Live Engine
Arm a strategy on a symbol and let it trade forward on paper, journalling every decision.
A backtest is a claim about the past. The engine is how that claim gets tested going forward, on candles nobody has seen yet, which is the only sample that was definitely not used to choose the rules.
It evaluates the armed strategies at every bar close, opens simulated positions when the entry conditions fire, manages them to the stop or through the target ladder, and writes an entry explaining what it did and which condition triggered it.
Paper only, and that is a structural fact rather than a setting: there is no exchange client anywhere in this codebase and no code path that can place an order. Fills are simulated against real prices.
What it will do
Pick a rule set and a market, and it is evaluated at each close of that strategy’s timeframe until you disarm it.
One evaluator, one trade planner, one position state machine, shared by both runners — so live results and backtest results cannot diverge because of an implementation difference.
Each position records why it opened, what the stop and targets were, and how it closed. A log of entries without reasons is unreviewable.
It cannot run while the machine is asleep, and it will not backfill an entry it was not present for. A late fill at a price the rules never saw would quietly flatter every result on the page.
On a public deployment this stays off by design. The engine is a single shared book of positions, so one visitor arming a strategy would change what every other visitor sees — it is meant for a copy of the app running on your own machine.