The Strategy of Conflict, by Thomas Schelling. ISBN 0674840313.
I expect everything in here is deeply familiar to today's game theorists, but it wasn't back in 1960 when this book was first pubished. Schelling rings the changes on a few simple but deep ideas in game theory:
- Restricting or worsening your options (in particular, by making commitments and arranging to be bound by them) can be to your advantage (because it stops your opponent reasoning "if I do X, he'll do Y, which would be awful" and allows you to get to (X,Y') which is better for you than you could have done if your opponent hadn't done X -- despite being worse than (X,Y), which you'd never have been able to get anyway). It can similarly be useful to have less than perfect information, or less than perfect rationality.
- In games with any cooperative element, coordination becomes important, and is to a considerable extent a psychological and social, not a mathematical, phenomenon.
- In general, the "human" aspects of game theory need at least as much attention as the mathematical aspects.
Schelling applies these ideas primarily to war (limited, nuclear, cold, ...), deterrence, and negotiation.
Many of the chapters of "The Strategy of Conflict" contain phrases like "this paper"; I assume they are derived from earlier publications. But the book holds together well; it doesn't generally feel like a set of loosely connected articles. It helps that Schelling writes well.
(Schelling was one of the two co-winners of the 2005 Nobel Prize in economics.)