Transcription
What is high frequency trading? Hello everyone. Um, this is my first YouTube video. Many of you know me um from social media, from Instagram and Tik Tok, but those for those of you that don't, my name is Jackson Semos. I'm a quant trader. I work in high frequency trading at a firm where I develop and trade strategies that operate automatically in the market. No humans pressing buttons. Note, it's purely just algorithms running code all the day whenever I choose to have these systems on.
And I know a lot of people have theories about how this stuff works from the outside, but I'm going to explain how it works from the inside because this is literally what I do. I trade medium frequency trading and high frequency trading. And the reason I wanted to make a YouTube channel is because I think I can better teach you guys in more long form content than in short form content. And there's a reason why I started with this video specifically. It's because I see a lot of content about high frequency trading online and most of it is either too surface level or just simply wrong. People explain it like they read a Wikipedia article, put it into chat and spat out an argument and it it does it is starting to annoy. So I want to explain it the way I actually understand it from the inside. Um, so enough yapping and let's get into it.
So, high frequency trading at its core, it's this. It's using computers to trade financial instruments, stocks, futures, currencies, options, whatever at extremely high speeds in extremely high volumes to capture tiny profits over and over again. And when I say high speed, um, I'm not talking fast for a human. I mean fast in absolute terms. We measure latency, which is essentially just delay in microseconds. A microsecond is 1 millionth of a second. Your brain takes roughly 200,000 microseconds to process something you see on the screen. A a high frequency trader system is making trading decisions in single-digit microseconds. There is literally no comparison to anything else on in the world.
And the key thing to understand um high frequency trading firms are not making big directional bets. They're not saying I think Tesla is going to go up this month so let's long Tesla. That's not that's not how they how they work. The game is find a tiny market inefficiency, a tiny edge that exists for a fraction of a of a second and exploit it. Then do it again and again millions of times a day potentially. Millions might be a little bit of an exaggeration. Think of it like this. Imagine bananas are costing $1 at one market stall and a $1.2 at the stall right next to it. If you could buy from the first one and instantly sell to the second and do that thousands of times a day, you'd make a lot of money on something that sounds quite trivial. That's essentially what's happening. Except instead of bananas, it's shares. Instead of market stalls, it's stock exchanges. And instead of you walking between them, it's a server making the round trip in microseconds. That person is called a market maker. Okay.
So, how does a high frequency trader trading system actually work under the hood? Three things. The algorithm, the infrastructure, and latency. And let me walk you through each one. The algorithm is the brain. It sits in a loop. It reads market data coming in from the exchange in real time. It looks for a set entry signal or a set of set entry signals or a condition that tells you a trade makes sense for your system. And if it finds one, it will fire an order. All of that happens in microseconds constantly all day. Um the signal could be a lot of things. A price discrepancy between two exchanges, a statistical relationship between two assets that's temporarily out of line, a pattern in the order book, which is a live list of all pending buy and sell orders. um that suggests price is about to move. That's what we're looking for all day long. There's so many different signals, so many different entries that you can uh make a strategy on. So, I could go on and on and on and on about that, but that's not what I'm trying to teach.
Next, infrastructure. This is where people's minds get a little blown blown out a bit because having a fast algorithm means nothing if your hardware can't keep up. High frequency trading firms use custom servers. A lot of them use FPGAS, field programmable gate arrays. These are chips that you can program to do one specific thing with almost zero overhead. No operating system sitting in between. No general purpose computation happening in the background. Just data comes in, decision gets made, order goes out in literally nanoseconds. And something on this, a firm in Chicago about I think it was 15 years ago actually built a direct landline through to through to the New York Stock Exchange to improve the um uh the speed at which they received these data and the speed at which their signals were executed. That's how finite the the gap of high frequency trading is.
Next up, there's there's collocation. This is where firms pay the exchange, literally paid the stock exchange, the NYSE, to put their servers physically inside the exchange's own data center. And you might be asking why? Because even at the speed of light, distance adds latency. If your server is 10 km from the exchange's matching engine, your signals have to travel 10 km of cable. Collocation means your server is a few meters away. That distance can be worth millions in this business. You think about these qu you think about Jane Street for example 20.7 billion in 2024 they have the money to do this and it will make them more money having collocation on the NYSE and latency is just the obsession um that ties all of this together. Everything in high frequency trading is about reducing the time between something happening in the market and your system reacting to it. Some firms have built microwave tower networks in straight lines across countries because microwave travels slightly faster than fiber optic through glass. That's how serious it gets. That's literally how nerdy this stuff gets and what you're compete what you're competing against. And this is why retail traders will never have won't have a good chance in high frequency trading. I don't want to say never will, but won't have a good chance.
Now, let's talk about what high frequency trading firms actually do to make money. There are a few main strategies. Um, there's market making. This is the most common one. Essentially, you place a buy order and a sell order simultaneously, a bid and an ask, and you collect the spread between them. If a stock is at $10, you might bid at $9.99 and offer at $10.01. Anyone who buys from you or sells to you, you make 2 cents, the price difference. Multiply that by millions of trades across thousands of instruments and it adds up very very very fast. The risk is getting picked off. Someone trades against you right before the price moves against you and you're stuck with a bad position. Managing that risk is where a lot of the real strategies live.
There's um statistical arbitrage. You find two instruments that historically move together. two related future contracts. For example, when when their prices diverge beyond a statistical threshold, you short the expensive one and go long on the cheap one, expecting them to converge back to their historical um average. Classic pairs trading, um it's done in microseconds as well. Um then there's latency arbitrage, and this is one I would consider myself well educated in, and it's I've built a model on latency arbitrage, but this one's a little more controversial. Different exchanges don't always update prices at exactly the same time. A faster system can see that a price has moved on one exchange before another exchange has reflected that update and trade on that stale price. You're not doing anything illegal. You're just faster than the exchange. But a lot of people argue this is essentially front running and the debate around it is very much ongoing and there's both positive and negative um theories on it.
You can't cover hyper frequency trading without talking about May 6th, 2010, the flash crash. In about 15 minutes, the Dow Jones Industrial Average dropped nearly a thousand points, close to 10%, and then recovered most of it within 20 minutes. Trillions of dollars of value, evaporated, and restored in half an hour. And what happened? A large asset manager placed a massive automated sell order for futures, 4.1 billion worth to be specific, using an algorithm that didn't account for market conditions that triggered a cascade. High frequency trading firms detecting the unusual volatility did exactly what they were designed to do. They pulled out. They withdrew their liquidity. And when you remove liquidity from a market that's already free, prices will collapse. And that's the double double-edged sword of high frequency trading. Um, in normal conditions, it provides enormous liquidity. It tightens spreads. It makes markets more efficient. But in a crisis, those same algorithms can make things significantly worse because they're optimized to make money, not to stabilize markets. When the rational move is to step away, they will step away. The they are risk first models.
Well, all right guys, that's that's high frequency trading. um from the perspective of someone who who works in this space. Um we covered what it is, how the technology works, the main strategies, and what happens when it all goes wrong. Um there's a lot more to go into um like how a machine engine works, what a book dynamics, market microstructure. Um and I'll be covering covering all of that in future videos. Um this channel is only only going to be about quant trading explained by someone who's actually working in it. Um, so if that sounds interesting interesting to you, subscribe. Um, would genuinely mean a lot for the first video. Um, and just drop any questions in the comments. I'll I'll answer them. And um, I guess I'll see you guys in the next one. Thanks, guys.