What it does: System C is the most advanced ML system. It uses a GRU neural network with attention — a type of artificial brain that reads price history like a sentence, learning which past moments matter most for predicting the future. Unlike XGBoost (which looks at a snapshot), the neural net understands sequences and patterns over time.

What is GRU? GRU (Gated Recurrent Unit) is like a short-term memory system. Imagine reading a book one word at a time — a GRU remembers the important plot points and forgets irrelevant details. For trading, it reads 100+ candles in sequence and learns temporal patterns like "after 3 red candles with increasing volume, price usually bounces."

What is Attention? Attention lets the model "highlight" the most relevant candles. If a whale moved 5 hours ago and that's more important than what happened 1 hour ago, the attention layer will focus on that. It's the same technology that powers ChatGPT and modern AI.

What does the heatmap show? The confidence heatmap shows how confident the neural net is about each symbol at each timeframe. Brighter green = higher confidence in the prediction (BUY or SELL). Red = model is uncertain or predicts a move against the signal.

25 input features: The model ingests: normalized OHLCV, RSI, MACD, Bollinger %B, volume z-score, hour of day (sine/cosine), BTC return, Fear & Greed index, funding rate, Fibonacci position, and lagged returns. It predicts 3 things simultaneously: entry signal, take profit level, and stop loss level.

Current status: TRAINED — Model file (gru_attention.pt) exists. System C has a trained neural net and may be generating predictions, but Discord notifications are disabled pending performance review.