Game theory is an amazing way to simulate reality, and I strongly recommend any business leader to educate herself on underlying concepts. However, I have found that the way that it is constructed in economic and political science papers has limited connection to the real world–apart from nuclear weapons strategies, of course.
If you are not a mathematician or economist, you don’t really have time to assign exact payoffs to outcomes or calculate an optimal strategy. Instead, you can either guess, or you can use the framework of game theory–but none of the math–to make rapid decisions that cohere to its principles, and thus avoid being a sucker (at least some of the time).
As Yogi Berra didn’t say, “In theory, there is no difference between practice and theory. In practice, there is.” As a daily practitioner of game theory, here are some of its assumptions that I literally had to throw out to make it actually work:
- Established/certain boundaries on utility: Lots of games bound utility (often from 0 to 1, or -1 to 1, etc. for each individual). Throw away those games, as they preferenced easier math over representation of random, infinite realities, where the outcomes are always more uncertain and tend to be unbounded.
- Equating participants: Similar to the above, most games have the same utility boundaries for all participants, when in reality it literally always varies. I honestly think that game theorists would model out the benefits of technology based on the assumption that a Sumerian peasant in 3000 BC and an American member of the service economy in 2020 can have equivalent utility. That is dumb.
- Unchanging calculations: In part because of the uncertainty and asymmetries mentioned above, no exact representation of a game sticks around–instead, the equation constantly shifts as participants change, and utility boundaries move (up with new tech, down with new regs, etc). That is why the math is subordinate to structure: if you are right about the participants, the pathways, and have an OK gut estimate of the payoff magnitudes, you can decide rapidly and then shift your equation as the world changes.
- Minimal feedback/second order effects: Some games have signal-response, but it is hard to abstract the concept that all decisions enter a complex milieu of interacting causes and effects where the direction arrow is hard to map. Since you can’t model them, just try to guess–what with the response to the game outcome be? Focus on feedback loops–they hold secrets to unbounded long-term utilities.
- The game ends: Obviously, since games are abstractions, it makes sense to tie them up nicely in one set of inputs and then a final set of outputs. In reality, there is really only one game, and each little representation is a snapshot of life. That means that many games forget that the real goal of the game is to stay in it.
These examples–good rules of thumb to practitioners, certain to be subject to quibbling by any academic reader–remind me of how wrong even the history of game theory is. As with many oversights by historians of science, the attribution of game theory’s invention credits the first theoretician (John von Neumann, who was smart enough to both practice and theorize), not the first practitioner (probably lost to history–but certainly by the 1600’s, as Pascal’s Wager actually lines up better with “game theory in the wild” in that he used infinite payoffs and actually did become religious). Practitioners, I would ignore the conventional history, theory, actual math, and long papers. Focus on easily used principles and heuristics that capture uncertainty, unboundedness, and asymmetries. Some examples:
- Principle: Prediction is hard. Don’t do it if you can help it.
- Heuristic: Bounded vs. Unbounded. Magnitude is easier to measure (or at least cap) than likelihood is.
- Principle: Every variable introduces more complexity and uncertainty.
- Heuristic: Make decisions for one really good reason. If your best reason is not enough, don’t depend on accumulation.
- Principle: One-time experiments don’t optimize.
- Heuristic: If you actually want to find useful methods, iterate.
- Principle: Anything that matters (power, utility, etc.) tends to be unequally distributed.
- Heuristic: Ignore the middle. Either make one very rich person very happy (preferred) or make most people at least a little happier. Or pull a barbell strategy if you can.
- The Academic Certainty Principle: Mere observation of reality by academics inevitably means they don’t get it. (Actually a riff on observer effects, not Hiesenberg, but the name is catchier this way).
- Heuristic: In game theory as in all academic ideas, if you think an academic stumbled upon a good practice, try it–but assume you will need trial and error to get it right.
- Principle: Since any action has costs, ‘infinite’ payoffs, in reality, come from dividing by zero.
- The via negativa: Your base assumption should be inaction, followed by action to eliminate cost. Be very skeptical of “why not” arguments.
So, in summary, most specific game theories are broken because they preference math (finite, tidy, linear) over practice (interconnected, guess-based, asymmetric). That does not mean you can’t use game theory in the wild, it just means that you should focus on structure over math, unbounded/infinite payoffs over solvable games, feedback loops over causal arrows, inaction over action, extremes over moderates, and rules of thumb over quibbles.