Would a node-based workflow for automated trading be useful?

General Discussions
1

Hello everyone,
I’m currently exploring an idea for automated / quant trading.
The concept is a node-based workflow where users define:

market data → indicators/features → signal → risk rule → execution → monitoring

The framework would then handle the operational parts around it:

  • data preparation

  • backtesting

  • dry-runs / paper trading

  • execution logs

  • risk checks

  • live monitoring

I’m not asking anyone to share their strategy or alpha.

I’m mainly curious:

  1. Is the bigger pain building the strategy logic, or operating/testing it properly?

  2. Would a node-based workflow be useful, or would you prefer pure Python/code?

  3. Which part would you want automated first: data prep, backtesting, dry-run, monitoring, or execution logs?

Any feedback would be appreciated. Thank you for reading this!

1 Like
2

The strategy logic is easy. Personally, I extract the data from NT into a database and have Python use that data for back-testing. If the strategy seems profitable, then convert that strategy to C# and test it with market replay.

2 Likes
3

That is the most basic approach you can take.

Also You can export the data into a CSV file, then build code around that CSV so the model can read your layout, process the information, and learn from it more deeply and you also can run it with your own terminal so u can get results since its already in a spreadsheet,

Personally, I would rather move toward a more predictive model instead of wasting time building something that everyone already has.

If you are going to call something “quant,” it should actually provide productive value. Nowadays, a lot of people misuse the word quant when they are only doing basic data organization.

A real quant approach should focus on structured data, probabilities, prediction, testing, and improving decision making through measurable results. What you are describing is a workflow orchestration framework, not a quantitative, A true quantitative is what happens inside that workflow, example your quant should kick in from once your probability model to prediction all the way reaching your executions

4

Start by collecting data from NinjaTrader. Make sure you have a CSV exporter and a dedicated folder where every trade is automatically saved and organized.

That folder should capture everything you may need for future analysis, not just the basic trade information. The more complete your dataset is, the more you can learn from it later.

Run your strategy in auto execution and let it collect between 1,000 and 2,000 trades. The larger the sample size, the more reliable your analysis will be.

Once enough trades have been collected, stop the strategy and close NinjaTrader.

Open the folder containing your CSV files. Review the data to make sure everything was recorded correctly. Check for missing values, duplicate trades, incorrect timestamps, or anything else that could affect your results.

You can also create a scanner to validate the CSV files automatically instead of reviewing everything manually.

Once your data is clean, run your analysis through your terminal so you can get your information , At this point, you are not trying to predict the future. You are simply learning what happened and identifying patterns in your trading.

Once you can successfully collect, organize, clean, and analyze your data, you have completed the foundation.

Now build your first analytical model.

Its job is not to predict the market. Its purpose is to discover relationships and identify which market conditions produce your best and worst trades.

Think of this model as your research assistant.
After your first model consistently identifies , use those insights to build a second model whose purpose is prediction.((((Quant))))

once u build your models your models the model should give all the adjustments u need for each one of his ,

Beginners will always choose nodes ,

[quote=“icarus, post:1, topic:7119”]
Which part would you want automated first: data prep, back testing, dry-run, monitoring, or execution logs?
[/quote] Data prep I always first, or else it would be like I want to drive a car without gas

Persist data between indicator resets?