, which is like your HTML, your CSS, you know, making things look pretty. Um, but then I can also do the back end, which is all the logic and, you know, all the complicated things. And then on top of that, most people's career is normally confined there, but I now also do networking and things, so I can set you up servers, I can set up all this other stuff as well with internet and things. And yeah, I basically do what, apparently 3 jobs, I guess. But I get the title for it, so it's worth it.

And you deliver, of course.

Yeah, yeah. So that's the next thing. My title also says I'm the fintech product lead right now, which is basically like, like Ant said, I create quant or even algorithmic trading products now and then. We go to our clients and deliver them because I want to help other people program. I think it's a great skill to have. And, you know, I think the more people that understand how code works and hopefully they don't get headaches from looking at it, um, it just makes everyone's life easier and kind of good for the future, right?

Could you have done what you're doing now without studying computer science?

I don't think so. I think, um, It's a very difficult thing because I never touched any code until I was 18. And I think a lot of people have this misconception where if I don't do my, my like programming A-level, for example, that's my, my career gone out the window. So back in A-levels, I was primarily into economics. That was my, my strong subject, right? Um, and, and I only went out, kind of went to clearing on that day. I decided I'm gonna go from accounting and finance as my degree to computer science just like that on the day because I didn't get the results I wanted for accounting. Um, and obviously I think, you know, kind of to get to the point where I am at, yeah, I think I need my degree. And the only reason is because a lot of, uh, like probability and statistics and kind of the way you problem solve is what gets taught at uni. And I think that's the most valuable thing you walk away— it's not kind of knowing how to code, it's the, the way to look at a task break it down and go from there. But it's not something you can't just— I mean, you can learn it online. I guess it's just there's no real structure when you learn online. Obviously, with a degree, you just get that structure and path.

Yeah, that kind of structure and way of thinking you just mentioned, are we talking maths then?

No, so I don't use too much maths, but it's more like if you pose a question to me, like, a simple thing like, I want this button to do something. I think a lot of people would get fixated on, on the wrong thing. They might look at it and say, uh, um, you know, I've, I've seen in the past with some people, they get focused on the end result rather than how they're going to get to the end result. So I think when I say structured thinking, I mean you need to be able to look at the task, say I need to do X, Y, Z, I need to go in this order. Uh, and, and the problem with, with, with of our jobs, I guess, when you're in your engineer is it needs to happen now. I don't have, you know, a week. I can't say to, to, to a manager, I'll be back in a week, I'm gonna go and see what this button does. You know, I need to look at it and say, you know, off the top of my head, this will take me a few days maybe, or, you know, a couple hours. Um, and I think that's that structured thinking because I need to be able to quickly say, you know, I need A, B, and C completed before I can do you know, whatever, whatever I need. And I think that's the hardest thing that I find that a lot of people can't get is advice, you know, when I run the sessions is a lot of people get stuck on the how does this go to that and then go to this, and it's just because they feel like there's a lot of magic in between. Um, so it's a bit of a difficult one, I guess.

Is there any way in your mind to improve that way of thinking, like things you can be doing like even outside of that actual task, that are just general, using that same memory muscle or brain muscle?

Yeah, I think, well, that's where the kind of the, you know, like a business analyst kind of role comes in. I think one of the lucky things of doing my placement year with Amplify was I had a small time where I was a business analyst, and my main job was kind of, you know, to talk to someone like yourself and say, what do you need done? And then obviously I need to convert what you need done to what the engineers need. And being that kind of middleman translator almost from, you know, just like one or two sentences to maybe a whole paragraph detailing what needs to be done is what builds that skill. And I guess, how do you do that yourself at home? Well, it's like if I say to you, I want to build an algo strategy, the first thing I need to know is where do I get my data? Googling where do you get the data, reading up on it, putting that into maybe like a notebook and saying, okay, I can do XYZ, And then from there you have to pick which one I'm gonna use. And it's kind of just building projects from, from the start to the end without kind of cheating. Um, rather than just saying go to a YouTube video, right? For— let's watch this for 4 hours, the guy's gonna teach me everything I need to know. Um, I used to do that, but there's no fun, right? Because you don't have that research phase and looking around and figuring things out yourself.

Okay. And then I'm just thinking about, then you mentioned before, just going back to computer science, because I know quite a few people are kind of thinking of that as a subject to potentially study or study further. But I heard you say before about you've got a friend who was studying at a different university and he was— the difference between having good algorithm skills or good programming skills. What do you mean by that? Like, how is that different?

Yeah, so I think this is one of those things where when I got to uni, I thought, you obviously, you get taught the same thing everywhere but with slight differences. But what I found with, with my friend, he went to Loughborough, um, and the way they taught his degree and the way he's kind of come out at the end of the, the 3 years is if you had a question about algorithms or any more of a mathematical kind of problem, I would say You can ask me, but it might take me a bit longer than him. Likewise, when it comes to programming, basically, you ask me, I'll get the code written and working quicker than he does. And the reason being is I think he had more mathematical modules, so he learned more about like the theory and the key understandings behind computer science, like what makes things tick. And, um, you know, for example, execution time you know, there's a whole maths and a bunch of formulas behind how fast should something be and how do you calculate that, um, and things like that. But I had, I think, maybe 2 terms of kind of lessons on that, um, whereas he might have had maybe— I, I know, like, for in my final year, I didn't touch a single bit of maths. Um, my primary kind of modules offered to me were around information security, so I have a lot of knowledge in terms of being very like safe on the internet, you know, like hacking and things like that. Whereas he was offered more, uh, like I said, algorithmic stuff. So his knowledge on that, on that stuff varies. And I think that's the one of those things when you're 18 and you pick a uni, you don't know what, what are they gonna teach. Um, and it just happened to be Royal Holloway taught a lot more programming. So I think when it comes to— he's come to me in the past actually, when it comes to writing code, he'll come to me and pose a question, be like 'Can I do this in Python? Can I do this in Java?' Even though on paper we're the same person in terms of our degree and what we achieved, um, I'm better at that unfortunately, and he's better at the, the algo side of things. And that's the tricky part, is when you're kind of looking for, for talent, I guess, is you've got a bachelor's in computer science, you've got one. Who's the better programmer? But then, okay, I might be the better programmer, but then I could be absolutely rubbish at my my algos, which is not going to do well at like a place like Citadel, right?

So what, what's the most common then? Is it that most people are specialized in an area like that rather than a generalist? I mean, do you— in a typical like Citadel team then, you have multiple different specialists covering them? What would those areas be?

Yeah, so, so Citadel, just from a, a top-level breakdown, I guess they kind of split people into investment and trading. These are— they still have computer science degrees, but these will might be people who are on kind of the front desk, any trading job, right? Um, and they'll be good at risk management and actually reacting to markets at the same time. Uh, and then another role is, is your quantitative research. So quant research is pretty straightforward. It's going to be someone who creates these algos and researches them and, and finds all the maybe, you know, discrepancies in the market through statistical analysis and models, right? So this might be someone who's really good at maths. They'll, like I said, they'll still have the same background. They know how to program, they did computer science. And then you'll have software engineers. And then software engineers are those who can program and kind of create the systems for the other two to use perhaps. And then you can see like everyone would have the same background almost, but they've split them into kind of these three groups of people. So you've got your maths heavy. So I would tell my friend, for example, if he got the— if he understood finance a bit more, you go for a QR role because, right, you'd excel there. Myself personally, I'd probably be split between software engineering and investment and trading. I'd have to kind of try both out. But, um, you know, other people are, uh, who kind of program more than anything, I'd say go do a software engineering one, right? Because investment and trading is a very— I guess that's the hard one to find because you need someone who can program manage the risk and understand the finance and react to things like you would as a trader because you're basically managing live code.

Yeah, I remember being pulled in because there was a market event that had happened and they wanted me to come in and give a quick soundbite. I think when one of you guys were training with them and I remember I was talking to one girl and she was previously interning at NASA and I was just like, some of these CVs are ridiculous. My mind's blown now. I used to work as a 16-year-old in McDonald's.

Yeah. I worked at Toby Carvery. Humble beginnings, I guess.

But talking of these people then, when we're talking about these elite-level financial institutions, from your experience then, because I know that's not the only one that you've dealt with, what's been your experience, your interactions with these people? Because I think there is a degree of mystique around particularly the quant side of the industry.

Yeah, I think, um, like obviously from my cohort at uni, I can't think of these people that would work at these companies because they're kind of more— I don't know how to describe it. I couldn't pick someone from my whole cohort and say, you know, he has the skills for Citadel. And like you said, because there is a— I don't know where they find these people, uh, the, the, the, like the, the talent and the way they, you know, they come across is very much like, you know, you— I was speaking with Piers once. I was like, you look at them and you think they're just your average geek and kind of, um, you know, you'd expect them to work in like a FAANG or whatever, right? Do it, do it kind of a more comp sci stereotypical job. And then you speak to them and you realize these guys have immense knowledge about finance already at this such a young age. Um, and I think that's Obviously that's why it's such a lucrative role almost, right? You're, you're, you're looking for a needle in a haystack that can kind of do something that's not— I mean, I, I can't— I was looking at master's, for example. It's very hard to find a degree where they'll teach you comp sci and apply that to a finance background. So computational finance, for example, there was only a— I mean, it's taught widely, but there's only a few places where it's actually taught well. Um, and I think that's the tricky part, right? Like, I can go do a master's in comp sci, but it doesn't mean I'm gonna get this job at Citadel because I need that specific kind of angle, I guess. And then that's very hard to find. And I think the way you'll find someone like that is someone who, like myself, I guess, is someone who's been interested in finance for a long time and kind of, you know, in their free time, you know, watches the news and keeps an eye on things like markets. We've got, what, 10, 15 engineers now at the firm, and I'd say only myself and one other actively keep up with markets. It's because we have a shared interest, I guess, in that. And the rest are just, they do their engineering things and whatever interests them.

So the key that you're saying then is that if you want to apply that technical knowledge to the financial industry, you've got to still have, like in any other traditional role, like in global markets, as a normal standard trader, you've just got to have that curiosity and interest in the subject matter itself as well as the technical.

Sure. I think if you, like, obviously if you're applying for software engineering, I don't think you need it as much, right? Because you're mostly going to be in a back office. But these lucrative front office roles, I'd say it's the skills are the same, which is why I think, you know, without bragging, I think I do pretty well in because I did the program with you guys, went back when I wanted to do finance, right? So I took all the knowledge in and everything that we teach and kind of, you know, I guess I owe the pro— if it wasn't for the internship, I guess I wouldn't know half the things I do and I wouldn't be able to do it because I got in early, I guess, when I was 18 and fucking my head and You have to carry on, right?

Because, you know, well, look, let's talk a little bit about what you've got cooking in the background then, because I know you've always got your fingers in a couple of pies. And I know you were, you were early on the whole crypto move, you were early on the whole busting out of a lot of the GameStop action when COVID hit. From, from a just general market interest point of view, I don't know, NFTs or just Any algorithms you're working on? Is there anything at the moment that's floating your boat?

NFTs and crypto I've kind of left behind. I think I've told you a few times, I've kind of burnt myself out on that. And I think it was the peaks kind of last year around March, right? It was when I had my crypto hype back. I was like, I was in this. I was like, I'm going to hold this for the next 10 years. And then just suddenly out of nowhere, one evening, I think I told you, right? I was like, I was just sitting here at 4 AM in the morning I woke up and I was like, I should sell my Ethereum. I don't know why, but I looked at the price and I was like, I've been watching this for a couple days, it's not moving. And since I sold it, I haven't touched it again.

So you're the Ethereum whale who got out at $4,000 or whatever it was, then they dropped the market.

Yeah, yeah, yeah. That's the thing, it's quite funny because I remember especially I just woke up randomly in the middle of the night and I was like, getting rid of this. It's, it's, I have my hype has just disappeared during my sleep.

So have you actually run any models over the crypto data? And what has that told you, if any information?

I had a few algos. They were primarily technical. And the reason being is I don't want to go into the fundamental side of algos because fundamental analysis, I guess, in terms of news, is I can't compete in speed. You know, like the oil release, for example. All right, cool. I could probably get the number as fast as you can read it. But, you know, there's so many people out there who— and this is one of the interesting things, I guess. I spoke to Citadel, one of the engineers. He said they don't care about seconds between trades. They care about like the nanoseconds between trades. And because, you know, to the point where obviously, you know, these firms have their exchanges in the same building as the actual exchange. When they're executing these trades, they're like maybe across the corridor from them. Um, in terms of myself with the models, I guess I— most of the strategies I've had kind of, uh, they used to work, I guess, up until this year. Um, you know, they looked good on paper, uh, but unfortunately I developed them this year, so when I, when I let them out and, and I was expecting them to do things they just, they just failed. And I think that's, that's the hard part. I think a lot of people forget is, you know, I— well, I've been working on algos now for 4 years technically, on and off. But, you know, you, you need to, to keep trying because— and even when it does click, it might work for a week, and then, you know, markets change, and, uh, and then it'll stop working, even though you'd think, oh, it's, it's the same strat. You know, I might be looking at something like, you know, whatever you're doing. And, and I think that's the, the bit where you get frustrated you get annoyed, you feel like, you know what, I've wasted my time almost, you know, countless hours of coding. And you're like, I'm just gonna shelve it. So I have like on my— I have a folder that has a lot of code in it and different strategies over the years. And the problem is one of them might be working right now, and I don't know because there's so many, right? And that's why you have these— now it clicks in my head why they have these massive teams, because you need people to to go back and look at these strats and say, "Okay, it used to work, but what's slightly changed?" Obviously a one-man job on that is quite difficult.

Just a general question. I'm not expecting a definitive answer, but take where we're at in the market at the moment, where we got an inflation situation where it hasn't been this high in multiple decades. Now, from a data perspective, what would be interesting is to go back and look at Fed rate behavior when inflation was last in modern history at these types of levels in order to then calculate what might the Fed do in the future. From a data perspective, I can see how that would make logical sense, but how do you start to then overlay in the aspect of that was 1980, we're now in 2022, the market dynamic is incredibly different now in terms of the dynamics of who's involved and how they're executing and managing these positional trades. What would be your first go-to then when you're looking to try and spec out that type of question?

Yeah, I guess when it comes to analyzing the future and things, obviously this is when you throw in machine learning and things like that. Um, and the reason being is the machine learning, you know, it's a buzzword and it gets thrown around, and literally everything has machine learning these days, right? Apparently. Um, but you want to take your data like you're saying and just basically just extrapolate it into thousands and, you know, hundreds and thousands of different outcomes. It's kind of like, you know, um, when you, for example, in the movies, right, when, when Doctor Strange does something and timeline goes into, you know, thousands of different ones or anything like that. It's the same thing. You want to take this data and say, you know, predict, you know, a million possibilities that can happen. And then the thing is, then the hardest part in my opinion is then you have to rank them and you have to— and that's, I guess, your secret sauce, right? How do you look at these a million possibilities and say this is the best one, or, you know, this is the worst one, you know. I guess if you can crack that, that's how you'll, you'll make the money. And I haven't cracked that yet, uh, in terms of being able to say like— obviously, because the problem is, right, you might Google it and you'll be like, oh no, there's a, there's an answer right there, but it's on Google, it's— that answer is not going to work, right? I guess it's like, like anything, if you want to make a good algo, even you need to— it needs to be your own, and, and that will be your edge. It's just like trading, right? Every trader might have their own edge. The same thing applies here, I guess, is I can Google it and I can tell you, oh yeah, let's apply this model onto it. But if it's on Google, that's not going to be the one that works. And there's a reason, I guess, it's publicly available. I'm sure like Citadel have stuff that would rank things and yeah, these models, and that's where the value is, I guess.

So that kind of, again, that terminology like machine learning or AI Well, I mean, is this just fanciful to like to be talking about? Because I know what you mean. Like you get the sales team involved and all of a sudden your product has AI and machine learning like potential. But what's the reality of that in your eyes in terms of where we're at at the moment and then to where it's heading over a longer period? Are we like in the infancy part of it at the moment or is it more developed than that?

No, I think we're still in the infancy. Like obviously if you look at NVIDIA, right, their graphics cards are just now putting the power, I guess, of being able to actually do like fast computations in the hands of your average Joe, right? You know, as we progress, and I've seen this, you know, they're focused less on gaming these days and performance and focus like, you know, our graphics card can also do X, Y, Z in computational power. I think as time goes and, you know, this computational power trickles down to your average Joe, um, then I guess is, is when kind of machine learning makes more sense. Like, you know, even with the iPhones, right? Like, only the, the new iPhones have those, you know, they obviously, they added that dedicated chip just for like, you know, machine learning and things like that in your phone. Um, only now that because this stuff's happening, I would say that machine learning makes more sense to put into a product. But like you said, they slap the title on literally everything and Sometimes the machine— like, machine learning is, I'd say, it's misunderstood because there's different ways and different levels. Like, you can slap a label and I could have the smallest amount of machine learning, which is— which might just say, you know, Ant likes coffee, someone else likes coffee. Like, that's a trend, um, you know, because that's all machine learning is, right? It's like, okay, 5 people like coffee now and they all live in one area. You know, there must be good coffee or something there. But like, that's not the machine that you think of in your head, right? And same with AI. I think at the end of the day, I think, you know, it's an oversimplification, but all AI— I probably might get some slate for this, but it's just a bunch of conditional statements at the end of the day, um, in terms of decision making, right? So I could make a robot and you could ask it like 5 questions, it'll make decisions. I can— I say that's an AI. You know, you, you asked it 5 different questions and it made a choice. Obviously then if you try and do other stuff, we'll say it's not implemented yet.

Yeah, you saying that then in that kind of more narrow structured way and, uh, being a consumer of Netflix content, so, so then at this point, how do— how, how— I know this is a massive question, but How do you then start to counteract then the designer's influence on the instructions that create the underlying architecture of these, of these systems? I understand these systems can grow out and then design their own pathways, but the core architecture has to come from like a you, for example, and you will have some degree of biases. So how do people address— where's that conversation at the moment? Because I know it's a big one.

Um, I think that's the, that's the tricky one. Like, uh, even, yeah, I don't know, like, if, if I make an algo, like you said, I will have bias because I think one way is right, and, and obviously you might think another way is right. Um, but that's why most of the time when, when you see these people, um, making, say, AI, that there are rules kind of in place, you know, not officially in place, but, you know, like guidelines that an ethical programmer would adhere to. And, you know, uh, in terms of, you know, you don't want to make an AI that just hates everyone. I mean, I've seen the AI that hates everyone, wasn't it, on Twitter or something, right? Yeah, yeah. So like, like, it's a very— like, it's a very hard question to answer. I guess it depends on it. I think, like, I'm just assuming anyways, I'd assume that if you're a massive company and you're working on AI, you'd have a team because you'd want to make it as neutral as possible. You'd And, and that's unfortunately the thing. Like, you're basically just taking someone's brain, copying it, but making it unbiased. And that sounds so hard to do because, you know, even from an— from a trading perspective, I have to make an algo that I might flinch off to say hitting my stop loss. Like, you know, let's say I hit 5 trades in a row on my stop loss. This algo needs to hit those 5 stops and keep going. At the same rate it did the first trade, right? It needs to be unaware rather than— I might be like, oh, if my algo hits 5 stop losses, I want it to now use a tighter spread. That's putting my bias into the program, and then it might make the, the algo worse, essentially. Um, whereas I should be taking it as the algo has no emotion. Every trade is— should be the same trade, right? Regardless of the previous results.

So, um, yeah, I guess I got two things that I read recently. One was just trying to recall the details, but it was Nomura put out a note about they were using AI for ESG scoring. And I get that because then that was basically more about just these monumental datasets and it was generating certain pattern recognition that could then identify trends that a human would never be able to see. That then similar to the other example I had was when, I mean, When I started using Twitter for markets, I mean, this was a long time ago, over a decade ago, I guess. And back then it was like not a well-recognized source of information as it is today. And there was lots of pop-up companies coming out and they'd have software aggregation tools that would then pick out on the hosepipe and alert them, a team of analysts that would basically then add the human layer of then of the million tweets, the machine's punching out 500 important ones it deems, and then the human picks out the 10.

Yeah, for sure.

There was always like, there's always the human element. There's almost like the pilot. It's almost like being in a super sophisticated aircraft where the plane can kind of fly itself, but the human still needs to sit in it to a certain extent because of the the what-if scenario, or is this actually the right decision for the context and things like that? So yeah, I mean, that's how I've seen it develop in my career. And that's what I was still reading with the Mirror as well the other day.

Yeah. I think it's like you said, you've hit the nail on the head there. It's exactly the same still. Obviously Citadel hire their investment and trading people and they do what you said. They'll have the strat running, but you have that individual you can rely on. And if something, you know, goes haywire or it places the wrong trade, you need that, you know, the pilot in this case, I guess, to put things back where it should be. Um, and I just had a quick look at this, this Numera thing that you mentioned, actually.

Oh yeah.

Um, I attempted this with Reddit back, back during the—

so go on, talk me through it then.

Um, I try— well, there's other websites that already do this, but I was like, I want my own one that kind of does it for me. And it would just scrape Reddit for every new post coming into WallStreetBets and just say, what's the ticker? Analyze the sentiment. So you have natural language processing, which is basically you look at text and you compare it to a dictionary or a library collection of words that will say the word bad is negative. If I said the stock is good, that's a positive sentence. Um, so I'd look at a headline on Reddit and it might say like, you know, everyone uses the dollar sign, for example, just to make it easier. So I'd say, okay, in this text, look for the dollar. What are the— what's the ticker after the dollar? It might be like $AMD. Um, then what, what is the rest of the sentence? And then from that, I'd give it a score saying, okay, that was a very positive one. Um, and then I'd add to a counter saying this was mentioned once in the last 24 hours. It'd run, you know, constantly all day reading these posts, and, and all I'd get at the end of the day, still manually like you said, right? I'd log on to a— I made a web, web panel for it, like an admin panel. Uh, I'd log on and say, okay, he's put— this ticker was mentioned X amount of times, this, this is how many were positive, this is how many were negative. I can click on it and view all the, all the different things. And the reason being is, I guess, back then I was like, okay, You know, yeah, yeah, who's next? What I want to know over the, the next couple days, right, there might be something that was mentioned maybe 4 times, and then as it starts climbing and trending and you're like, you know, that's the next one, and I hopefully hop on it early. Um, the problem I ran into, I guess, was like I said, as a, as a one-man job, um, I didn't have the power to, to, to run this thing and, and the data and you know, leaving my computer on 24/7, for example, just scraping data and analyzing it. And then I think I had so much data, I just didn't know what to do with it myself because, you know, it becomes too much. And that's when, you know, like big firms would have these different teams, I guess, where I need someone who is a data scientist essentially and takes my data and says, am I analyzing the correct way, right? Because obviously I did it from a very, very simple, point of view. Um, but like I said, there are way more intense ways, I guess, of ranking things. And the problem I ran into was— and this is the one that still bugs me today because I couldn't fix it— uh, was what happens if you don't put a dollar sign? Or, you know, when WallStreetBets kind of clocked that people were checking on the Reddit and running these algos, they started using code. By, um, for example, like, instead of GME, right, they'd write something else. Um, 'How do I get my algo to pick up these ones?' Because it might say that GME was mentioned 5 times when actually it might be 500 because they're using their own code, right? Um, and then problems ran into, okay, what if someone puts 2 spaces between the letters G and M and E?

Right, so the permutations start building out and you need— that's where you need that maintenance, I guess.

Yeah, and that's where machine learning kind of— you want your, your algo to get smarter to say, okay, look, someone's—

yeah.

'GME' with one space yesterday, but now they're writing it with two. It needs to correlate that to the same. I was doing it manually. I had a text file where I would add tickers. So if I saw something on Reddit, I'd be like, okay, people are typing it like this. I'd add it to the text file and then my program would start looking at it. Obviously, that's not sustainable. So that's another one of my many shelved projects. I can say I've been there and I tried it, I guess.

Okay. Well, look, just to conclude things, I guess there was a practical element of— you get told quite a lot, rightly or wrongly, if you're interested in finance, particularly in markets and trading, and this doesn't necessarily need to be quant, this could even be just more standard investment bank sales trading, that you need to know a programming language. I think finance students now get told that, just like vanilla, you should learn programming. Now, A, is that true? And then B, If you're going to learn a language, which one? And then I guess, C, how much do you actually need to know of that language?

Okay, that's fair. Is it true? I think so. I think, well, so I like, you might learn your programming language and go into your job and you might not use it. But I think what it prepares, you're preparing for a worst case scenario where let's say in 6 years, you know, we don't know what the landscape will be like, but If you already knew now how to code and maybe have just— all I ask of people is just have a basic understanding. So when you're talking to an engineer, you don't feel lost, perhaps, you know. And what I kind of preach, I guess, to our interns, I guess, and just anyone I meet is learn a little bit. And you never know, at your job you might come up with an idea. You know, you don't have to code the idea. Just coming up with the idea could lead to, to you revolutionizing a company, right? But if you didn't know how to think from a kind of— what, like, if you don't know what can code do for you, how will you come up with these ideas, right? You wanna— like you said, we're always pushing forward, I guess, as a, um, as a race. Um, but, you know, you need to— yeah, that— I guess that's my point is, you know, Don't think, oh, Milan says go learn to program, I never use programming for the next 2 or 3 years. I'm just preparing you and kind of giving you the opportunities to say, right, this is clicks. Because, you know, people come up to me all the time like, I got an app. I'm like, because you don't have an understanding of how programming works, your app is too ludicrous, I guess, for me to do anyways on my own. Or, you know, you're talking a couple million in terms of development for, for your little idea. If you had a better understanding for programming, maybe you might have come up with a different idea and things like that.

So where do you actually start then? What's like the base language you'd recommend?

So I start— so I think this surprises people. I actually started with Java, even though I don't use Java anymore. That was— I spent a whole maybe year and a half at uni learning that. Then I did JavaScript and Python and C and and everything. And I think when it comes to learning a programming language, I think I always recommend Python. It's not the fastest, the, you know, C is hands down, I guess, the, the better one for, for finance, but it's the easiest to get into, right? You download, you go to the Python website, I think it's like python.org, you download the thing, you open up a notepad, you write code, and it just works. Like when it comes to Java and things, right, there's a lot more steps involved. You know, the last thing you want if for someone who wants to get into programming is have issues installing the language and getting it running, right? You just want to— I just want to point people to this place and say, look, download this, you get this, this little Notepad window, write your code and crack on. Um, I say Python, but I think the, the one thing I always emphasize is Regardless of what language you start, you just want to learn the fundamentals. So things like variables, data types, you know, data structures, um, you know, writing simple conditionals, for example, and then loops and things like that. Once you understand these concepts, when you go and pick up another language, all you need to learn is how to write it, right? I, I always say to people, it's the same as English UK and English American, right? It's the same thing at the end of the day, but they, they, they say, uh, aluminum and aluminium. Um, and at the end of the day, like, you know, you can understand it. And my point is, if you learn Python, and I think going on to your question of how much, I guess, do you need to know, um, I always say to people about— I say 50 hours, um, would probably make you very not very good, like you'd get a good understanding for programming. So for example, let's just say with Will, right, I'd say when we started this whole, uh, quant shenanigans, I guess, as a firm, um, Will knew, you know, very minimal code and it would like kind of wouldn't be able to look at like maybe 3 or 4 lines and understand what's going on. Obviously he's been doing the sims with me and, and I, I remember once when we were meant to go to the States, but obviously we got denied. Um, we went to have lunch and I sat down with Will and maybe spent about 2 or 3 hours of me just reading line by line each thing in our quant program and explaining to him how does this work and things like that. And then kind of Will's understanding, I guess, is improved and to the point where he was even saying to me he could run the sim almost, even though he's, you know, he's not an engineer, obviously. Um, he doesn't even code for a living, right? He doesn't do it every day, but you know, repetition. And I would probably say I've spent maybe 20 to 30 hours with him just looking at the code. Has got to the point where, with our quants in many ways, that Will almost sometimes answers the questions without even coming to me. But when we ran it initially, it would be to the point where Will would have his hand up every 5 seconds being like, 'Millen, can you come here?' So I think in terms of learning, I'd say go for Python just for the ease of getting into things. And then about 50 hours, I'd say.

But so 50 hours, I'm just thinking about the maths there. So you're— I know we're coming to the back end of summer, but you could do it as a summer project, right?

Yeah. So I think the easiest way to learn programming is don't— okay, so go do your Code Academies, you know, your different courses online. Don't pay for anything ever in terms of like, you know, £15, £20 to learn Python because you can Google how to learn Python, right? It, you know, engineers have helped each other out, you know, all the resources are online. Um, but basically pick a topic that you want to do. Don't just go in with the idea of I'm going to learn programming and then follow the, the, the course. I'd say, say, okay, I want to build, let's just say, a very simple, uh, text analysis bot, right? You give it a sentence and it'll tell you it feels bad or good. That sounds horrendous and really hard, I guess, from, from, say for example, you were to do it, right? It's pretty— sounds really difficult. Um, but then break that down, right? Okay, I need to get my Python code to read text. That's now one like subtask on its own, right? Um, okay, I want to read the text from a text file. Okay, then that's another subtask. And all I would say is just do that little bit by bit, you know, like Lego. We're gonna start with one little, little piece and just start building it up. The next thing you know, you'll have your, your Lego Bugatti or something made. Um, but like, start with that, and then you Google, okay, how do I read a text file in Python? Like, you know, internet's amazing, everything's there. Um, and you start from that, and the next thing you know is you've done that. Then the next task, right now, how do I How do I score this text? How do I determine if this text is bad or good? Then you go down a rabbit hole of that, and then you'll— you know, I think that's the best way to program because you're not doing it for the sake of, like, I'm sitting down to learn something, right? I'm doing this because I want to achieve something, but at the same time, I'm going to learn from scratch. Um, and I think that's the biggest difference, I think, between myself and many programmers is I spend a lot of time outside of work, I guess, just writing random stuff, right? And like I said, I've got hundreds of scrapped projects and just shelved, but I learned something doing all of them. And that's what kind of builds you up into being a good programmer almost. And it's just like finance. I'm sure you can agree, like, you've probably spent countless hours, I guess, um, just researching and reading for your own enjoyment as well at the end of the day.

Yeah, much as my wife hates that. But look, look, we'll wrap it up there. Milan, thanks very much for giving up some time and sharing some of your insights and thoughts. What I will do is I will put your LinkedIn profile, if you don't mind, into the show notes on the podcast. So if anyone wants to reach out to Milan, I'm sure he doesn't mind helping.

That's fine. I always tell anyone, if you ever need anything, message me. I think I had someone from about 2 years ago I met, and she sent me an algo she did for a dissertation or something, and I was like, okay, well, I had a read of it and I gave some feedback. Um, but like I said, my inbox is always open. If I don't reply, don't take it personally. It's just quite busy. Uh, but eventually I normally tend to try and get back to people. Uh, or you can always, yeah, message me on Instagram if you're really, really desperate.

Cool. All right, we'll wrap it up there. Uh, thanks everyone for listening and we'll be back as normal next week. All right, have a good weekend everyone.