| Article ID: | iaor20003654 |
| Country: | Netherlands |
| Volume: | 14 |
| Issue: | 1 |
| Start Page Number: | 87 |
| End Page Number: | 102 |
| Publication Date: | Jul 1999 |
| Journal: | Computational Optimization and Applications |
| Authors: | Flam Sjur Didrik |
| Keywords: | gradient methods |
We model repeated play of noncooperative stage games in terms of approximate gradient steps. That simple format requires little information and no optimization. Moreover, it allows players to evaluate marginal cost or profit inexactly and to move with different velocities. Uncertainty can also be accommodated. Granted some crucial stability, we show that play converges to Nash equilibrium.