so i built a learning indicators a while back, watches the tape and builds up probabilites for what price does next given the current context. the math is solid, beta posteriors kernel smoothing parental priors (I learned this from AI) for cold start, im actually proud of the structure. but using it live taught me something i didnt expect. markets dont sit still long enough for it to learn anything stable, but i’m only using 1458 context keys.
I haven’t put additional time into it, because it started to seem like I’d probably need to expand this 1 billion context keys and do a whole bunch of other things, that won’t guarantee much. But with 1458 context keys, it takes over 3 weeks to stabilize, saves to file, but its experimental.
Its an idea anyone can run with if interested.
I don’t share code, I share ideas.
The idea is more important than the code in the age of artificial intelligence.
For me its the idea, once I see or hear an idea, we (AI and I) can infer everything else on first principles. (And there’s enough in my description to interrogate the concept)
When you read books, literature, or whatever, it imbues one with ideas, and ideas lead to action. Relatively intelligent action, for instance along the lines of first priciplies, can get one closer to the thing.
The key here is, given a premise, can we imagine contingent possibilities. Imagination is an asset with these AI tools. It one of the important ingredients it doesent have.
But, I hear your point, I’ll start releasing some tools on the Ninjatrader Ecosystem in the future.
Have you tried running it and then labeling the frames, I know this will be more work, but it will be more accurate, thank you for the idea,
1 Like
It already does that inside the 1458 context keys. Each key captures a context and it is then reinforced with the repetition of that context and price move continually, until it stabilizes.
After stabilization, it’ll continue adjusting but not at the same rate as at the beginning.
It saves to an external csv which I load up every session to it’ll continue building on its context matrix. (Learning)
yes this is perfect, some might say it’s a lot of work but it’s a great method, and I’m mind blown that u only have 1458 context keys, great work ,
I share ideas, because it could help some minds break the pattern of simply recreating a different form of whatever they’ve seen, instead seeing something like this, it could light a fire in their psyche and blow open their sense for what is possible.
This idea I’ve shared is enough to run with, share it with a high level AI, mix in some context building and let your mind play at creative creation.
This isn’t an indicator, it is a research module. If one wants to harness it, you’d have to put it together yourself because that’s the only way it’d be understood and valued.
Case in point: once you take this idea, play with it… imagine taking the learning component and attaching to any other ninjatrader tool. You’d a new class of learning ‘indicators’. Can you see the possibilities?
