# How Machines Learned to Play Chess Better Than Anyone
Deep Blue beat Garry Kasparov in 1997. That moment changed everything.
Before that match, grandmasters thought computers would always need humans. Kasparov himself believed no machine could match his intuition, his feel for the position, his centuries of human chess knowledge compressed into his mind. Deep Blue proved him wrong. It won 3.5-2.5. The 1997 rematch ended any serious debate: machines had arrived.
Deep Blue used brute force. It evaluated millions of positions per second, checking move after move after move. IBM engineers hand-coded evaluations for endgames and opening positions. The machine was powerful but narrow. It couldn't learn. It couldn't adapt. It simply calculated faster than any human ever could.
For the next two decades, that remained the chess engine playbook. Engines got stronger by getting faster. Stockfish emerged from the open-source community in 2008 and became the gold standard. It still works this way today. Traditional engines like Stockfish evaluate 50 million positions per second on a modern computer. They use alpha-beta pruning and other search optimizations to avoid wasting time on obviously bad moves. They're brutally effective.
Then AlphaZero arrived in 2017 from DeepMind, Google's AI subsidiary. It worked completely differently. AlphaZero didn't memorize opening theory or endgame tables. It didn't calculate in the human sense. Instead, it learned chess from scratch through self-play. It played millions of games against itself, gradually discovering which positions were winning and which were losing. No human knowledge. No hand-coded rules. Just neural networks learning patterns.
The results shocked everyone. AlphaZero played chess with style. It sacrificed material in ways no traditional engine would dare, finding compensation in initiative and pawn structure that took seconds to understand. It beat Stockfish convincingly in a 100-game match. For the first time, chess players watched an engine think like an artist rather than a calculator.
But here's what matters for competitive chess: AlphaZero never became the standard tool. It required enormous computational resources. DeepMind kept it mostly proprietary and academic. Stockfish remains freely available and runs on any computer. Most serious players still use Stockfish for preparation.
The real impact on the game has been subtler. Engines have become so dominant that human intuition barely registers anymore. Players memorize engine lines 20 moves deep. The romantic sacrifices and mysterious moves that defined Kasparov's era get analyzed away within seconds. Computers haven't made chess worse, exactly. They've made it different. More precise. Less mysterious. More about who prepared better the night before.
That's the trade-off we've accepted since Deep Blue won in 1997. We got stronger play and deeper understanding. We lost some magic.