Support/Resistance as ML Features

S/R drawn by hand is subjective. Quantify: pivot high/low clusters, volume-weighted price shelves, max-pain as a magnet. Feature sr_distance = |spot - nearest_sr| / atr. The model uses sr_distance to avoid buying into a known shelf (fade) or to size a breakout when volume confirms. This removes the 'I think this is support' bias with a number the chain and price agree on. Subjective lines become testable features -- that is the ML advantage over the cursor.