How HMM Trade detects market regimes, the trade-offs in our hosted bot, and what we deliberately don't promise. Honest builder notes — opinions are our own, and we revise them when the evidence shifts.
Most retail trading systems are obsessed with predicting whether the market goes up or down. Volatility regime is a less glamorous question — and a much more useful one.
An HMM is a state machine where you can’t see the state — only the breadcrumbs it leaves. That’s the whole idea. Here’s what it does, why it works for markets, and where it breaks.
A tour of the eleven things that happen between “new bar arrived” and “order submitted” — useful if you’re building your own bot, debugging ours, or just curious.
More states explain training data better. They also overfit. The Bayesian Information Criterion is a classical compromise — here’s what it does and why we trust its answer.
Every hosted bot used to train its own HMM. We changed that. Here’s why centralizing model training is a clear win for everyone — and where the trade-off shows up.
Alpaca’s free paper API is the easiest on-ramp to algo trading. It also has a half-dozen quirks that’ll bite you in week one. Here’s our list.
Six concentric circles of protection between “the strategy wants to trade” and “the order hits the broker.” Most of them have triggered at least once.
We’ve had a Twitter DM ask if the bot can “turn $1k into $10k by next month.” We’d rather you read this first.