Thats refinement versus improvement. Practice requires understanding, blindly copying and refining is always limited. "Hitting a wall" is like brain neural netw

Thats refinement versus improvement. Practice requires understanding, blindly copying and refining is always limited. "Hitting a wall" is like brain neural network stuck at local optimum, which requires forgetting and relearning. Lets take example with Sudoku: if you know the basics(marking candidates and eliminating rows/columns), you can solve easy puzzles with practice only, but you'll be stuck solving mid-level puzzles. Mid-level puzzles require intuitive understanding of column-row exclusion(where X number candidates in a block are exclusively on one row or column they exclude X from entire row or column), but aren't sufficient for solving high-level puzzles. High-level play requires knowledge of subsets, X-wings,XY chains,coloring, etc to solve the hardest puzzles(and practicing doesn't help, it requires intuitive understanding to find such patterns).