A catalog of weird-ass corners of game theory research


Originally published at: http://boingboing.net/2016/07/25/a-catalog-of-weird-ass-corners.html


Re the Winners Curse, the winning bidder might not be overpaying. If the bidders are all estimating item value independently, the other bidders may be undervaluing the item. This is particularly true if the winner is an experienced bidder. Even more likely, the winner may be making some forecasts of long-term trends and feel the the item is a good investment over the next five or ten years. He may be proved wrong, but the fact that he was the winner doesn’t guarantee that.


quantify empirically optimal outcomes

That string of words in the post is bugging the crap out of me. It’s hard for me to express why. It sounds a bit like someone made a word cloud of some game theory papers, took the most common ones and tried to mimic how those people speak without really understanding what they do.

It kind of reminds me of this:

Game theory is a methodology for modelling strategic interactions. What makes it’s “seductive,” is that it can show how/why people who are independently pursuing what’s in their personal best interest can result in situations where both people have “bad” outcomes despite the fact that there does exist some perfectly attainable other outcome where both people ended up better off. Conversely, it can also show that to achieve the best possible outcome, you may have to do something that, on the surface, seem like it’s the wrong thing to do. It’s also used to explore how feelings of guilt, justice, fairness, etc. often cause people to do things that aren’t in their best interest, e.g., “I think it’s unfair for you to do that, so I am going to stop you, even though it makes me worse off to stop you.” Or maybe it’ll even show that this does in fact make you better off because that purposeful choice of harming yourself now prevents a more serious harm in the future.

It can also show why the same situations can lead to different outcomes. Imagine something like dog poop. If you’re someplace clean, you like the cleanliness, so you pick up your dog’s poop. Because everyone thinks like that, the place stays poop-free. But if you’re someplace with tons of dog shit, you think “why am I going to go through the gross hassle of picking up and carrying poop when the area is going to be covered with poop no matter what I do?” And because of that, it gets even poopier, which leads others to think the same way, so it just gets even poopier still, even though all involved would prefer a poop-free area and are even willing to pick up poop to maintain that. That is, there’s a “good” equilibrium and a “bad” one, and game theory can help explain what leads to one vs. the other, and maybe offer insights on how to push things towards the good one.


This is really interesting, especially since I’m in sales and several of these are directly relevant. The big problem with most of these is that they tend to assume rational actors. The assumptions used for the calculations are correct, if people behave rationally, but as we know, people often don’t.


In a common value auction, the auctioned item is of roughly equal value to all bidders, but the bidders don’t know the item’s market value when they bid. Each player independently estimates the value of the item before bidding.

This summarizes game theory’s applicability to real life neatly. You make an unwarranted assumption and you arrive at an unwarranted conclusion.

Similarly, know more about the value of your car than a buyer doesn’t make it impossible to sell, but it might mean that you end up having to sell for less than you hoped.

Regarding biased media, here is one of the assumptions:

The media firms can withhold acquired information in their reports, but cannot lie.

Go watch Fox News and tell me how well that holds up in real life.

Game theory is about games. It is not about life. Life is a million times more complicated than any game theory simulation. And even when it isn’t, game theory ignores our ability to change our system of analyzing a problem when we discover that the system leads us to bad conclusions. Game theory tells us that in single iteration prisoners dilemma we should defect, but considering how much better off we are if we cooperate, this alone should tell us not to listen to game theory.

And in war, it’s much better to turn around and shoot your commanding officer. Carry on in that direction until you get whoever ordered the damn war to start in the first place. No game theory needed.


I only clicked through on three of the bullet points, but all three grossly misstated the principle they purported to illustrate. But they were clever!


That’s basically criminology’s Broken Windows Theory.


Well, sort of. People see litter and they think, “fuck it, I guess I can litter too.” That’s pretty well established, I think. Broken Windows Theory is that people see litter and think, “fuck it, I guess I can head up a drug empire and become a serial rapist.”


Slapdash popularizations of game theory like this are more infuriating the more you know about the subject… which I guess is probably true of all crappy popularizations. This is sort of like an article that says that the Kepler project has discovered thousands of new planets… some of which may contain exciting new cancer-fighting drugs. Well, yeah, technically all true, but really misleading.

I do an exercise in some of the classes I teach. I tell them that we’re going to play two games, for real cash prizes.* The first one is a “Prisoner’s Dilemma” tournament, which I almost always win, because I alone know the secret optimal tit-for-tat strategy. The second is the “game” where you have people pick a number between 0 and 100 that is closest to two-thirds of the average of all numbers. I always pick 0, which is what optimal strategy tells you to do.** The winner is usually around 20 or so.

The lesson here is that some “games” are more easily adapted to the actual chaotic irrational world than others (and that most aren’t, in any obvious way). Then we talk about nuclear deterrence and stock-picking robots.

* If I’m feeling REALLY educational, and if I didn’t win either game, I tell them that the cash prizes part was a lie, and it’s not my fault if they proceeded on faulty assumptions.

** The logic goes: Since the object is to pick a number two-thirds as high as other people’s choices, it can’t be larger than 66 2/3. But since everyone knows this, no one will pick a number higher than that. So it really can’t be any higher than two-thirds of THAT, or 44 4/9. But then nobody will pick a number higher than that, either. So it has to be less than two-thirds of THAT… etc. etc. etc., and that sequence converges on zero. A surprising number of students pick exactly 44.44444444, having figured out the trick but not why they need to keep applying it more than twice.


You just tripped the pricing robots for the “Trench Warfare” book. There are not too many copies out there, and I am going to have to wait a while to get one.


…it offers the seductive promise of being able to quantify empirically
optimal outcomes from thorny problems ranging from whether to go to war
to how to split the tab at a restaurant.

Unfortunately, this is completely wrong. Game theory has no way of mapping back to reality, since the payoff structure of agents in the real world is completely unobserved. Different payoffs, lead to different equilibrium outcomes. You can’t tell from observing two people playing a game whether they are engaged in the prisoner’s dilemma or a coordination game.

Payoffs are hard to define in games and are the biggest determinant of equilibrium outcomes. It is clear that we cannot simply ascribe payoffs to agents whose preferences we have never observed.


First, assume each human actor is an homogeneous perfect sphere 1.6 m in diameter…


Game theory research is all corners! The results are worth considering, but a failure to unpack implicit assumptions tends to suggest monoculture.

What constitutes a game is a number of actors with similar goals and/or rules. But any given model of economics is already a game in itself. Do the other “participants” share the same goals or methods? Of course, for the game researcher, it is convenient to assume that they do. But how applicable is this? Can you ever know with much certainty that others are playing the same game that you think you are playing?

So it becomes a hugely reductive process. Encouraging people to use the same systems while punishing those who can’t be easily quantified. In this sense, game theory research becomes yet another game itself.

It might be worth considering why one might bother valuing a sense of individuality if 99.99% of its motivations are still going to be redundantly hoarding resources, mating, approval of others, comfort, and survival when there are already billions of other “individuals” already treading those same well-worn paths.


I am so glad to read your post. I read through about four of these and they all just seem to be riddled with BS assumptions, and missing obvious factors. How exactly do the car owners trying to sell “Peaches” just “withdraw” from the market? How many sellers can afford to hold high value stock just because the market won’t come to their price?

A lot of smart people seem to be invested in it, so I can only assume there are genuine insights to be had, but the examples I read here today are very poor.


As if littering was the gateway drug of crime.

Yeah, the theory has gotten plenty of due criticism, not the least of which is overreaching for the ties to larger crimes.


I think Game Theory’s approach to human interaction is well summarized by this xkcd strip:


That’s what speciation is for!

“You lot can play that game amongst yourselves, we’re going to play this game over here instead.”


This topic was automatically closed after 5 days. New replies are no longer allowed.