The history of Stratego

L'Attaque: a French game, patented by a woman in 1908

Stratego's ancestor is L'Attaque, and its inventor is one of the few named women in the history of board game design. Hermance Edan, a Frenchwoman who had developed the game in the 1880s, filed a patent in 1908 for "a battle game with mobile pieces on a game board". The patent issued in 1909 and the game was published in 1910.

L'Attaque was played on a 9×10 board with 36 pieces a side, and it already had the essential idea: hidden ranks, revealed only on contact. That is the mechanic the whole family is built on, and it was a genuinely original invention — nothing else at the time hid the identity of pieces in plain sight.

British editions followed after the First World War, published by H. P. Gibbs and later Waddingtons, and spin-offs appeared for naval and air warfare (Dover Patrol, Aviation).

From L'Attaque to Stratego

The Dutch publisher and designer Jacques Johan Mogendorff adapted the game, renamed it Stratego, and placed it with a Dutch manufacturer. From 1938 it was marketed in the Netherlands, with post-war editions following from 1946.

His changes made the modern game:

• The board grew from 90 squares to 10×10.
• Four pieces were added — two more Bombs, an extra Major and an extra Miner — bringing each army to forty.
• The presentation was standardised, and later editions adopted the Napoleonic imagery that most players recognise.

The game became a Dutch institution: Stratego is to the Netherlands roughly what Monopoly is to the United States, with a national competition scene that has produced a disproportionate number of world champions.

1961: Milton Bradley and the American boom

Milton Bradley licensed the game for American distribution and launched it in 1961 (the trademark dates from 1960). It became a staple of American family game cupboards for the rest of the century, sold alongside Battleship as the other great game of hidden information.

The physical design did as much work as the rules. The upright plastic pieces with the rank printed on one side only are a small piece of engineering genius: they hide information without requiring a screen, a card or a referee, and they let a player see their own army at a glance.

Competitive play followed the game around the world. World championships have been held since the 1990s, and Dutch, Belgian and German players have dominated them — the countries where the game became a childhood standard.

Why computers found Stratego so hard

Stratego resisted artificial intelligence for far longer than chess, Go or poker, and the reason is a rare combination: it has imperfect information like poker, and an enormous game tree like Go, and an initial deployment with an astronomical number of possibilities.

Search-based methods cannot enumerate the possible enemy armies. Poker-style methods, which solve for an equilibrium strategy, hit the problem that Stratego's tree is many orders of magnitude larger than poker's.

In 2022, DeepMind published DeepNash, which learned Stratego by self-play using a model-free reinforcement learning method designed to approach a Nash equilibrium, and reached human expert level in online play. Its behaviour is instructive for players: it bluffs, sacrifices high-ranking pieces for information, and deliberately randomises its deployment so that no pattern can be exploited.

Common questions