Example strategies
Worked prompts, what the builder does with each, and what to notice. Every figure on this page is illustrative, none of these are real results.
1. The minimum viable request
eth 1m bullish
Three of the six inputs. The builder records them and asks for what is missing rather than guessing.
What it asks for: risk level, and any preference on indicators.
What to notice: you can start this terse. Each missing input costs one turn, and the builder asks for the specific thing rather than restarting.
2. A complete single-instrument request
create a long-only strategy on RELIANCE, 5 minute candles, entry when RSI(14)
crosses above 55 and price is above the 20 EMA, exit when RSI drops below 45,
1.2% stop-loss, risk 1% per trade, max 3 trades a day
Everything stated. No clarifying turn needed.
What to notice:
- Every field except take-profit is marked stated. Take-profit shows as defaulted, and the read-back says so.
max 3 trades a daybecame a strategy control. That is a build-time control; the matching execution gate is separate and off unless you configure it. See Risk defaults.- Both conditions carry numeric thresholds, so there is no repair pass.
3. Crypto, continuous session, aggressive
create a bullish momentum strategy for ETH/USDT, 1m, aggressive
Entry: RSI(14) above 60 AND EMA(9) above EMA(21). Exit on condition reversal. Trailing take-profit arms at 1.5%, trails 0.8%. Stop-loss 1.2% from entry. Per-trade risk 2.0% (calibrated to "aggressive"). Session: continuous.
I defaulted the stop, target and per-trade risk from your risk level, state them explicitly if you want different values.
What to notice:
- "Aggressive" calibrates defaults; it does not set them. The 2% per-trade risk is a defaulted value derived from your risk level, and the read-back says so.
- "Bullish momentum" resolved to two indicators. That resolution is the platform's, not yours. If you want specific indicators, name them.
- Crypto gets a continuous session with no trading window, because the venue never closes.
4. A request that needs a threshold
buy RELIANCE when price is near the 20 EMA on strong volume
Two MISSING_THRESHOLD gaps in one sentence: near and strong.
What happens: the builder identifies both spans and asks. Note that it correctly reads the
20 in "20 EMA" as the indicator's period, not as the threshold. So it knows the threshold
is genuinely absent rather than treating the span as complete.
The version that compiles first time:
buy RELIANCE when price is within 0.5% of the 20 EMA and volume is above
1.5 times the 20-period average volume
5. A request that is refused, and why
buy RELIANCE when RSI crosses above 55 unless volume is falling
AMBIGUOUS_SCOPE. Never retried.
"Unless" could negate the whole entry, or only the RSI clause. Both readings produce a valid strategy and one of them is inverted. And nothing in the backtest would tell you which you got.
What to say instead:
buy RELIANCE when RSI(14) crosses above 55 and volume is rising
Positive form. One reading.
6. Stating a backtest period
backtest the last 6 months
Works, because the state given to the model carries today's date. Absolute ranges work too:
backtest from 2025-01-01 to 2025-06-30
And you can override capital and costs in the same breath:
backtest the last 3 months with 5 lakh starting capital
See Choosing a test period for the default window and the 1 January 2024 floor.
7. Multiple symbols: not a portfolio
backtest this on ETH and BTC
Runs the same strategy independently on each symbol, each with the full starting capital. Two separate results you can compare side by side.
It does not model allocating capital across both, does not net exposure, and does not compute a combined equity curve. Each run is a standalone experiment on one instrument. If you want one strategy trading a selected set of instruments, that is a Dynamic strategy, a different thing.
8. A dynamic universe
build a momentum strategy across all large-cap crypto
You gave an instrument selection rule rather than an instrument. The builder recognises that and resolves a dynamic universe instead of asking you for a single asset.
Dynamic strategies evaluate every asset in the universe on each candle and trade the top-ranked ones. See Dynamic & universe strategies.
9. Using alternative data
build a strategy on RELIANCE that only buys when the current ratio is above 1
and revenue is growing
"Current ratio" and "revenue" are not price indicators. The builder searches the tenant's custom and alternative-data indicator catalog (fundamentals, corporate events, macro, on-chain, news) and resolves real indicator codes.
It never invents a code. If a metric is not in your entitled catalog, it says so.
See Indicators.
10. Options: buildable, backtestable, not live
build a long straddle on NIFTY, monthly expiry, ATM
Multi-leg options structures are supported by the builder. The symbol is the underlying
(NIFTY), and the derivative detail (instrument type, right, strike rule, expiry rule) is
captured alongside it.
You can build this, backtest it, and run it in paper. You cannot trade it live. The live exit path is not implemented, so live entry is blocked rather than left reachable. See Derivatives.
11. A request that gets the safety boundary
which stock should I buy today?
Routed to the safety boundary. You get a composed reply explaining that buy/sell calls are out of scope, plus the supported universe and an invitation to build a strategy on an instrument you choose.
Rephrasing does not change the outcome, the boundary reply is a template, not a generation. See What the AI will not do.
12. An educational question
what is an opening range breakout?
Answered concisely, with no backend action. Education is explicitly in scope; the AI is instructed to explain rather than to just assemble.
Patterns worth copying
| Pattern | Example |
|---|---|
| Name the indicator and the comparison | "RSI(14) crosses above 55" |
| Give every vague word a number | "within 0.5%", "1.5× average volume" |
| State rules positively | "volume rising", not "unless volume falls" |
| State risk controls explicitly if they matter | "1.2% stop, risk 1% per trade" |
| Put the period in the backtest request | "backtest the last 6 months" |
| Correct the builder rather than starting over | "make the stop 0.8% instead" |
Next
- Creating with AI, the builder end to end
- Indicators, the full vocabulary
- AI limitations