Why the financial markets cannot be universally understood

Grabbing OI/CVD, I suppose many bloggers have written about it as well, I myself once implemented a series of OI/CVD monitoring strategies, which indeed proved effective (early last year I had a signal notification group because the strategy could detect faster than Vida's OI monitoring by 3-5 minutes)
When do I think this kind of strategy is no longer suitable?
When all exchanges start to equalize skills and move towards agent-based systems, it inevitably leads to MM strategy adjustments
A short-term surge in OI volume — quick liquidation, all short-term automated bots following in are "trapped"
This is the new era of perp sandwiching
I think the chairman's open-source skill is very well done @connectfarm1, including @xincctnnq, who is also working on it, and the strategy is profitable
But the market's rhythm keeps changing, MM strategies keep evolving, and the OI of "monster coins" can jump from 3M to 10M within hours — still impressive in the morning, but gone by the afternoon (such cases are increasing recently)
The problem with small coins is liquidity + information/chip asymmetry (info edge). Analyzing dynamic indicators is still a "technique" study, which is scripting thinking. This is also what I discussed with @thecryptoskanda some time ago — LLM should be a systematic learning to analyze the real on-chain chip structure, understand MM's movements, and start positioning before OI stacks up
Don’t think script-wise, don’t bring script thinking into it, or else AI will always be AI, and you’ll never be able to fully utilize it