rbit

Post mortem of a (almost) 1 million dollar profit HFT strategy

In the summer of 2023 I started posting some monthly screenshots of this (at the time) 40+ sharpe ratio HFT strategy. I used to only publish % returns, but if you ever wondered what the $ amounts were, there you go: 

The first 6 months of running this live were exhilarating, making 3k$ a day (upwards of +10% returns/day) while I enjoyed the summer with my feet dangling off a sailboat. Good times. But all good things come to an end! (I see you @quantymacro)

This strategy was a hybrid maker/taker, mostly mechanical, HFT strategy on a tail venue. A bunch of EMAs in a trench coat. It wasn’t purely a latency arb, it had to do with some exchange specific mechanic that could be exploited.

It was scrappy and a huge pain in the ass to trade: at the beginning I had to scrape my way into it because the venue had barely any API documentation up, it required to hold a bunch of inventory that was very hard to hedge, and it made sense that nobody else wanted to...

...spend this much effort into trading this. The venue was pretty small, and I’ve squeezed every single cent I could out of this, becoming almost 25% of the total exchange volume at some point.

To the best of my understanding, in this whole period I had a total of 2 competitors going for the same trade. Guess when they turned on their bots. Also, you would’ve never guessed that 1M$ could be extracted from this venue by looking at it’s stats. I was very surprised too.

The whole process started in September of 2021, and for the first year it was basically me trying to poke at their API in my free time a few hours a week on Friday nights.

By the summer of 2022 I was collecting data and placing trades, but I wasn’t massively profitable on the platform. Then FTX collapsed and I lost a lot of my money with which I was trading my actual strategies (the “serious”, non-scrappy, high capacity ones)

For 4 months I was in a really though spot, but that’s when things clicked and I had the idea for this strategy. In a few months I integrated it, debugged it and tuned it and at the start of Q2 2023 I put some leftovers into it and off it went. First day live: +15%.

After that glorious summer, around the last week of September 2023 I suddenly hit a wall. I didn’t realise that while I was enjoying my vacations I had scaled up so much that my scrappy Python infra was becoming seriously backlogged.

It became very apparent when a competitor appeared and flatlined my PnL immediately. So in October I had to waste my perfect tan and sit down to do a proper rewrite of the entire infra.

Luckily it wasn’t very big, barely 60k lines of code, so within 2 months I rewrote everything from scratch and swapped it. Up and to the right again.

Around February 2024 volume on the exchange starts to tank heavily. Competition also becomes more fierce, and to the best of my understanding we are now 3 people competing for the same opportunity. For the rest of the year volumes keep shrinking and profits keep reducing.

In September 2024 I do one last optimisation attempt: I take 4-5 days to rewrite the heaviest part of the infra in C++ to shave off some latency. Number go up again, but it’s short lived: there’s barely any non-HFT trading on the platform anymore and volumes keep going down.

So is this really dead? No, I’m still trading it. It still makes around 1-3k$ a month, which is more than the 400$ a month I spend to host it, it doesn’t require any maintenance and doesn’t tie up much capital.

I guess it’s a race to see if my competitors get tired of paying for their AWS bills before I do. Bad news for them: 

So the bottom line is: there’s probably more money than you think laying around in small opportunities, everything @robotjames says is sadly true, the gnarliest stuff is some of the best trades you can possibly do, fancy models are often not the answer...

...deploying something 1 month earlier could mean a lot more profits even if it’s not perfect, Python is a man’s best friend, everything has an expiration date, and max out all the capacity you can while you still can.

Besides this occasional tail stuff, my main focus for the last 6 years has been MM and statarb on top3 venues, and I hope one day I’ll be able to post something more related to that instead. But if you are just a guy fucking around and looking for some good trades...

...I suggest you go scrape the bottom of the barrels like I did here and look for ideas. If I found 2-3 more of these and implemented them it would’ve been quite a good living, not too technically difficult, and maybe sustainable for a few more years by itself.