How we test and rate bots

Two kinds of post, labelled on the post itself

I can't run live capital through every platform I write about, and I don't think you should have to guess which ones I did. So every post that reaches a verdict carries a label near the title telling you what the conclusions rest on.

• Hands-on test with real capital: I funded an account and ran the bot. Any P&L figure in the post is money that actually moved. The pair, the allocation, the settings and the window are stated so you can judge the context.
• Documentation and fee analysis: no live test behind it. The post works through the platform's own documentation, fee schedules and disclosures, does the arithmetic, and links the primary sources so you can check every number yourself. Worked examples are labelled as worked examples, not results.

The second kind is still worth writing, because most of what determines whether a bot makes you money is knowable before you fund it: the fee per fill, how funding is charged, what happens when price leaves the range, what the subscription costs against your account size. But it is not the same as having run the thing, and it shouldn't be dressed up as if it were.

What a live test involves

When a post is labelled as a hands-on test, the bot ran for a defined period on a stated amount of capital, on my own accounts, not a demo. Paper trading hides the things that actually cost you: slippage, funding fees, fills that don't happen, and the psychological pull to override the bot. A bot that prints in a ranging month can bleed in a trending one, so I say which conditions it was tested in and where it would likely struggle.

What the rating actually means

Ratings are out of 10. They are not a vibe. A score weighs real performance in the test window, fees and total cost of ownership, how hard it is to set up and trust, and who the tool is genuinely for. A 9 would be rare. Most decent bots land in the 6 to 8 range because most bots are fine, not magic.

• 8 to 10: I'd run my own money on it and recommend it widely.
• 6 to 7.9: solid for a specific use case, with real caveats.
• 4 to 5.9: works, but something (price, reliability, fit) holds it back.
• Below 4: avoid, or only for a narrow edge case.

The affiliate part, stated plainly

Some links here are affiliate links, and I earn a commission if you sign up through them. That is how the site pays for itself. The score never bends because of a payout. I have rated bots I have an affiliate deal with below bots I don't, because the rating reflects the testing, not the commission. The full affiliate disclosure spells it out.

What would make me change a verdict

Bots change. Pricing tiers shift, features ship, platforms degrade. When something material changes, I update the review and note the date. If you think a verdict is stale or wrong, tell me on the contact page. I'd rather fix it than defend it.

Corrections

This page used to say that every review ran on live capital. That was broader than what the site actually did: a number of reviews were built from documentation and fee analysis rather than a funded account. Those posts have been relabelled and the figures in them corrected, and the per-post label above exists so the distinction is visible rather than assumed. If you read something here before that change and made a decision on it, I'm sorry, and the contact page reaches me directly.