Platform

Strategies

Playbooks on /strategies go from templates and Studio rules → Simulate → paper → live. This page describes what ships today — not a wishlist.

Building a playbook

Start from a template (DCA, grid, momentum, mean reversion, blank) or open Edit rules for Python Studio. There is no drag-and-drop canvas yet — templates plus Studio (and Ask my agent) are the path.

Ask my agent

Ask my agent and the in-page AI tab help you draft and explain rules. When Studio rules are honored by the engine, agent-drafted code is what Simulate runs — not a separate theatre path.

Risk rules

Configure stop-loss and take-profit (percent, points, or absolute), trailing / lock-and-trail, session windows, opening-range breakout, and re-entry modes on the playbook. Live requires both stop-loss and take-profit.

Simulate

Standard — historical backtest on daily bars with commission and slippage (Compute Units reserved from your simulation pack).
Custom — same backtest plus accelerated Monte Carlo paths (TPU when available).

Results include return, Sharpe / Sortino / Calmar, drawdown, win rate, verdict, overfitting meter, parameter sweep and cost-sensitivity studies, and a trade log with CSV export. You can compare multiple past runs side by side.

Backtest vs live fills

Simulations typically evaluate and fill on bar close. Paper and live may fill on later prices (and on a ~15 minute tick today). Treat a strong backtest as evidence, not a promise of identical live fills.

Deploy: sandbox, paper, live

Sandbox — rules and UI without broker orders (config flag; does not wipe your strategy).
Paper — executor ticks on a market-hours schedule against live prices with a paper broker.
Live — gated by backtest, stress, paper days, and filled trades. Orders are staged for approval unless you choose auto mode.

Execution modes: alert (notify only), confirm (propose then approve), auto (paper-style auto when allowed).

Pause, manual, flatten

These are different actions:

Pause — stop checking conditions; positions stay.
Listening — check conditions, do not open new entries (forward-test style).
Switch to manual — disconnect the algo; positions stay at the broker for you to manage.
Pause all / Stop all / Flatten all — account-wide panic controls (stop cancels working orders; flatten requests closing positions).

Account-level daily loss and target are persisted on your account — they do not need to be re-entered every morning.

Marketplace

Listing and purchasing strategies is a separate flow at /strategies/marketplace. Plugins and sandboxed analytics live under Plugins.