Scale into a position only as your thesis is proven right (add when the price moves toward your model's fair value) and cut size when it moves against you, the opposite of averaging down a loser. The edge is convex exposure: you get big on the trades that are working and stay small on the ones that aren't, which suits prediction markets since every market eventually settles to a clear yes-or-no truth.
Markets on Polymarket where anti-martingale conviction scaling is the natural play:
