Subscribe now

Haven’t read Part 1 yet? Start there first. This won’t make as much sense without it.

You’ve got an indicator on your chart and it looks good!
But here’s the question nobody asks until, “they’ve lost money” — does it actually work? That’s what a backtest answers.

Quick Recap: Indicator vs. Strategy

Before we start, one distinction worth nailing.

An indicator is observational. It doesn’t trade anything.

A strategy is the same logic, wrapped in a framework that TradingView can actually test. It has entry rules, exit rules, position sizing and a performance report you can read.

The indicator you built in Part 1 is already 80% of the way there. We’re just adding the other 20%.

Step 1: Open Claude and Load the Skill File

We’ll be using the exact same setup as Part 1. You can find the link to download the skill file below.

Important: Don’t forget to upload the entire .zip folder! The skill isn’t just a single file, and Claude needs the complete folder to understand the full context.
Download the skill File

Open a new Claude conversation, upload the Pine Script skill file (the full .zip), and wait for it to load. If you still have your Part 1 conversation open, you can continue there, but a fresh session with the skill file reloaded tends to give cleaner output.

In this session, we built a custom indicator inspired by our recent post on VWAP lines and the Supertrend! You can find the link to the original post below, along with a Google Drive link to download the indicator.
Download the indicator

“Here is a Pine Script indicator I built. I want to convert it into a strategy so I can backtest it in TradingView’s Strategy Tester. Use /pinescript to help me plan it.”

That’s the whole prompt. Claude takes it from there.

Step 2: Claude Asks Its Questions

In the previous post, we defined our entry conditions based on our indicators. In this one, we need to define our exit conditions so our strategy works well in the real market.
Also, Claude won't touch the code until it knows exactly what to build. This time, the questions are about how the strategy should behave, not just what it should show.

Step 3: Review the Strategy Spec

Before any code is written, Claude sends back a Strategy Spec a plain English summary of exactly what it’s about to build.

S .crossover(close, vwap) + direction < 0. Short: ta.crossunder(close, vwap) + direction > 0. Both gated by per-leg lock vars.
Exits: Opposite signal only. strategy.entry("Short", strategy.short) auto-reverses the open long (and vice versa) — no separate strategy.exit() needed since pyramiding = 0.
Filters: None by default. Optional NSE session filter input (9:15–15:30 IST) — off by default, user can enable.
Notes: process_orders_on_close=false (default) — entries fire at next bar's open, which is realistic for NIFTY futures. lookahead_on from the Earnings/Dividends/Splits anchor branches is stripped — those introduce future bias in backtests. alertcondition() calls removed (not valid in strategy scripts).

Read this carefully. This is your only checkpoint before Claude writes any code.

If everything looks right, reply: Looks good, go ahead.

Step 4: Paste Into TradingView

Once Claude gives you the strategy code:

1. Open TradingView
   

2. Click Pine Editor at the bottom of the screen

   

3. Clear the existing code and paste in the strategy code

   

4. Click Add to Chart
   A new panel will appear at the bottom of your screen labelled Strategy Tester.

Step 5: Backtest Setup

Most people open the Strategy Tester and jump straight to net profit. Don’t. The numbers mean nothing until you’ve told TradingView what assumptions to run them on.

Click the Settings icon on your strategy panel. You’ll see a Properties tab. Three inputs to set before anything else:

• Initial Capital: Enter what you’d actually deploy. ₹5L and ₹50L will produce very different profit figures, and neither is wrong. What matters is that the number reflects your real situation, so the results stay meaningful to you.

• Commission: For NIFTY futures, 0.03–0.05% per side is a reasonable starting point. Leave this at zero and every strategy looks better than it actually is.

Step 6: Read the Strategy Tester Report

This is the step most people rush through.

The report has two tabs. Here’s what actually matters in each:

Net Profit: The starting point, not the finish line. A positive number tells you the strategy didn’t lose money historically. That’s it. Check everything else before drawing conclusions.

Win Rate: Easy to misread. A 70% win rate looks strong until the losses are 3x larger than the wins. It only means something when you read it alongside Profit Factor.

Profit Factor: Total profit divided by total loss. Above 1.5 and the strategy is generating more than it’s giving back. Below 1.0, it lost money overall and no further analysis needed.

Max Drawdown: The worst stretch the strategy went through in the entire backtest period. Before you move on, ask yourself: if this was happening to your actual capital in real time, would you have stayed in? If the honest answer is no, the strategy isn’t right for you! regardless of what the net profit looks like.

Number of Trades: Sample size matters. A backtest with 15 trades over two years isn’t a backtest, it’s a handful of data points. You want at least 50–100 trades before the numbers carry any real weight.

List of Trades Tab

Click individual trades and watch where TradingView marks them on the chart. Ask yourself:

• Are entries happening where your logic says they should?

• Are exits cutting trades too early, or holding too long into reversals?

• Are there obvious bad trades that all cluster around one market condition (e.g., sideways chop)?

This is where you catch logic errors that the numbers alone won’t show you.

What You’ve Built

You took a trading idea. You turned it into a strategy. You checked whether it held up across historical trades.

In Part 3, we go one step further. We will create a telegram bot which messages you everytime your strategy gives a signal!

(Disclaimer: Any video/image/text content is for educational and informational purposes only and does not constitute financial advice. Refer - https://stocko.in/stocko-disclaimer/. Please do your own research or consult a qualified financial advisor before making any investment or trading decisions. Trading in stock markets involves the risk of loss.)

Subscribe now

Just a heads-up before we start, you can definitely do this on the free version of Claude without needing Pro subscription. Just keep an eye on your usage so you don’t hit the message limit too quickly!

If you’ve ever opened TradingView’s Pine Editor, stared at it for five minutes, and closed it, this one’s for you.

Pine Script is TradingView’s built-in scripting language. It’s how you build custom indicators, set alerts, and run backtests. But it’s also notoriously finicky. The rules change between versions, functions have strict naming rules, and one wrong spacing will break your entire script with an error that tells you almost nothing useful.

But this changes today.

This post closes that gap. We’ll walk through how to use Claude as a Pine Script co-pilot, from describing your idea in plain English to having a backtested strategy running in Strategy Tester. Two steps, one workflow.

First: The Skill File

Claude on its own will write Pine Script. Claude with the right context will write Pine Script that just works.

What is a skill file?
“A skill file is a document that instructs AI agents on how to perform specific tasks, workflows, or follow company standards”

The skill file we’ve built for pinescript is a reference document that tells Claude exactly how Pine Script v6 works and common mistakes it should avoid when writing it. The skill file also has a 10-pass self-review checklist that Claude runs on its own code before showing it to you (it means minimal errors).

Step 1: Download the Skill File

We’ve dropped the skill file in our Google Drive. This is the file that turns Claude into your Pine Script co-pilot download it before anything else.

Uploading the whole .zip folder is important since the skill is not a single file but multiple files, which give the right context to claude

Download the skill File

Step 2: Load the Skill File in Claude

Once downloaded, follow the video below to get it set up in Claude. It takes under two minutes. Do it once and it’s ready every time you open Claude.

Step 3: Call the Skill

Once the skill is loaded, activating it is just one word. Type /pinescript before your message and Claude instantly knows it’s in Pine Script mode with all the rules, checks, and context already loaded.

Step 4: Describe What You Want

This is where most people expect it to get complicated. It doesn’t.

Just describe your indicator in plain English, the same way you’d explain it to a friend. You don’t need to know any code, any syntax, or anything about how TradingView works under the hood.

For example:

“hey I want to create a indicator which shows the historical probability that the nifty on 1d timeframe has bounced from their EMAs. 10,21,50,100,200
---
use /pinescript to help me plan it”

That’s enough to get started.

Some other examples you can use:

• “I want to plan an indicator that highlights when volume is unusually high compared to the past 20 candles using /pinescript.”

• “I want to see when the 9 EMA crosses above or below the 21 EMA, with arrows on the chart using /pinescript. Help me plan it”

• “I trade breakouts. I want an indicator that draws a box around the high and low of the last 20 candles and alerts me when price breaks out using /pinescript. Help me plan it”

Also, the trick to make claude write efficient code is be as detailed as possible! One more thing to be cautious of is to use the word “PLAN” while writing your prompts!

Step 5: Answer Claude’s Questions

Here’s what makes this workflow different from just asking claude to write you a script… Claude won’t write a single line of code before it asks you clarifying questions.

It asks you a focused set of questions first:

• What should it plot: a line, histogram, background colour, shapes on the candles?

• Should it mark buy/sell signals? At what condition exactly?

• What RSI length and thresholds? Should those be user-configurable?

• Any overbought/oversold levels to draw?

Answer them in one reply. The more specific you are, the better the output.

If Claude doesn’t ask you these questions and directly gives you the script, tell it to do the following:

“Please read the /pinescript file again and ask me the questions”

Step 6: Review the Script

Before writing any code, Claude sends back a Script Spec — a plain English plan of exactly what it’s about to build. Think of it as a blueprint you approve before construction starts.

Script Spec — EMA Bounce Probability Table

Type: Indicator (overlay=true)
Logic: For each of the 5 EMAs (10, 21, 50, 100, 200), scan the last 252 bars on the 1D timeframe. A touch is a candle whose low dips below the EMA but closes above it (wick crossunder + body recovery). A bounce is confirmed when the next candle's close is also above that EMA. Probability = confirmed bounces ÷ total touches.
Visuals: A single stats table (top-right corner) with rows for each EMA showing: EMA period | # Touches | # Bounces | Bounce Probability %
Inputs: Lookback period (default 252 bars), toggle to show/hide each EMA line on chart, table position
Filters: Minimum 1 touch required to show a % (otherwise shows —)
Notes: No short-side / resistance logic. Works on any symbol on 1D timeframe; EMAs computed natively so no request.security() needed.

What you should check

Read through the spec and ask yourself:

• Does the logic match what I described?

• Are the signals correct?

• Are the visuals what I pictured?

• Are the default settings what I want?

How to respond

If everything looks right, simply reply:

Looks good, go ahead.

If something is wrong or missing, tell Claude what to change:

Change the buy signal condition to when RSI crosses above 40 instead of 30.

⚠️ Important: Do not skip this step. This is your only chance to check the plan before the code is written. It is much easier to change the spec now than to fix code later.

One thing that makes this skill unique is that it forces claude to run a 10 step internal checklist before it gives you the code!

Step 7: Get Your Code

Claude writes the complete Pine Script and before showing it to you, silently runs a 10-pass review on its own output, checking for errors, outdated syntax, and common Pine Script traps. If it catches something, it fixes it on the spot.

What you get is clean, ready-to-use code.

Step 8: Paste Into TradingView

• Open TradingView

• Click Pine Editor at the bottom of the screen

  

  

• Paste the code

  

• Click Add to Chart

  

Your indicator is now live!

Troubleshooting if something goes wrong

In most cases the code will work perfectly the first time. But if you do see an error, here is what to do.

If TradingView shows a red error message

Do not panic. Simply copy the exact error message text, go back to Claude, and paste it in the chat:

“TradingView is giving me this error. Please fix it: [paste the error here]”

Claude will read the error, identify the problem, fix the code, and give you a corrected version to paste back in.

If the indicator doesn’t look right

Go back to Claude and describe what is wrong. Be specific:

“The indicator is working but the buy signal is appearing too frequently. I only want it to fire when RSI has been below 30 for at least 2 candles and then crosses back above 30.”

Claude will update the spec, confirm with you, and give you revised code.

If Claude seems to forget the skill File

This can happen in very long conversations. If Claude starts making Pine Script errors again, simply say:

“Please re-read the skill File and apply all Pine Script v6 rules before continuing.”

Stuck somewhere?

Refer to the below chart!

Now Try It With Your Own Idea

You don’t need to start from scratch. Already trade a setup manually? Describe it. Seen an indicator you liked but couldn’t replicate? Describe that. The skill handles everything in between.

Next, we’ll take this exact indicator and turn it into a backtest so you can see whether your setup actually holds up historically.

(Disclaimer: Any video/image/text content is for educational and informational purposes only and does not constitute financial advice. Refer - https://stocko.in/stocko-disclaimer/. Please do your own research or consult a qualified financial advisor before making any investment or trading decisions. Trading in stock markets involves the risk of loss.)