Torres, winner of the 2000 Spiel des Jahres award, is essentially a game of competitive collaboration in tower-building. Played on an eight-by-eight grid, players place plastic "tower blocks", building both outward and upward.
During your turn, you have five "action points"; placing a tower block costs one; moving a "knight" (you start with one) one square -- or through a set of connected tower blocks to the other side -- costs one; and drawing an action card costs one. You can also spend two to place a new knight (you start with one on the board). Playing an action card costs nothing (though you can only play one per turn), and they allow, in effect, one-time special moves.
Tower blocks can be placed orthogonally adjacent to an existing one, or atop an existing one, but never in such a way that two separate castles are united. As well, the highest block in a castle cannot be higher than its surface area (e.g., a castle covering three squares cannot be higher than three levels).
After each player has taken three or four turns, there is a scoring round: basically, if you have a knight on a tower block, it scores as many points as the surface area of the castle times the height of your knight. (But you can score only once per castle, even if you have multiple knights there.) So while you may be collaborating with others to build castles, you compete to occupy the high points. The game ends after three rounds of scoring.
There's a bit more to it than this, but this should give you a sense of the essence of the game: the limited number of action points each turn forces you to make difficult decisions about whether and where to build, and when and where to move. You also have a limited supply of tower blocks to place (replenished after scoring rounds), and have to plan placement carefully to maximize your score at the end of the round. Aside from the action cards, it's a game of symmetrical, non-random strategy -- like most Eurogames, an abstract one, despite the supposed theme (building medieval castles).
Setup also takes mere moments, and a game can be completed in half an hour, unless you have players susceptible to analysis paralysis (the tendency to hold up a game as you obsess over what's the perfect move to make).
It is, to my taste, a little too dry a game to qualify as perfect -- but it is certainly worth playing, and worthy of its award.