0
$\begingroup$

How one can find each player’s optimal strategy and the value of the two-person zero-sum game in the following Table:

\begin{array}{c|c|c|c|c|} & L & CL & CR & R \\ \hline T& 4& 5 & 1 & 4 \\ \hline M & 2& 1&6 &3\\ \hline B & 1& 0 &0 &2\\ \hline \end{array}

I know this game has no saddle point. I want to use mixed strategy. But, how can I reduce this matrix using iterated elimination of dominated strategy first? Moreover, what would be the Nash-equilibrium of this game?

Anyone help?

$\endgroup$
  • $\begingroup$ Delete the third row because it is strictly dominated for Player 1 by either of the two top row. In the next round, delete the right-most column because it is strictly dominated for Player 2 by a convex combination of the two central columns (can you see why?) $\endgroup$ – mlc Mar 18 '17 at 23:08
  • $\begingroup$ Thanks. I understand for the last row why it is deleted. I haven't realized for the right most column. I thought the first column should have been deleted because the right most column dominates the first column. $\endgroup$ – UserAb Mar 18 '17 at 23:47
  • $\begingroup$ The Column player wants to minimize so the rightmost column is weakly dominated by the leftmost column. (Rightmost should be deleted, not leftmost.) $\endgroup$ – mlc Mar 19 '17 at 8:59
0
$\begingroup$

As discussed in the comments, delete B because it is strictly dominated. Then delete R because it is (iteratively) strictly dominated. You are left with the matrix \begin{array}{c|c|c|c|} & L & CL & CR \\ \hline T & 4 & 5 & 1 \\ \hline M & 2 & 1 & 6\\ \hline \end{array}

For $0 \le p \le 1$, suppose that Player 1 adopts the mixed strategy that plays $T$ with probability $p$ and $B$ with probability $1-p$. Then, if 2 plays $L$, the expected utility of Player 1 is $u_1(p,L) = 4p+2(1-p)=2+ 2p$; if 2 plays $CL$, it is $u_1 (p, CL) = 1 + 4p$; and, if 2 plays $CR$, it is $u_1 (p,CR) = 6-5p$. Compare these three functions on a graph and check that $$\min \{ 1 + 4p, 2+ 2p, 6-5p\} = \left\{ \begin{array}{cc} 1 + 4p & \mbox{if } 0 \le p \le 1/2 \\ 2 + 2p & \mbox{if } 1/2 < p \le 4/7 \\ 6 - 5p & \mbox{if } 4/7 < p \le 1 \\ \end{array} \right.$$ (The graph of this function is the lower envelope of the three component functions.) Recall that Player~1 wants to play maximin, so he chooses $p$ in order to $$\max_p \min \{ 1 + 4p, 2+ 2p, 6-5p\} = \max$$ which is attained at $p^*=4/7$ for a value of $22/7$. Given $p^*$, 2 is indifferent between L and CR and thus she randomises between these two strategies with probability $q^*=5/7$ and $1-q^*$, respectively, to make 1 indifferent. The value of the game is $22/7$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.