Not All Bots Survive: Lessons From the Other Side of the Order Book

In the previous post, I walked through how high-frequency trading operates within crypto - how it creates liquidity, how exchanges support it, what kind of structure it relies on, and why for some, it offers a more systematic alternative to "manual" trading. But that story mostly focused on the visible part. But what's worth exploring now is what happens when everything looks set up - yet the core logic is missing.

Building bots, integrating with APIs, and optimizing for low latency isn't the hard part anymore. Anyone with a decent technical background can spin up an engine, place a few thousand orders per second, and look busy in the market. But speed, as it turns out, won't help if your decisions aren't grounded in anything. When you're operating without a model — without a statistical backbone, without any structure for understanding market behavior — you're not trading, you're volunteering.
But of course crypto has no shortage of volunteers. Some of them build technical setups that look impressive from the outside, but run empty inside. Others subscribe to signals, Discords, or their favorite influencer's "edge", mistaking a Twitter (sorry, X) thread for a strategy. You can watch this play out in real time - and in some cases, you can even watch the liquidation happen live.

The recent blow-up of James Wynn 🐳 is a good reminder: having capital and confidence means nothing if your framework doesn't exist, or worse - if it's borrowed from someone with less skin in the game than you. It's one thing to be wrong on a thesis, but it's another to be fast, leveraged, and exposed - all without any coherent sense of where the edge is supposed to come from.

And while it's tempting to believe that speed alone can compensate for everything else, this assumption tends to fail quickly when your actions start to repeat themselves. Markets are exceptionally good at detecting routine, and once your behavior becomes recognizable, it becomes manageable - and eventually, tradable. Without a framework to adapt or defend, even small decisions turn into structured opportunities for others.

Which brings me to the hard part: real HFT - the kind that survives - is less about execution and more about research. And research, inconveniently, is expensive. Because it's slow, it requires iteration, calibration, and sometimes humility. But it is the difference between becoming someone else's liquidity - and staying in the game long enough to matter.