Algo
Write a strategy down, prove it on history, then let it trade on paper.
The idea behind this section is that a trading rule you cannot state precisely is a rule you cannot test. So a strategy here is an explicit spec — an entry condition, a stop, a target ladder and a risk budget — rather than a feeling about a chart.
Once a rule is written that way, three things become possible that are not possible otherwise: it can be run over years of history, it can be compared against every other rule on identical data, and it can be handed to an engine that executes it the same way every time without getting bored or scared.
It is being built so the backtester and the live engine share one evaluator and one position state machine. That constraint matters more than any feature on this page: a backtest is only evidence about live behaviour if the two cannot quietly drift apart.
What it will do
Rule sets stated in full, in English, before anything runs them. Fourteen ship with the app as worked examples of the grammar.
Every fill pays a taker fee and slippage, signals fill at the next bar open rather than the close they fired on, and a bar holding both stop and target is assumed to have hit the stop first.
Walk-forward splits and parameter sweeps, so you can see whether an edge survives being measured on data it was not chosen on.
No exchange client exists anywhere in this codebase and no code path places an order. Positions are simulated against real prices.
What each part will be
Nothing in this section is live yet, and none of it is advice. A profitable backtest is a description of the past, not a prediction — four years of crypto is roughly one cycle, which is a small sample for anything that depends on the regime it is in.