1: \begin{abstract}

2: In this paper we consider the problem of distributed channel allocation

3: in large networks under the frequency-selective interference channel.

4: Performance is measured by the weighted sum of achievable rates. First

5: we present a natural non-cooperative game theoretic formulation for

6: this problem. It is shown that, when interference is sufficiently

7: strong, this game has a pure price of anarchy approaching infinity

8: with high probability, and there is an asymptotically increasing number

9: of equilibria with the worst performance. Then we propose a novel

10: non-cooperative M Frequency-Selective Interference Game (M-FSIG),

11: where users limit their utility such that it is greater than zero

12: only for their M best channels, and equal for them. We show that the

13: M-FSIG exhibits, with high probability, an increasing number of optimal

14: pure Nash equilibria and no bad equilibria. Consequently, the pure

15: price of anarchy converges to one in probability in any interference

16: regime. In order to exploit these results algorithmically we propose

17: a modified Fictitious Play algorithm that can be implemented distributedly.

18: We carry out simulations that show its fast convergence to the proven

19: pure Nash equilibria. 

20: \end{abstract}