ChatGPT & Claude for Trading Indicators — Complete Tutorial

You used to need months learning Pine Script to build a custom indicator. Now you describe what you want in plain English, paste the response into TradingView, and have a working indicator in 2 minutes. This tutorial shows exactly how to do it well — including prompts that produce production-quality code on the first try.

ChatGPT vs Claude for Pine Script

Both AI tools can produce Pine Script. After testing dozens of indicators on both:

The Anatomy of a Good Prompt

The single biggest factor in code quality is prompt structure. A bad prompt gives bad code. Here’s the formula we use:

1. Role: “You are a Pine Script v5 expert.”
2. Goal: “Create an indicator that…”
3. Specifications: exact parameters (periods, thresholds, colours)
4. Display: what should be plotted, where (overlay or separate pane)
5. Alerts: what conditions trigger alerts
6. Constraints: “use only built-in functions, no external libraries”
7. Output format: “Return complete .pine file ready to paste into TradingView”

Prompt Template (Copy-Paste Ready)

You are a TradingView Pine Script v5 expert.

Create an indicator with these specifications:
- Name: [DESCRIPTIVE NAME]
- Purpose: [WHAT IT DOES IN ONE SENTENCE]
- Indicators used: [LIST: RSI, EMA, Bollinger Bands, etc.]
- Parameters (inputs): [LIST WITH DEFAULTS]
- Plot location: overlay=[true/false]
- Plotted items: [WHAT LINES/SHAPES TO DISPLAY]
- Alert conditions: [WHEN TO FIRE ALERTS]

Constraints:
- Use only built-in Pine Script v5 functions
- All inputs must be configurable via input.* functions
- Add inline comments explaining each section
- Make it work on any timeframe and any asset

Return the complete .pine file, ready to paste directly
into the TradingView Pine Editor.

Example 1: RSI with Smart Alerts

Let’s build a real indicator. Here’s the prompt:

Result (paste this into TradingView’s Pine Editor):

//@version=5
indicator("RSI Smart Alerts", overlay=false)

// === INPUTS ===
rsiLen = input.int(14, "RSI Length", minval=2)
osLevel = input.int(30, "Oversold Level")
obLevel = input.int(70, "Overbought Level")

// === CALCULATIONS ===
rsi = ta.rsi(close, rsiLen)

// === PLOTS ===
plot(rsi, "RSI", color=#E040FB, linewidth=2)
hline(obLevel, "Overbought", color=color.new(#FF5252, 50))
hline(osLevel, "Oversold", color=color.new(#00E676, 50))
hline(50, "Mid", color=color.new(color.gray, 70), linestyle=hline.style_dotted)

// === BACKGROUND COLOURING ===
bgcolor(rsi > obLevel ? color.new(#FF5252, 90) : na, title="OB Zone")
bgcolor(rsi < osLevel ? color.new(#00E676, 90) : na, title="OS Zone")

// === ALERTS ===
bullishCross = ta.crossunder(rsi, osLevel)
bearishCross = ta.crossover(rsi, obLevel)
alertcondition(bullishCross, "RSI Bullish",
  "RSI crossed below {{plot_0}} — potential BUY")
alertcondition(bearishCross, "RSI Bearish",
  "RSI crossed above {{plot_0}} — potential SELL")

Example 2: Multi-Indicator Strategy

More complex prompt for trading strategy:

//@version=5
indicator("EMA + RSI Combo — IA Trader Pro", overlay=true)

// === INPUTS ===
emaFast = input.int(9, "EMA Fast")
emaSlow = input.int(21, "EMA Slow")
rsiLen = input.int(14, "RSI Length")
rsiMid = input.int(50, "RSI Midline")

// === CALCULATIONS ===
ef = ta.ema(close, emaFast)
es = ta.ema(close, emaSlow)
rsi = ta.rsi(close, rsiLen)

// === PLOTS ===
plot(ef, "EMA Fast", color=#00E676, linewidth=2)
plot(es, "EMA Slow", color=#FF5252, linewidth=2)

// === SIGNAL CONDITIONS ===
emaCrossUp = ta.crossover(ef, es)
emaCrossDn = ta.crossunder(ef, es)
buySignal = emaCrossUp and rsi < rsiMid
sellSignal = emaCrossDn and rsi > rsiMid

// === ARROWS ===
plotshape(buySignal, "BUY",
  location.belowbar, color=#00E676,
  style=shape.triangleup, size=size.small, text="BUY")
plotshape(sellSignal, "SELL",
  location.abovebar, color=#FF5252,
  style=shape.triangledown, size=size.small, text="SELL")

// === STATUS TABLE ===
var table dash = table.new(position.top_right, 2, 3,
  bgcolor=color.new(color.black, 20),
  border_width=1, border_color=color.gray)
if barstate.islast
    trend = ef > es ? "BULL" : "BEAR"
    trendCol = ef > es ? #00E676 : #FF5252
    table.cell(dash, 0, 0, "EMA", text_color=color.white, text_size=size.tiny)
    table.cell(dash, 1, 0, trend, text_color=trendCol, text_size=size.tiny)
    table.cell(dash, 0, 1, "RSI", text_color=color.white, text_size=size.tiny)
    table.cell(dash, 1, 1, str.tostring(rsi, "#.#"),
      text_color=color.white, text_size=size.tiny)

// === ALERTS ===
alertcondition(buySignal, "EMA+RSI BUY",
  "EMA crossover up + RSI < 50")
alertcondition(sellSignal, "EMA+RSI SELL",
  "EMA crossover down + RSI > 50")

Pitfalls of AI-Generated Code

How to Debug AI Code

If the code doesn’t compile in Pine Editor (red error), do this:

1. Copy the error message from TradingView (bottom of editor)
2. Paste back to AI: “This code returns the following error: [paste error]. Fix it and explain what was wrong.”
3. Iterate until it compiles
4. If 3+ iterations fail, simplify the request and start over

Claude in particular is excellent at debugging — paste broken code and ask “what’s wrong with this?”. Usually gets it right on first try.

Backtest Before Trusting AI Indicators

Common Indicators You Can Generate

AI can generate any of these in seconds with proper prompts:

• Trend: EMA Crossover, MACD, SuperTrend, Heikin Ashi
• Momentum: RSI, Stochastic, ROC, Awesome Oscillator
• Volatility: Bollinger Bands, ATR Channels, Keltner Channels
• Volume: OBV, Volume Profile, VWAP, Accumulation/Distribution
• Levels: Pivot Points, Fibonacci, Support/Resistance Auto
• Patterns: Engulfing detection, Doji finder, Inside Bar
• Combos: Any of the above combined
• Custom logic: “Buy when X AND Y AND Z” — describe it, get it

See our 10 Pine Script Indicators library for production-ready examples.

What AI CAN’T Do (Yet)

Free Tier vs Paid — SA Cost Analysis

At ~R18.50/USD, here’s what each tier costs SA traders monthly:

• ChatGPT Free (GPT-3.5): R0 — sufficient for simple indicators
• ChatGPT Plus (GPT-4): $20/mo ≈ R370/mo
• Claude Free: R0 — sufficient for most Pine Script tasks
• Claude Pro: $20/mo ≈ R370/mo
• API access (developer): pay-per-token, ~R5-R20 per 1000 generations

If you generate 1-2 indicators per week, free tier is enough. Heavy users (10+ per week) benefit from Pro plans for higher message limits.

Iterative Prompting (Pro Technique)

Don’t try to get the perfect indicator in one prompt. Iterate:

1. Prompt 1: “Create basic RSI indicator with alerts.”
2. Prompt 2: “Add background colouring when overbought/oversold.”
3. Prompt 3: “Now add a status table showing current RSI value.”
4. Prompt 4: “Convert to strategy script with 2% position sizing and 50-pip stop loss.”
5. Prompt 5: “Add multi-timeframe RSI (show H1 RSI on M5 chart).”

Each iteration builds on the last. AI maintains context. End result is far better than asking for everything at once.

Generate Trading Strategies, Not Just Indicators

Once comfortable with indicators, level up to strategy scripts. These actually place simulated trades and let you backtest. The key prompt:

Paste your indicator first, then this prompt. AI converts it. Open in Strategy Tester to see realistic backtest results.

Privacy & Data Considerations

SA-Specific Use Cases

AI prompts particularly useful for SA traders:

• JSE stock analysis: “Analyse this Naspers price action: [paste OHLC]. Identify trend, key levels, possible setups.”
• ZAR pair indicators: “Build indicator for USD/ZAR forex pair. Range-bound strategy with breakout filter.”
• Tax categorisation: “Help categorise these trades for SARS: [paste list]. Income vs CGT?”
• Education: “Explain Bollinger Bands in simple terms for a SA retail trader who knows RSI.”

Save Your Best Prompts

Build a personal prompt library. Save prompts that produce good results. We use a Google Doc with categories: Indicators, Strategies, Debugging, Conversions, Explanations. Over time, you’ll have 20-30 reusable prompts that save hours.

What’s Next

1. Try our 5 Specific AI Prompts
2. Build indicators with our 10 Pine Script library
3. Backtest properly: Backtesting guide
4. Move to Python automation: Python + Deriv API
5. Test live: Deriv Bot setup

Related Reading

• 5 Specific Prompts for Traders
• 10 Pine Script Library
• Backtesting Guide
• Python + Deriv API
• Deriv Bot Setup