Short, because the important parts are short.
Last updated: 2 September 2026.
We are infrastructure, not advice. backtrade.dev replays historical market data and simulates orders against it. We do not provide strategies, signals, recommendations or financial advice of any kind, and nothing here is an offer to buy or sell anything.
You own the outcome. Your prompts, your models, your strategies and your results are yours. Whether anything you build here makes or loses money is your responsibility, and we accept none of it — simulated or live.
Bar replay fills an order in full or rejects it under its prior-minute liquidity
budget; it does not model partial fills or observe intraminute spread. Supported
sip-partial-v2 tick sessions model partial fills using recorded liquidity
and quote spreads. Legacy sip-full-v1 tick sessions wait for a single
quote with enough displayed size to fill the entire order; they do not fill it partially.
There is no order-book queue position, no market impact and no auction simulation.
Configured basis-point costs are already reflected in cash and net P&L. They are
embedded in market-order fill prices; any portion constrained by a limit price is
charged separately to cash. Do not subtract the reported total cost again from net P&L. Simulated fills do not establish executable live returns.
Past performance of a historical session says nothing about the future, and a simulated result says less than that.
The engine refuses to show a model information it could not have had at that instant. A bar arrives only once its interval has closed; a market order fills at the first interval open strictly after submission time (13:40:00 submission → 13:41 open on a contiguous minute tape); the account marks from completed bars only. We describe this in the quickstart. Limits and stop activation require intervals starting strictly after submission; limits are reported at interval completion. A coherent moving-price bar with zero reported volume supplies no volume-derived cap, because its volume is unknown; buying power and other checks still apply. This does not verify real liquidity. We do this because a backtest that leaks is worse than no backtest — not because it makes the results tradeable.
Trial credentials are free, time-limited and session-limited, issued without an account. They are rate-limited per source. Do not attempt to circumvent those limits, resell access, or use the service to attack it or anyone else. We may revoke a credential at any time.
You may create a password account to keep account-bound credentials and completed run history. An account is optional: free anonymous trials remain available without an email, password or account. Creating an account does not turn a simulation into advice, transfer responsibility for a strategy, or promise paid access.
A successful self-service deletion removes Backtrade-owned password and recovery authority, browser sessions, account keys, grants, usage and credential audits, completed runs, device links, and free billing-account ownership.
Deletion is refused while Stripe or payment state, checkout records, subscriptions, or other linked authority still needs to be resolved. Those cases require cancellation or support rather than silently deleting a financial relationship. Stripe and payment records that must be kept for legal, accounting or financial reasons are handled through that process and are not silently erased.
The shared users identity may remain after Backtrade access is removed if
another application refers to it, or if shared email-verification or terms-acceptance state
depends on it. A retained shared identity carries no remaining Backtrade password, session
or credential authority.
Market data and headlines are licensed from third-party vendors for replay within this service. You may use what a session delivers to drive and evaluate your own agent. You may not redistribute or resell the underlying data as a dataset.
The service is provided as is, without warranty of any kind. It is a beta. It may be interrupted, changed or withdrawn, and it may contain defects — when we find one that would have changed a result, we say so publicly rather than quietly patching it.
We will date material changes on this page rather than silently editing it.