Chess vs AI

Chess vs AI controls

The story behind Chess vs AI

Chess in its modern form emerged in Europe in the late 15th century, evolving from the older Indian game chaturanga through Persian shatranj and Arab influence. The rules for the queen and bishop's modern long-range moves stabilised in Spain and Italy around 1475. The first world chess championship was played in 1886 between Wilhelm Steinitz and Johannes Zukertort. Computer chess as a serious field began with Claude Shannon's 1949 paper "Programming a Computer for Playing Chess", which established the minimax search framework that all classical chess engines still use. IBM's Deep Blue defeated then-world-champion Garry Kasparov in a 1997 match, a landmark moment in artificial intelligence. Modern engines like Stockfish and Leela Chess Zero now play at superhuman strength — Stockfish has been rated over 3500 Elo, roughly 700 points above the highest-rated human player in history. The AI in this browser game is deliberately much weaker than that — it is designed to give beginners and intermediate players a friendly, solvable opponent rather than a perfect one.

Master-level Chess vs AI

The single most important strategic idea in any chess game against a computer opponent is piece activity. Beginners think in terms of material — how many pieces they have on the board — and try to protect every piece from capture. Improving players think in terms of activity — how many squares each of their pieces influences and how many threats they collectively create. A well-placed knight in the centre of the board attacks eight squares; the same knight in the corner attacks two. Against the Medium and Hard AI on this site, the winning approach is to actively develop your pieces into the centre in the first ten moves, castle to safety, and then look for tactical shots that gain material through fork, pin and skewer patterns. Passive play — moving the same piece back and forth, keeping everything safe — loses to any competent AI because it lets the AI build a slow positional advantage that eventually cracks your position open.

The second key concept is trade evaluation. When you have a choice to trade pieces, ask three questions: does the trade favour me materially, does it improve my piece activity, and does it improve or worsen my king safety? Beginners often assume that trading queens is always safe (it usually is when material is roughly equal), that trading rooks is neutral (it depends on open files), and that trading bishops for knights is fine (it depends on the position — bishops are better in open positions, knights are better in closed ones). The Hard AI on this site plays trades correctly, so if it initiates a trade, it usually thinks the trade helps it. When you have a choice about whether to accept a trade, defaulting to "keep pieces on when I am winning, trade when I am losing" is a decent heuristic — the losing side wants fewer pieces on the board to reduce the winner's options.