Game Theory and Strategy

Strategy is about competition, and, ideally, cannot and should not, be cast in stone. Static strategy is suboptimal. While it should be concrete enough and clear enough for all to follow, it should, however, be dynamic and capable of evolution and change over a given time horizon. Strategy set ex-ante mostly becomes more or less obsolete once the rubber hits the road, once additional market information becomes available, once other players, be they competitors or other ‘neutral’ players like customers, regulators, take their respective actions in response. In such circumstances, it would be foolhardy, even potentially suicidal, to hang on to what was decided at the strategy session or board, without making appropriate changes to respond timely and appropriately.

 

While it is true that in developing strategy, a market and competitor analysis is undertaken, this tends to be static or backward looking. And yet, strategy is a game-plan for the future.

 

Game theory modelling, while mathematically perhaps too robust for everyday application, is however very pertinent for strategy modelling. Clearly, it is important to understand the game you’re playing so that you choose the appropriate form of the game. Once chosen, the model can guide you throughout the interactions as your strategy unfolds. 

 

Take, for instance, a duopoly. You can think here of Coke versus Pepsi. Or iOS versus Android. In geopolitics, during the cold-war, it was the US versus the USSR, but now it has morphed into the US versus China, and so on.

 

In practice, neither of these can have, or should have, a strategy that is static. Yes, there is an overall goal: dominance. But each can only use past knowledge to model or anticipate or create scenarios based on the other’s reaction to their own moves. Player-1 wouldn’t know beforehand what Player-2’s reaction would be to Player-1’s move, neither would Player-2 know Player-1’s move beforehand, and how Player-1 would react to Player-2’s move in reaction to Player-1’s initial move, and so on. Neither side would also know the severity or extent or impact of the move each makes (No move is also a move). Moreover, they would not know the reactions of ‘third parties’ like regulators or consumers, which would impact or have to be factored into subsequent moves. 

 

What they have to work with are scenarios or sequences, and their likely payoffs. Some situations may have finite scenarios/sequences, while others may have an infinite number of scenarios/sequences, and in order to simplify and be able to make a move, they may have to narrow down to a few scenarios/sequences. All this sounds philosophical and convoluted, but it can be modelled. But the modelling presupposes flexibility. A calculation of the payoffs of each action and reaction, without knowing what the competition will do, and a preparedness to adapt and change to circumstances as they unfold. 

 

So, a once-off annual strategy session, largely based on your own decisions, without factoring in what the competition may do, and then putting it on the shelf won’t cut it. Besides the all-too-well-known failure to execute strategies, a failure to model appropriately may also be a factor in not implementing strategies: they are generally not implementable.