How Stretus works
Every strategy on Stretus moves through the same five stages. This page walks each one, says what the platform decides and what you decide, and names what can go wrong.
The lifecycle
| Stage | What happens | What you do |
|---|---|---|
| 1. Design | Plain language becomes a structured strategy object | State the rule |
| 2. Validate | The strategy is backtested with real costs applied | Read the result and decide |
| 3. Allocate | A capital ceiling is set for the strategy | Choose the amount |
| 4. Execute | The strategy runs in paper or live, isolated from each other | Choose which mode, and when |
| 5. Monitor | P&L, open positions and risk state are tracked | Watch it, and stop it when needed |
Nothing advances on its own. Every stage begins with an action you take.
1. Design
You describe intent. The AI asks clarifying questions and assembles a complete strategy object.
What gets assembled:
| Element | Example |
|---|---|
| Instrument and timeframe | RELIANCE, 5 minute |
| Direction | long only |
| Entry condition | RSI(14) crosses above 55 AND close > EMA(20) |
| Exit condition | RSI(14) < 45 |
| Stop-loss | 1.2% from entry |
| Take-profit | Fixed, laddered, or trailing |
| Position sizing | 1% risk per trade |
| Trading window | 09:15-15:30 IST |
| Gates | Max trades per day, cooldown after loss |
What the platform decides: which indicators express a qualitative phrase like "bullish momentum", and default values for anything you did not state. Both are labelled.
What you decide: the rule, and whether the assembled version says what you meant.
What can go wrong: an ambiguous phrase. The builder asks rather than guessing, see AI limitations. Read the read-back before you accept it.
See Creating with AI and AI engine.
2. Validate
The strategy runs over real historical data. Results are graded.
What you get: total return, Sharpe ratio, max drawdown, win rate, profit factor, trade count, a monthly breakdown, every individual trade with its entry, exit, P&L and reason, and a six-grade assessment (A, B+, B, C+, C, D).
Costs are applied. Modelled slippage, per-venue commission, and in Indian markets STT, stamp duty, exchange transaction charges, the SEBI turnover fee and GST. This is not a frictionless simulation, and the cost model is versioned so results from different statutory rates can be told apart. See Fees & charges.
The window: by default 1 January 2024 to now, a fixed floor, the same for every timeframe and asset class. You can state your own period in the chat. See Choosing a test period.
What can go wrong: the run produces zero trades. Usually an unreachable condition, the validator catches the structural causes, and Backtest limitations covers the rest.
See Backtesting, Strategy grades and Reading results honestly.
3. Allocate
You assign the strategy its own slice of capital.
The property that matters: each strategy is isolated. It cannot exceed its allocation and cannot borrow from another strategy. Two of the nine risk gates enforce this and cannot be disabled.
What you decide: the amount. Size it against a drawdown you can sit through, not against the return you want, the allocation is the only control that scales your loss directly.
What can go wrong: allocating to the return rather than the drawdown. A 22% max drawdown on a ₹10 lakh allocation is ₹2.2 lakh. Decide whether that sentence is comfortable before you deploy, not after.
See Capital model.
4. Execute
Two independent paths.
Paper
Filled by the internal simulator against live prices. Nothing leaves the platform, no broker connection is needed, no capital is at risk.
Available immediately after a backtest.
Live
Real orders to your own connected broker or exchange.
| Market | Live requires |
|---|---|
| Indian equity | A connected broker and per-strategy broker approval |
| Crypto spot | Valid exchange API keys, no approval gate |
| Futures & options | Live entry is refused. Paper only |
If you deploy live with valid keys, the strategy starts placing real orders on the next candle close. There is nothing between you and the market.
Both paths run the nine risk gates identically. That is what makes a paper run a real test of your configuration and not just of the strategy.
Isolation: separate capital, separate positions, separate history. A position opened in one mode never migrates to the other.
What can go wrong: the risk limits you assumed exist do not. Seven of nine are off until you configure them. See How risk is enforced.
See Paper vs live, Order lifecycle and Connect a broker.
5. Monitor
P&L, open positions, trade history and risk state, in real time.
Where: the dashboard for the portfolio view, the strategy detail page for one strategy, the live terminal for order flow, and the algo execution console for operational detail. Optional Telegram notifications push events to you.
Controls:
| Action | Effect |
|---|---|
| Pause | No new entries. Open positions run to their exit conditions |
| Resume | New entries re-open |
| Stop | No new entries. Open positions are not closed, they run to their exits |
Neither pause nor stop closes an open position. To be flat, close the position at your broker.
What can go wrong: assuming a strategy that stopped entering has broken. Usually it has hit a risk gate, or it is outside its trading window. See Alerts & breaches.
See Monitoring.
What the AI actually does, in one example
You say:
"create a bullish momentum strategy for ETH/USDT, aggressive"
It resolves:
| Question | Resolution |
|---|---|
| Which indicators express "bullish momentum"? | RSI above a threshold, and a short EMA above a longer one |
| What constitutes entry? | All conditions true simultaneously, evaluated on candle close |
| How to exit? | Condition reversal, a trailing take-profit, or both |
| How much to risk per trade? | Calibrated from "aggressive", and labelled as a default |
| What trading window? | Continuous, because crypto spot never closes |
Then you iterate in plain language ("tighter stop", "reduce the drawdown", "less aggressive") and it adjusts and can re-run the backtest.
The resolutions marked as defaults are the platform's, not yours. If a control matters to you, state it. See How the AI reads your words.
Your capital, throughout
Funds stay in your broker or exchange account for the entire lifecycle. Stretus places orders using credentials you connect; you keep custody. See Data, custody & security.