strategy-dev

How to write a great AI Builder prompt

Describe a strategy in plain English and get a config that loads, backtests and runs exactly as you meant it

10 min · beginner

What you'll have when finished

  • Know the 6 parts every good strategy prompt includes
  • Know what the AI Builder can and cannot build, so nothing gets silently dropped
  • Have 9 copy-ready prompts from beginner to advanced
  • Refine a strategy in follow-up messages instead of starting over

Before you start

  • The AI Builder builds what you describe. It does not pick a winning strategy for you, and nothing it builds is financial advice.
  • Always backtest at your exchange's real fee across at least two different periods, then paper-trade, before using real money.

Walkthrough

  1. The 6 parts of a good prompt

    A prompt that includes all six loads cleanly, with no guessing: 1. **Pair**: BTC/USDC, ETH/USDC, SOL/USDC… 2. **Timeframe**: 15m, 1h, 4h or 1d. Slower timeframes lose less to fees. 3. **Entry**: when to buy, e.g. "RSI below 30 AND EMA20 above EMA50". 4. **Size**: how much per trade, e.g. "10% of my balance". 5. **Exits**: take profit and stop loss in %, e.g. "take profit 6%, stop loss 3%". 6. **Limits**: max trades per day, daily loss limit, and a drawdown % that pauses the bot. Leave one out and the AI either asks you about it or fills in a conservative default. Either way, you'll see exactly what it chose before anything is saved.

    Success criteria: Your prompt names a pair, timeframe, entry, size, exits and limits

  2. What the AI Builder can and cannot build

    **Can build:** - RSI (14), EMA 9 / 20 / 50, price, Bollinger Band width - Price vs an EMA or a Bollinger band ("price above EMA50", "price crosses above the upper band") - EMA crossovers ("EMA20 crosses above EMA50") - Volume vs its recent average ("volume at least 1.5× average") - AND / OR logic - Market buy or DCA buy with safety orders - Take profit, trailing take profit, stop loss, trailing stop - Daily trade, daily loss and drawdown limits **Cannot build (yet):** - Short selling or leverage. Strategies are long-only spot. - MACD, ATR, VWAP, or custom periods like "RSI 7" or "EMA 200". - Indicator-based exits like "sell when RSI hits 70". Exits are take-profit / stop-loss percentages. - Time schedules like "every Monday". Use the DCA Bot for that. If you ask for something on the "cannot" list, the AI builds the closest version and explains each difference in the warnings, so read them.

    Success criteria: You can tell whether an idea fits the Builder before you type it

  3. Beginner prompts (copy and paste)

    **Dip buyer:** "Buy BTC/USDC on the 4h chart when RSI is below 30. Use 10% of my balance per trade, take profit at 6%, stop loss at 3%, max 2 trades per day." **Trend follower:** "Buy ETH/USDC on the daily chart when EMA20 crosses above EMA50. Use 15% per trade, trailing take profit that arms at 8% and sells after a 3% pullback, stop loss at 5%." **DCA dip ladder:** "On ETH/USDC 4h, start a DCA when RSI drops below 35: first buy 5% of my balance, then 3 safety orders every 3% lower. Take profit 4% above my average price, stop loss 15%."

    Success criteria: One beginner prompt loaded and backtested

  4. Intermediate prompts

    **Dip buyer with a trend filter:** "On BTC/USDC 4h, buy when RSI is below 35 AND EMA20 is above EMA50. 15% per trade, take profit 6%, stop loss 2.5%, max 2 trades per day." **Volume breakout:** "On ETH/USDC 4h, buy when price crosses above the upper Bollinger band AND volume is at least 1.5 times its average. 10% per trade, take profit 8%, stop loss 3%." **Pullback in an uptrend:** "On SOL/USDC 1h, buy when price is below EMA20 but above EMA50 and RSI is below 45. 10% per trade, take profit 4%, stop loss 2%, daily loss limit 3%."

    Success criteria: Compared a filtered strategy against its unfiltered version in backtest

  5. Advanced prompts

    **Squeeze breakout:** "On BTC/USDC daily, buy when Bollinger width is below 0.04 AND price crosses above EMA20 AND volume is at least 1.5 times average. 20% per trade, trailing take profit that arms at 10% and sells after a 4% pullback, trailing stop 6%. Pause the bot after a 12% drawdown." **Trend stack:** "On ETH/USDC 4h, buy when EMA9 is above EMA20, EMA20 is above EMA50, RSI is above 50 and below 70, and volume is at least 1.3 times average. 15% per trade, trailing take profit at 6% with a 2.5% pullback, stop loss 3%, max 1 trade per day." **Mean reversion inside a trend:** "On BTC/USDC 1h, buy when price closes below the lower Bollinger band, RSI is below 30, and EMA20 is above EMA50. 10% per trade, take profit 3%, stop loss 2%, max 3 trades per day, drawdown limit 8%."

    Success criteria: Read every warning the AI returned before saving

  6. Refine instead of starting over

    After the AI builds a strategy, keep talking to it. Each message changes the current strategy: - "Make it more conservative." - "Change the timeframe to 1d and widen the take profit to 10%." - "Only buy when EMA20 is above EMA50." - "Lower the position size to 5% and add a 3% daily loss limit." Not sure where to start? Ask: "What makes a strategy lower-risk?" The AI explains the trade-offs and builds a starter example you can backtest. Then: **Use this configuration** → Save → Backtest at your real fee on two periods → Paper trade for a week → only then consider Go Live with a small budget.

    Success criteria: Refined a strategy at least once · Backtested it on two different periods

What's next

Compare what you built against the Templates tab. Every template shows its backtest scorecard, so you can see how similar ideas performed on real prices. Then follow [Paper trade → live capital](/learning-library?guide=paper-to-live-30-day-plan) before putting real money behind anything.