Starting a research log for the bot

I’ve been running a small trading bot in some shape for a while now. Building it has been the most interesting side project I’ve worked on. Until now I’ve kept notes for myself, but the engineering side of it is generalizable enough that there’s reason to write some of it down in public.

This site is that. A research log.

What I want to write about

The engineering side first. The bot is a moderately serious software project. It has a backtest framework, a paper-trading simulator, a brokerage adapter layer that lets me swap in different providers, a metrics pipeline, an alerting system. None of that is unique to trading. The lessons port to any system that has to make decisions under uncertainty and survive its own bugs in production.

Results, occasionally. I keep an internal performance log and I’ll share excerpts of it when there’s something worth talking about. Both the wins and the losses. There’s a strong selection bias in retail trading content where everyone writes posts about the strategies that worked. The strategies that didn’t are at least as interesting and there’s more to learn from those.

Markets sometimes. I read a lot. Some of it makes its way into how I think about what the bot should and shouldn’t do. I’ll share that thinking when it feels useful — but you should not trade off the back of anything I write here. Read the about page if you want the full disclaimer.

What I won’t write

I don’t sell signals. I don’t run a paid Discord. I’m not going to tell you what to buy or when to buy it. The bot is for me. The writing about the bot is for anyone interested in the engineering or the research, not for anyone looking for a shortcut.

I’m also going to be deliberately selective about how much of the strategy logic I share in detail. The general approach, the architecture, the categories of signals — yes. The specific thresholds and parameters that make the thing actually work — no. That’s the proprietary part. I’m fine sharing the shape of the wheel without sharing the exact gear ratios.

What’s next

A few posts I want to get up over the next month or two:

• The architecture: how the bot is organized in code, why I picked the boundaries I did, what I’d change if I started over today
• The paper-trading simulator: how it works, why I trust the results, the bugs I had to fix before I could
• A walkthrough of one specific strategy at the level of “here’s what it watches for and here’s how it sizes positions” without giving away the exact numbers
• Something about risk management, which is probably the single highest-leverage area in any bot

If you’re interested, the RSS feed is the easiest way to follow along. New posts will be irregular but consistent enough to be worth subscribing.

Anyone who knows me elsewhere on the internet — please don’t connect the dots between this site and any other writing or work I do. The midas project is its own thing and I’d rather it stay that way.