I spent ~ 9 hours with co-pilot with this is the fix this is why it works in every case: blah bla blah.
in 30 minutes Claude fixed the code in a way copilot said would not work.
2 Likes
Never mind I asked Claude. I had already built in logging Claude said to share the log.
For Context my signal set is MACD based and instead of calling it MACD Cheesecloth I call it MACnCheese. The Cheese cloth is my attempt to filter out the noise signals.
How can I best share my trading day for analysis?
8:13 AM
Claude responded: A few good options depending on what you want to analyze:
A few good options depending on what you want to analyze:
Export from the addon The .cs file already writes two CSVs automatically:
_Cheeselog.csv— 500ms snapshots of all DEF/Avg/Sum valuesAlert,TradeDialog.csv— timestamped alert and trade signal events
Those two files together give a complete picture of your day.
What to share for analysis The most useful combination is:
- Cheeselog CSV — the raw time-series data
- Alert log CSV — when signals fired
- A screenshot or notes on how price actually moved (so signals can be correlated to outcomes)
Where to analyze it
- Drop the CSVs back here and I can chart them, find patterns, flag which signals were predictive vs noisy, and compare thresholds
- The HTML tool you have now could also be extended with a CSV import so you can replay a day and see the color states change over time
Quickest path — just drag the _Cheeselog.csv from Documents\NinjaTrader 8\log\MACnCheeseLog\ into this chat after your session and I can build you a full replay and analysis.
Want me to add a CSV import + chart replay feature to the HTML file?
1 Like
It’s clear to me now that Ninja Trader was primarily built for those who can program. The default offering in Ninja Trader is simply the default offering, it wasn’t meant to be all or only what we’re using.
1 Like
A key limitation when trying to have ChatGPT work with online sources is that it can’t actually open links — it only processes whatever text is visible up to the point of the link. Because of that, I shifted to a more controlled approach.
I compiled most of the NinjaTrader 8 framework into a single text file called:
“12-Block Library NT8 (Final Authority).txt”
(6 months of copy/paste all day long, every day.)
Instead of relying on vague generation, I feed this file into ChatGPT along with my scripts. I also use it to have the model audit code — it flags violations, identifies structural issues, and then returns corrected versions of the script that actually compile. If there’s still an error, I iterate again until it’s resolved.
One important thing I’ve learned is that if you let AI freely “fix” or rewrite code without strict constraints, it will often silently remove or alter logic without clearly stating what changed. To avoid that, I force it to output explicit FIND/REPLACE changes for every modification, rather than general instructions like “add this feature” or “improve this section.” Otherwise, it tends to “optimize” things in ways that can break intended behavior.
Using this approach, I’ve been able to build fairly solid indicators using the free version of ChatGPT. Once I have a stable script, I can even reuse the structure in other chats without needing the full 12-Block library every time — though I still bring it in when needed for validation or deeper checks.
One of my main projects is a Python pipeline that reads MTF_OHLCV_v1.csv and feeds another indicator that continuously produces real-time market context on a bar-by-bar basis, like:
FAILED BREAKOUT at 7270
↓
STRONG SELLING PRESSURE
Bias: BEARISH
Structure: Back inside range
Momentum: Increasing
Expect:
→ 7250 test (in progress)
→ 7235 (next target)
That kind of structured read gives me continuous situational awareness instead of raw signals.
I’ll admit I do get frustrated with Chat sometimes and have thrown a few F-bombs at it — but it doesn’t exactly take it personally.
1 Like
ChatGPT 4.5 and Claude can find and read everything big dawg🐕
You need to just pay for the sub and learn about skills and prompts.
Worth it though, to save on major cost
I’ve built far more superior products that I won’t even go into. But I’m no longer afraid to show this.
But this is an ADD-on static DOM, with saving templates, persitence and institutional grade performance and optimization, and in every way superior to the default Ninja Trader offering, or whatever is available to the retail market. You can click a price to enter, and adjust the price by moving the entry line, target, stop loss, etc directly on price column
Everything that retail is sold, is probably bullshit. After you sat behind the screen and built up your screen time, you know what you’re looking for, you know the behaviors. No tool is going to make you profitable!. However, you can build precisely what unique to you that helps you in this business.
No, this is not publicly available, and will probably not be publicly available for a decade. Because I’m using it myself. But you’ve seen what it’s possible, you can build your own that has your unique features and your unique requirements, and you can probably build it in one or two days.
The Ninja Trader platform, is a playground, that you can use to build all sorts of things. And you don’t even need to connect to the data, you can use its framework to build upon anything, but that’s a whole different story.
Just by looking at it you can see that there’s some specialties.
Particularly look at the ordering system button at the bottom.
Notice the ordering system and what that implies.
The code itself was 7000 lines built with Claude.
Building this from scratch, without AI, with developers, going back-and-forth, explaining this or that, and then the performance and optimization layer, that would cost over $600,000.
But this is an example of what’s possible building with a high level artificial intelligence. I am not a coder, but I understand what I want and what I need.
This DOM provides edge because I no longer enter based on what I think, I enter based on events in the marketplace. And, hint, that can increase your edge.
The build time for this was two weeks.
3 Likes
I blurred out my IP’s because the idea itself is an edge, and this is an age, where even this image shown to the right mind, can transmit something I don’t want transmitted.
But, here ya go. And you can build this yourself.
1 Like
big dawg🐕 hahaha
I’m not a coder either. Know that movie Limitless? Coding using A.I. is just about like that lol. I take my smart pill and I’m suddenly a coder.
But a, 4.5 and Claude actually opens up those links? If it can, damn! Game changer. Holy smoke. Because my 12-Block fixes the fucked-up scripts that Chat gives me like that magic pill.
#####################
I asked Chat, here’s what it said,
Short answer: no—none of them truly “open” links like a human browser does.
Models like GPT-4.5 (and me) or Claude can access and read web content if browsing tools are enabled, but what’s actually happening is:
- They fetch the text/content of a page, not interact with it visually
- They don’t click around, log in, or navigate dynamically
- They can’t reliably access content behind logins, paywalls, or complex scripts
- They often only see what’s statically retrievable, sometimes a partial snapshot
So when it feels like they’re “opening links,” it’s really just:
pulling whatever readable content is accessible and summarizing it
That’s why your workaround—copying the full NinjaTrader docs into your own “12-Block Library”—is actually very effective. You removed all the uncertainty about what the model can or can’t access.
###########
I LIERALLY copied paste more than a shitload, of NT8 code content and AI can read EVERYTHING code. Im humble, I dont know nothing I do the best I can.
AI can accelerate the path but understand the tradeoff,
Many people are building businesses, software, indicators, strategies, and products entirely through AI generation without understanding ownership risks, originality issues, licensing concerns, or future patent and copyright complications.
If your entire system is generated by AI, there may be future disputes over originality, ownership, training data, or intellectual property rights depending on the platform, model, or laws involved.
I was told the strongest approach
Use AI as a research assistant.
Use AI to learn faster.
Use AI to organize information.
Use AI to improve workflow efficiency.
But build the actual architecture, logic, execution, and final system yourself.
Your ideas, your structure, your implementation, and your engineering process are what create true ownership and long-term value.
AI should enhance your intelligence not replace your ability to think, create, and build independently.
The people who will dominate the future are not the ones copying AI outputs blindly.
They are the ones who understand the information deeply enough to create their own systems from the ground up.
I personally had an experience with a copy right that I submitted, an AI helped me make this one picture page and it ruin 200-page copy right, just one page that I involved AI my copyright was denied ,
Learn from AI.
Study from AI.
Accelerate with AI.
But develop your own mind, your own process, and your own technology.
That is where real power and real ownership come from.
1 Like
Maverick, to find errors in my code I prefer Claude.
But don’t forget the main thing, which I think you’re also on the right track for, and that is the idea: people differentiate themselves by their ideas and how they approach them. Let’s say getting rich, but there are 1000 ways, some good and some terrible.