What HMM Trade is *not* — honest limits of regime detection
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.
Most trading-bot websites lead with returns. Yearly backtest, 99% drawdown-free, six-figure passive income, you get it. We've deliberately not built that page. This post is the inverse: what HMM Trade definitely is not, what it won't do for you, and what we explicitly refuse to claim.
We're a tool, not a guarantee. If you came here for guarantees, we'd rather lose you as a customer than lie to you.
It's not a way to predict the market
Volatility regime detection is about recognizing the present, not predicting the future. The HMM tells you whether right now looks like a calm uptrend or a crisis crash; it does not tell you what tomorrow brings. Strategies that act on the regime label make sizing + entry decisions based on what kind of market this is, not the market we hope is coming.
If you want a system that says “NVDA goes to $400 by Friday,” we're not it. If anyone says they have such a system, run.
It's not a high-frequency trading bot
Our default cadence is daily bars on stocks, hourly bars on crypto. The bot sees a closing bar, computes regime, decides positions for the next bar. We don't do tick-level execution, we don't trade off the orderbook, we don't care about millisecond latency. The strategies running on top wouldn't benefit from finer granularity even if we provided it — regime classification stabilizes on multi-bar windows, not within-bar microstructure.
It's not a substitute for a financial advisor
We don't know your tax situation, your retirement timeline, your risk tolerance, your existing portfolio, or your income. We don't personalize the bot to those things and we don't pretend to. If you have meaningful capital at stake, talk to a fiduciary financial advisor first. They cost money. They're still cheaper than the mistakes you won't make once you've had one good conversation.
Even if the bot makes you money, that's a small part of getting your finances right. Asset allocation, tax-loss harvesting, retirement contributions, insurance — none of which an HMM model has any opinion about.
It's not a backtest oracle
Look closely at our pricing page and you won't find a backtest curve. Three reasons:
- Backtests on retail-accessible strategies are almost always overfit. They show what worked on a specific historical window, often the same window the strategy was developed on. Out-of-sample, the curves shrink.
- The numbers people quote (Sharpe, max DD, win rate) mostly come from the period 2013-2019, which was an anomaly. Show me your bot's 2008 or 2020 March performance and I'll trust the rest.
- Even with an honest backtest, future returns aren't past returns. Volatility regimes shift, market microstructure changes, fee structures evolve. Backtests are a sanity check, not a forecast.
We do publish the BIC and log-likelihood of every model in the catalog, and the regime distribution + transition matrix on the bot's detail page. Those are model-quality metrics, not return promises.
It won't save you from an undisciplined trader
The bot does what we tell it. It does not adapt to your psychology when you watch it lose three days in a row and decide to disable the risk caps because you're “sure” the next trade will hit. It does not stop you from connecting a live brokerage account before paper-testing for a month. It does not stop you from increasing the position size cap from 5% to 25% because the conservative cap “leaves money on the table.”
Every one of those is a thing we've seen retail traders do. The bot can't protect you from yourself; it can only enforce the rules you set when you were thinking clearly.
It's not magic
The math under HMM Trade is well-understood, forty-seven-year-old statistics. We didn't invent it. The edge — to the extent there is one — comes from disciplined application: clean feature engineering, conservative model selection (BIC, not just maximum likelihood), regime-aware sizing, layered risk caps, honest validation gates that block bad fits before they reach the fleet.
These are all engineering decisions, not insights. Anyone with a few months of focus could replicate the architecture. We're a hosted version that does the engineering for you so you can focus on whatever else you want to focus on. That's the value, and that's all the value.
What it actually is
It's a trading bot that classifies market volatility into regimes, sizes positions according to the regime, refuses bad trades via layered risk caps, and runs on hosted infra so you don't have to. It's a tool. A useful one, we hope — but a tool, not a future-printer.
If that sounds underwhelming, good. The trading bots that sound underwhelming are the ones that don't blow up. The ones that sound exciting are usually selling you the backtest, not the bot.
The pricing-page disclaimers, repeated here
- Past performance is not indicative of future results. (Real disclaimer that we mean.)
- Trading involves risk. You can lose more money than you start with, especially with leverage. (Same.)
- We don't provide investment advice. Nothing on this site or in the product constitutes a recommendation to buy or sell any security. (Same.)
- We're not registered as an investment advisor. We're a software vendor. The distinction is meaningful and we take it seriously. (Same.)
If after all this you still want to spin up a paper bot, the free tier is on us. If you decide we're not for you — also fine. We'd rather you make an informed call than the other thing.