An open-source, candle-only crypto pump detector–looking for ML contributors

There's something profoundly strange about the idea that a pump-and-dump scheme—a coordinated act of financial deception—leaves a detectable fingerprint in nothing more than candlestick patterns. No insider gossip, no on-chain whale tracking, just the geometry of price movement over time. This tool operates on a deeply weird assumption: that human manipulation, when scaled across hundreds of coordinated actors, becomes statistically legible, almost like a phase transition in physics. The market, it turns out, is a thermodynamic system where information flows downhill and bubbles form like supercooled liquids, waiting for a single tap to crystallize into collapse. What fascinates me here is the inversion of the usual narrative. We usually think of algorithms as the ones doing the manipulating—the bots front-running trades, the arbitrageurs exploiting latency. But this screener arms the retail observer with a kind of X-ray vision, turning raw price data into a confession. It's a small act of democratization in a domain that prides itself on being ungovernable, yet it also raises an unsettling question: if we can detect these patterns, how many more are we missing? The tool is "candle-only," which means it's deliberately blind to everything else—order flow, social sentiment, whale wallets. It's like trying to diagnose a disease by only listening to the heartbeat, ignoring all other vital signs. And yet, the fact that it works at all is a testament to the idea that markets are not random walks but complex, self-organizing systems with hidden structure. The call for ML contributors is where it gets truly speculative. Machine learning on financial time series is a notoriously treacherous endeavor—the data is noisy, non-stationary, and riddled with survivorship bias. But if the ML can learn to distinguish organic volatility from coordinated manipulation, we're essentially teaching a neural network to recognize the psychological signature of a mob. That's a weird and wild frontier: using artificial intelligence to detect the collective unconscious of traders, encoded in the fractal geometry of price charts. It's not unlike trying to hear the whispers of a crowd through the roar of a stadium. Source: https://github.com/aleks-ent/pump-dump-crypto-screener