Principal Variation Search is an optimization of alpha-beta pruning where the engine searches the most promising move (the “principal variation”) fully, but then searches all other moves with a reduced “null window” to quickly detect if they are worse. If a move unexpectedly beats the principal variation, the engine re-searches it in full. This hybrid method cuts down the number of nodes explored while still guaranteeing exact results, because most moves are inferior and can be disproven with minimal effort. PVS is one of the core techniques that allow engines to reach high depths without exponential slowdown.