Ah, of course, I forgot a prepositional phrase: he actually was less likely to score on that side because it was easier for the goalie to predict which side he would favor.
(Incidentally, this proposition has been empirically tested in G.C. Moschini, Economics Letters 85 (2004) 365–371)
However, we do have to be careful in games of strategy in selecting what we call capability improvements. Increasing my payoff in a single cell can change the relationship between cells, preventing me from credibly committing to a particular strategy and thereby diminishing the outcome of the game.
As an example, imagine we have a game defined as follows:
(U,L) ⇒ (1,1)
(U,R) ⇒ (0,0)
(D,L) ⇒ (0,0.9)
(D,R) ⇒ (1,0)
where the pairs are (x strategy, y strategy) mapped to (payoff to x, payoff to y).
The unique Nash equilibrium is (U,L), so each player receives a payoff of 1.
Now “improve” player y’s capabilities by making (U,R) ⇒ (0,1.1). Now there is no equilibrium in pure strategies, and the unique mixed strategy equilibrium is: Pr(U) = 0.9, Pr(L) = 0.5. Expected payoff to the “improved” player is 0.99, and to the other player 0.5, both down from their previous equilibrium values of 1 each, and the magnitude of the effect damaging player y increases as her payoff to (D,L) decreases (derivations available on request).
Of the top of my head, I suspect your heuristic applies in zero-sum games, but not necessarily elsewhere. Unless the players could read each other’s source code...
Ah, of course, I forgot a prepositional phrase: he actually was less likely to score on that side because it was easier for the goalie to predict which side he would favor.
(Incidentally, this proposition has been empirically tested in G.C. Moschini, Economics Letters 85 (2004) 365–371)
However, we do have to be careful in games of strategy in selecting what we call capability improvements. Increasing my payoff in a single cell can change the relationship between cells, preventing me from credibly committing to a particular strategy and thereby diminishing the outcome of the game.
As an example, imagine we have a game defined as follows:
(U,L) ⇒ (1,1)
(U,R) ⇒ (0,0)
(D,L) ⇒ (0,0.9)
(D,R) ⇒ (1,0)
where the pairs are (x strategy, y strategy) mapped to (payoff to x, payoff to y).
The unique Nash equilibrium is (U,L), so each player receives a payoff of 1.
Now “improve” player y’s capabilities by making (U,R) ⇒ (0,1.1). Now there is no equilibrium in pure strategies, and the unique mixed strategy equilibrium is: Pr(U) = 0.9, Pr(L) = 0.5. Expected payoff to the “improved” player is 0.99, and to the other player 0.5, both down from their previous equilibrium values of 1 each, and the magnitude of the effect damaging player y increases as her payoff to (D,L) decreases (derivations available on request).
Of the top of my head, I suspect your heuristic applies in zero-sum games, but not necessarily elsewhere. Unless the players could read each other’s source code...