#250： The Art of Strategy
Episode Stats
Words per Minute
164.89862
Summary
Whether you're a businessman, a statesman, or a general, you're strategizing on a daily basis. So how can you do it better? Well, my guest today will provide some insights. His name is Barry Nailbuff, and he's the author of the book, The Art of Strategy: A Game-theorist s Guide to Success in Business and Life. And on the show today, Barry and I discuss how Game Theory can help you make better strategic decisions in all sorts of situations. For example, we explore why threatening to punish your child s sibling for bad behavior might be a more effective strategy than threatening the child himself. We'll discuss what Donald Trump can teach us about the promise and perils of injecting randomness into your strategy. And we also talk about how you can employ Game Theory against yourself to lose weight or even quit smoking.
Transcript
00:00:00.000
Brett McKay here and welcome to another edition of the Art of Manliness podcast. Whether you're
00:00:18.700
a businessman, a statesman, a general, or a parent, you're strategizing on a daily basis.
00:00:24.460
So how can you do it better? Well, my guest today will provide some insights. His name
00:00:27.740
is Barry Nailbuff. He's a game theory expert and the author of the book, The Art of Strategy,
00:00:32.240
A Game Theorist's Guide to Success in Business and Life. And on the show today, Barry and
00:00:36.600
I discuss how game theory can help you make better strategic decisions in all sorts of
00:00:40.700
situations. For example, we explore why threatening to punish your child's sibling for bad behavior
00:00:46.040
might be a more effective strategy than threatening to punish the child himself. I know that sounds
00:00:50.640
Machiavellian, but we'll explain the reasoning behind that. We'll discuss what Donald Trump
00:00:54.780
can teach us about the promise and perils of injecting randomness into your strategy.
00:00:58.920
We also talk about how you can employ game theory against yourself to lose weight or even quit
00:01:03.380
smoking. After the show's over, check out the show notes at aom.is slash game theory.
00:01:20.780
So you're the co-author of a book called The Art of Strategy. It's about strategic thinking,
00:01:25.920
particularly game theory. It's a topic I've long been interested in. But before we get into the
00:01:30.580
specifics of game theory, what it is, let's talk about strategy broadly. How do you and your co-author
00:01:36.620
define strategic thinking in your book? I mean, what is strategy really?
00:01:41.900
Strategy is different from decision-making. And the reason is that there are other people's
00:01:48.800
decisions that end up mattering. So when a lumberjack cuts down a tree or an engineer builds
00:01:53.580
a bridge, that bridge isn't responding, isn't thinking. The tree isn't a strategic player.
00:01:59.560
But when you make decisions in the real world, the success of your actions depends on how other
00:02:04.900
people will respond. And so that interactive aspect of the decision-making is what makes for
00:02:12.680
Okay. And I mean, I can understand why business people or military strategists need to understand
00:02:18.080
strategic thinking or game theory, but why is it important for even lay people? Like just people
00:02:22.740
who are mom and dads, husbands, wives, and why is it important for them to understand strategy?
00:02:27.740
I think everyone is interacting with decisions you make. Certainly kids, whether or not they want to
00:02:34.100
eat something or not eat something or stay up late,
00:02:36.980
or how one divides up chores in a household, people are interacting with each other. And you don't
00:02:45.440
make decisions in a vacuum. In physics, they say that for every action, there's a reaction equal and
00:02:51.680
opposite. But in game theory, that reaction can be changed. It can be influenced. And since we don't
00:02:58.000
act in isolation, we better figure out how other people are going to respond to what we're doing.
00:03:03.000
Right. But the thing is, I think strategy has a bad, has a PR problem, right? Ever since ancient
00:03:08.760
Greeks, you know, Odysseus was the wily one, and his strategic thinking was often looked down upon as
00:03:14.840
sort of, you know, unmanly or wily. And, you know, we think of strategy, we think of Machiavelli and being
00:03:19.820
manipulative. Is that what strategy is? Or is that, can strategy turn into that? Or can strategy actually
00:03:26.460
Well, another one of my books, co-authored with Adam Brandenberger, is called Co-Opetition.
00:03:31.020
And it's about competing and cooperating at the same time. And so you need to understand strategy
00:03:37.000
for how to compete more effectively, but you also need to understand it for how to cooperate more
00:03:44.440
Okay. So let's get into, you know, what makes up strategic thinking. You focus on game theory.
00:03:49.600
And I think a lot of people might have heard of game theory if they've seen A Beautiful Mind
00:03:52.820
about John Nash. But what is game theory? And what's the history of its development?
00:03:58.740
Sure. Well, game theory was started, created by a brilliant polymath at Princeton named John
00:04:06.320
von Neumann. And of course, John Nash, also at Princeton, less than 100 years old. So it's a
00:04:14.860
relatively speaking, pretty new science. And initially, it started out thinking about everything
00:04:19.980
from how one would hide and find submarines in warfare, to now anything from how to raise
00:04:28.320
kids, to bid in auctions, to find smart compensation contracts for executives. I thought it might be
00:04:37.820
That could illustrate how to do this, what's going on. But it depends, actually, to the extent you've
00:04:53.760
Well, maybe we can link to something, an online game that people can play online if they haven't
00:04:59.940
done it. Well, I mean, okay, so it started off primarily math driven. But as I read your book,
00:05:04.400
it seems like game theory has developed in something more interdisciplinary. Is that correct?
00:05:08.880
Well, since political science, sociology, law, all of that requires thinking about interactions,
00:05:17.460
it does actually cross many disciplines. Yes, indeed.
00:05:20.740
Yeah. It seems like a lot of behavioral science is influencing it, psychology as well,
00:05:26.500
Well, remember, if you think that economics is supposed to be a social science, as opposed
00:05:32.680
to asocial, we're supposed to understand how other people interact with us. And in that
00:05:37.660
sense, we have to take them as they are, not as you wish they would be. And so, in that
00:05:44.040
sense, it's certainly, we don't have a behavioral game theory, but a simple reason that that would
00:05:49.860
be redundant, that how other people behave is intrinsic and central to any discussion of
00:05:58.480
Okay. And I'm sure we'll get into some games I'm sure people might are familiar with, but
00:06:03.140
let's start getting to the nitty gritty here. So, you start off saying in the book that the
00:06:09.020
first step when you find yourself in a strategic game. So, let's start, how do you know you're
00:06:14.220
in a strategic game? Like, is it just whenever there's someone else or other people in a decision
00:06:22.040
If you're acting with other people and they can react to what you're doing, or their actions
00:06:28.380
influence your success, then you're pretty much in a game. So, to give you a couple recent
00:06:34.420
public policy examples, much of the debate about the ACA or Obamacare actually is really
00:06:40.900
a game theory discussion. So, whether or not you require people to buy healthcare, well,
00:06:49.720
if you don't require healthy people to buy healthcare, then the only people who will end
00:06:55.520
up buying it are those who have pre-existing conditions, who aren't healthy. That means the
00:07:01.240
premiums are going to have to be very high. That means that only the even sicker people will end up
00:07:06.540
buying healthcare, which means the premiums will have to be higher still. And the end result is
00:07:11.780
you'll get what's called a death spiral, and nobody will end up being able to afford healthcare.
00:07:17.260
And so, the idea of understanding the interaction between who will buy and what the effective premium
00:07:24.440
is would be a classic example of what George Ackloff won a Nobel Prize for, something called the
00:07:31.260
market for lemons. And so, I mean, game theory, games can get very complex or very simple. I mean,
00:07:36.600
a simpler one would be just negotiating what time your child's going to go to bed, right? That's a very
00:07:41.580
simple one, but the Obamacare instance, that's very complex. There's a lot of different people
00:07:46.360
involved, a lot of different factors. So, you say that whenever you find yourself in a strategic
00:07:52.620
game, when there's decisions being made that involve other people, the first thing to do is figure
00:07:57.560
out what kind of game it is, and it's either simultaneous or sequential. What are the differences
00:08:02.800
between the two, and how will your strategy change based on what kind of game it is?
00:08:08.840
So, let me just go back for a second. I don't think the game with your kid about bedtime is so simple,
00:08:14.140
because remember, it's not just one night. This is a classic repeated game, and you might decide that
00:08:20.000
today it's not worth fighting it, but you need to be tough to have a reputation, because otherwise,
00:08:24.740
in the future, you won't have any credibility. And so, most games are neither simply sequential or
00:08:35.500
simultaneous. They're mixtures of both. They're not single shot. They go on again and again, so they're
00:08:41.480
repeated. In a sequential move game, it's much like checkers. I make a move, you make a move,
00:08:48.500
I make a move, and we alternate making moves. So, when I'm making a move, I have to think about how
00:08:54.680
you're going to respond. When your response is taken into account, it's thinking about what I'm
00:08:59.460
going to do in response to your response, and so on. In contrast, a simultaneous move game doesn't
00:09:07.420
really require us to move at the exact same moment, but it means I have to make a move without
00:09:13.380
knowing exactly what it is you've done at the time I'm making my decision. So, a very simple example
00:09:22.420
of that would be, I am placing a bid in an auction, and I'm bidding without knowing what your bid is.
00:09:30.280
Another example could be, quite soon, I'm voting. And when I place my vote, I don't know what the other
00:09:36.620
people have done in their voting booth. And so, when I'm thinking about, do I want to support the
00:09:41.360
candidate who I really like, or make a protest vote, I don't really know what other people's
00:09:48.280
decisions have been made at the same time. So, it sounds like with simultaneous games,
00:09:52.940
there's a lot more uncertainty. With sequential games, there's a bit more certainty than simultaneous.
00:09:57.840
Exactly. Note, again, just in voting, it's not that we're literally all voting at the exact same
00:10:02.560
moment. But when I'm voting, since I don't know what you've done, it's as if we're moving at the
00:10:08.600
same time. And you're absolutely right that in sequential move games, it's much, much easier
00:10:13.780
to solve, because I know everything in terms of what you've done.
00:10:19.380
Right. So, with sequential games, you can look forward and start reasoning backwards,
00:10:27.960
Exactly. So, the nice thing about sequential move games is we know how to solve them.
00:10:31.660
Yeah. And essentially, you can play out every possible scenario, and you can figure out what
00:10:37.500
is the best way of playing the game. And so, that's really easy to do in tic-tac-toe, which
00:10:44.980
is why nobody plays tic-tac-toe once they're above seven, because you can figure out going
00:10:49.920
to the center means we're always going to get a tie. In contrast, for a while, it was thought
00:10:55.660
that chess was too hard to solve, or Go was too hard to solve. And so, even though, in theory,
00:11:02.660
there was an optimal way of playing it, a way of guaranteeing either a victory, a tie, or a tie,
00:11:10.840
since nobody knew what that was, we could still enjoy playing the game. Pretty soon, I'd say that
00:11:20.160
Right. Because computers will allow them to map out all the sequences, possible sequences.
00:11:25.660
Right. I think the top six or seven best chess players in the world are all now computer
00:11:31.280
That's crazy. So, I mean, so I think we can all intuitively do this, you know, sequential,
00:11:37.060
you know, forecasting, right? You know, looking forward to reason backward, when things are
00:11:40.740
pretty simple. But, I mean, some of these sequential games can get really complex. I mean, chess is a
00:11:44.720
perfect example. There are millions upon millions of different sequences. So, how do you, as a game,
00:11:52.080
you know, as a game theorist, track those sequences and then figure out which sequence will probably
00:11:57.540
be the one that will play out in the real world?
00:12:01.340
Well, when we didn't have computers, you use heuristics. You say, I think that owning certain
00:12:09.580
parts of the board, certain positions are stronger than others. Certain pieces are worth more than other
00:12:14.980
pieces in terms of power. And so, you look for simple rules that are usually right. Maybe they're
00:12:22.420
not always right. In other cases, you can do simulation. In other cases, you base this on
00:12:29.720
experience. Depends a little bit on how important the game is and how often it's going to be played
00:12:36.560
in terms of how much you want to go and figure out how to solve it. A classic example, though,
00:12:43.800
sometimes failure to understand the right strategy occurs in sports, where when teams are thinking
00:12:52.440
about when it is to go for a two-point play rather than a one-point play after a touchdown,
00:12:59.940
sometimes they fail to look forward and reason backward. So, if you're down by two touchdowns
00:13:06.160
with not that much time left to go and you score one touchdown and make the extra point,
00:13:13.940
there are cases where the coach has then gone for the two-point play on the second touchdown
00:13:20.280
with just a few moments left to go so as to win rather than tie the game. And it turns out that that's
00:13:27.160
a fine strategy. You might argue it's worth taking that risk. But if you thought that's what you'd want
00:13:32.880
to do, then you should have gone for the two-point play on the penultimate touchdown. The reason being
00:13:38.580
that you have to make both a one-point and a two-point play. It doesn't really matter which order you make
00:13:43.680
them in. But if you miss the two-point play on the first shot, then you have another chance to make a
00:13:48.880
two-point play on the second one and still get a tie. Ah, okay. That makes sense. I didn't think about it
00:13:54.500
that way. I've had those moments where my coach decided to go for it later on and we ended up
00:14:00.000
losing. So, it sounds like sequential games. Are there any examples like real-life examples of like,
00:14:06.180
you know, I'm talking about, you know, parent to child or business to business where there's
00:14:10.440
sequential games? It seems like the examples we've been talking about are very, you know,
00:14:13.540
they're games, like literal games, tic-tac-toe, football, chess. Any examples where there's less
00:14:20.040
structure, but there's still a sequential game involved? Well, I'd say if you think about
00:14:25.840
appointing a Supreme Court nominee, the president goes and suggests a candidate. The Senate then,
00:14:33.720
in theory, advises and confirms. And so, sure, there are discussions and movements ahead of time,
00:14:42.600
but essentially the Senate doesn't get to, well, they can try and change this and preempt it by
00:14:50.720
saying, unless you pick X, we're not going to accept anyone. But essentially, the president moves
00:14:58.540
first and then the Senate moves second. In other cases, Congress and the Senate will pass a bill and
00:15:05.900
then the president decides whether to sign it or to veto it. And if vetoed, the Congress decides
00:15:13.360
whether to override the veto. So, it's still obviously a simplification, but various laws have
00:15:22.460
created a structure which puts some sequentiality into the moose. Right. But there's still some
00:15:27.920
simultaneous things going on. As you said earlier, games are usually a mixture of the both,
00:15:32.020
simultaneous and sequential. Absolutely. But I'd say in this case, there's still a predominant
00:15:37.840
aspect of sequentiality. Okay. So, let's move into specific games that I think people might have
00:15:43.320
heard or have experienced with that kind of highlight some insights into game theory. One
00:15:50.020
of them, I think a lot of people have heard of, is the Prisoner's Dilemma. For those who aren't
00:15:54.440
familiar with it, can you briefly explain what the Prisoner's Dilemma is and then what insights
00:15:58.560
about game theory does it provide us? I think there's a sense in which people think
00:16:05.500
the Prisoner's Dilemma and game theory are synonymous, which is unfortunate because game
00:16:10.880
theory is a lot more than the Prisoner's Dilemma. But to the extent that anybody has watched a crime
00:16:16.720
thriller or the wonderful English TV show Golden Balls, what you have is a situation where both
00:16:26.820
individuals, each individual, has an incentive to cheat or to confess no matter what the other side
00:16:33.360
does. So, here you're interviewing two prisoners in separate rooms. A crime has been committed. If
00:16:40.000
neither prisoner confesses, then they both get off. But if one confesses while, or they actually don't
00:16:46.520
quite get off, they get a light sentence. If one confesses while the other keeps mum, the one who
00:16:52.680
confesses gets off and the other one gets a very tough sentence. While if they both confess, then they
00:17:00.160
each get a medium sentence. And so, the idea is that if one side keeps mum, then I can get off by
00:17:07.320
confessing. And if the other side confesses, then I really had better confess, because otherwise I'm
00:17:12.520
going to get a really severe sentence. And so, since no matter what I think the other side is going to
00:17:19.760
do, I have an incentive to confess. But of course, when both confess, they're both worse off than if
00:17:25.920
neither one does. And this is this paradox, because in some sense, when we each act in our individual
00:17:32.220
interest, the end result is bad for us in a collective sense.
00:17:37.260
Right. And any instances where, I mean, the prisoners don't look off, you know, obviously
00:17:41.200
happens whenever you're doing the separating the prisoners and trying to negotiate confessions. But
00:17:45.460
any other like real life examples where you see prisoners dilemmas act play out?
00:17:50.100
Well, I think there's the tragedy of the commons, which is a multi-person version of this. And so,
00:17:58.920
if you think about global warming or air pollution, it's in my interest to not change my lifestyle,
00:18:07.640
to continue driving cars or flying planes or to support industry. And each of us does this.
00:18:20.100
And the end result is that we end up with global warming and we're all worse off. And so,
00:18:25.400
we need to somehow collectively agree to work and cut back carbon emissions as opposed to do this on
00:18:32.200
an individual basis. Now, the advantage that we have here is that we can actually talk to each other
00:18:38.460
and monitor what the other side is doing. So, if we had to make these decisions in isolation and not
00:18:44.040
do treaties, then we'd find ourselves in the multi-person version of prisoners dilemma. And the result
00:18:49.700
would be disastrous. And so, the good thing is we don't let prisoners talk to each other and make
00:18:54.880
a pact, but we do let countries do that. And that's why we need to actually do it via treaties
00:19:00.800
and alliances as opposed to counting on people acting in their own self-interest.
00:19:08.000
Right. So, it sounds like if you find yourself in a prisoner's dilemma situation is to open up
00:19:13.460
communication. That's how you avoid those scenarios.
00:19:15.920
Exactly. And of course, sometimes the goal is to put other people in a prisoner's dilemma and prevent
00:19:22.140
them from being able to communicate. So, you've hit on a key point, which is sometimes the best way
00:19:28.660
to play a game is to change the game. That if the game isn't working out for you, don't accept it the
00:19:35.600
And how would you, I mean, so what's an example of changing the game so you can make things better?
00:19:45.220
You can add more players to the game. You can say, okay, if in fact I discover that you snitched,
00:19:55.900
or you discover I snitched, and we go to prison, people will beat up snitches. And so, it may look
00:20:01.100
like it's a good idea to do the confession, but actually the game isn't over yet. And that we,
00:20:07.860
in fact, want to make sure that there are other players who will end up punishing us for doing
00:20:16.840
the strategy that at the short run seems to be in our interest.
00:20:21.540
Okay. So, you extend the game, make it longer. Yeah. I mean, I guess that you bring that,
00:20:26.340
that ties into you bringing the tit-for-tat approach that one game theorist developed back,
00:20:32.260
you know, a couple decades ago. One of these sort of prisoners dilemma type games where there was
00:20:36.960
multiple prisoner dilemma games. So, like, you know, you would confess one time, and the other
00:20:42.160
guy would confess, and then the other guy would know what you did, and then he would retaliate
00:20:47.240
for, you know, sticking it to you. And at the time, they thought that this tit-for-tat approach
00:20:53.380
was a good way to solve the prisoner's dilemma. But you argue in the book that it's actually not
00:20:58.600
that great of an approach. So, it's probably not the case that we're going to play tit-for-tat
00:21:03.100
with prisoners because they would have to be serious recidivists to be doing this tens or
00:21:09.960
hundreds of times in a row. Or it's more likely, the prison dilemma exists when companies are trying
00:21:19.080
to find ways to circumvent competition and to come up with, say, implicit, sometimes even explicit,
00:21:26.240
collusion. So, firms, let's take airlines as a case, might want to keep prices high.
00:21:32.800
And so, the question is, I want to go and do a little bit of a price cut and steal some market share
00:21:40.480
from you. And it may not be in my rival's interest who has a larger fraction of that particular route
00:21:49.520
or has more to lose by coming down and matching me. But if I understand that the person is going to do
00:21:57.280
that, they will do a tit-for-tat strategy. And if I come down, they're going to come down.
00:22:01.280
Then, I don't get the gain from doing any type of price cut. They're going to punish me. And as a result,
00:22:11.780
I will learn that this type of cheating, and cheating, by the way, here is, in some sense, cheating on the
00:22:17.920
collusion. So, it's cheating on the cheating, if you want. It doesn't actually pay off. Now, the problem with
00:22:27.140
simple, mechanical responses to when you think somebody else is cheating is that every now and
00:22:33.340
then, you're going to make a mistake. And you're going to think somebody cheated even when they
00:22:36.740
didn't. And so, you're going to punish them. And then what's going to happen is they're going to
00:22:40.840
punish you for punishing them, and then you're going to punish them for punishing you for punishing
00:22:44.500
them. And you're going to get yourself into one of these spirals, perhaps a little bit like what
00:22:50.540
we see in the Mideast, where it's hard to even figure out who started it. But now, we're just
00:22:56.980
in this endless cycle of retaliation. And how do we ever get out of it?
00:23:02.340
Right. This is interesting. With the airline thing, I mean, there's laws in place where companies
00:23:08.200
can't explicitly collude, right? They can't get together in a sort of a cabal and say, all right,
00:23:13.740
here's the price we're going to set the tickets at, so all of us can benefit from it. So, because
00:23:18.960
they can't do that, they have to do these sort of implicit collusion. Okay, well, if you're going
00:23:23.280
to raise the price, I'll raise the price, and it kind of evens out. Well, it's a little bit of,
00:23:28.520
I see my rival raise price. Now, I can do two things. I can take advantage of that and get some
00:23:35.180
extra share, or I can match the price. And while I don't want to match the price, I want to give my
00:23:39.760
rival incentive to have taken this action and encourage them to keep the price high. And so, even
00:23:46.760
without talking to the other side, I can figure out that this might be in my long run interests.
00:23:55.120
People have been talking about the effects of greater concentration in industry, and that
00:24:02.540
leading to higher profits and perhaps higher prices. But what often they've missed is something
00:24:09.900
called common ownership. And two of my colleagues here, Florian Etter and Fiona Scott Morton,
00:24:17.120
have been working on this. And one of the things they've discovered is that most companies,
00:24:23.800
most competitors have the same owner. So, Vanguard or Fidelity own huge fractions of all the different
00:24:32.420
airlines, or all the different pharmaceutical companies. And when you own A and its rival, B,
00:24:39.800
you know, you say, well, wait a second. Guys, I don't think this is such a good idea for you to go
00:24:45.340
and keep on cutting prices or try and steal share from each other or adding capacity. You know, just
00:24:50.740
lay off a little bit here. And so, essentially, now we have to worry not just about firms colluding with
00:24:57.780
each other, but the person who owns both of the firms, encouraging them to not really compete
00:25:05.520
So, another type of game that you mentioned in the book that's sort of a bit different from
00:25:08.920
the Prisoner's Dilemma is what you call a confidence game. What is a confidence game,
00:25:15.540
and how does it differ from a Prisoner's Dilemma?
00:25:17.480
You know, I think Maria Konnikova has written a great book about this, and she talks about how it
00:25:26.120
is that con artists end up fooling people. And this is some wonderful applications of game theory.
00:25:34.140
In particular, if you're ever wondering why it is that spam that's trying to get phishing exercises,
00:25:43.140
trying to go and get you to send lots of money to Nigeria or someplace else,
00:25:48.120
are full of spelling errors. You can say, well, okay, guys, you know, come on, run through a spell
00:25:54.540
checker. I mean, how bad, how stupid you have to be? And the answer is, they don't want to waste
00:26:03.960
their time with people who aren't gullible. And so, they do things that are particularly bad.
00:26:10.720
Because if, in fact, you can't spot the super obvious nature of the spam, then that's saying,
00:26:21.860
okay, I've got a real stupid fish here. And I've hooked a great one, and I can go after this person.
00:26:28.940
Whereas, if they find people who are particularly sophisticated, later on, those folks will catch on,
00:26:35.320
and they will have wasted a lot of their time. So, it's an interesting point that they make the letters
00:26:42.740
intentionally simplistic, unrealistic, riddled with spelling errors, because they're trying to find the
00:26:54.660
Right. So, it's sort of like they're signaling. They're putting out a signal to find out the signals
00:27:02.600
Exactly. It's the latter point that's the key, is that they're looking for their victims to signal
00:27:07.480
that they're not paying attention, that they're gullible. And if they don't have spelling errors,
00:27:14.820
then the other side can't really signal their gullibility.
00:27:19.820
All right. So, these spam guys from Nigeria, they're pretty smart.
00:27:27.380
Right. Right. So, let's talk about, I think, something that people might have heard before
00:27:33.400
because of popular culture. I think A Beautiful Mind might have helped. But like, this idea of
00:27:37.480
the Nash equilibrium. I've heard it before, over and over, and I never quite understood until I read
00:27:43.220
your book. But for those who aren't familiar, what is a Nash equilibrium in game theory? Does it
00:27:49.280
happen in sequential games or does it happen in simultaneous games or both?
00:27:54.400
The concept of a Nash equilibrium was developed to help us understand what will happen or a resting
00:28:02.380
point in a simultaneous move game. And so, the challenge is, what do we do in a world where
00:28:11.700
I think, that you think, that I think, and ad infinitum, will happen? I don't get to see what
00:28:19.940
you're doing. You don't get to see what I'm doing. And so, what move is it that I want to make
00:28:24.700
in a world where I have to anticipate what you're doing when you're anticipating what I'm going to do?
00:28:31.680
And that seems like it's an infinitely recursive logic. And it's not clear how you ever cut through
00:28:36.400
that not. And so, the brilliant insight of John Nash is, well, are there a set of strategies or moves
00:28:44.360
such that if I'm doing A and you're doing B, and I think you're doing B, and I think you think I'm
00:28:52.440
doing A, then I still actually want to do A. So, that is, if you've correctly anticipated what I'm
00:28:57.940
going to do, and I've correctly anticipated what you're going to do, neither of us wants to change
00:29:04.220
what we're doing. And that is an attractive candidate for how a game will be played when
00:29:13.660
we can't actually see what the other person has done. So, it sounds like the goal is to get to a
00:29:19.580
Nash equilibrium when you're strategizing. No. No? Okay. Right. It is not a goal. In particular,
00:29:29.300
in the Prisoner's Dilemma, the Nash equilibrium is that both confess. So, that's not a good outcome.
00:29:34.220
So, if we want to predict how a game might be played, then a Nash equilibrium is a good starting
00:29:41.180
place for what players might end up doing. But it is not necessarily desirable as an outcome.
00:29:50.660
Okay. So, as you work through the Nash equilibrium or trying to figure out what the Nash equilibrium
00:29:57.220
is, you're going to find, I guess, what you call dominant strategies in the mix. And then,
00:30:03.780
is that the thing you should take? Is the dominant strategy for you?
00:30:06.440
Well, if there's a dominant strategy, then life is easy. Because it says, whatever I think the other
00:30:13.120
person is doing doesn't matter. It's always the case that I want to do A. A is better than any other
00:30:19.960
strategy. And so, I don't have to consider what the other person is doing. And so, that allows me a real
00:30:24.520
cheat out of this Gordian Knot. The challenge, of course, is that there aren't that many games
00:30:34.760
where there really is a dominant strategy. And so, we have to, more often than not,
00:30:45.160
refer back to a Nash equilibrium to get a better sense of how we should play the game.
00:30:49.300
Okay. So, this idea, when you're strategizing, doing a simultaneous strategy, like you said,
00:30:58.080
you're doing this recursive thing in your head. I'm thinking A, and I think my opponent or
00:31:03.740
competitor is thinking that I'm thinking A. And if he's thinking that, then I'm going to do this.
00:31:09.920
It sounds like it would be good to inject randomness, right? So, you can throw people
00:31:15.300
off. Well, again, it depends. Is my goal to coordinate with you or to get an advantage over
00:31:23.000
you? So, here's, I'll give you two versions of a simultaneous move game. Here's one. You and I
00:31:32.600
both have to pick a number. And if we pick the same number, then we both get that amount of money
00:31:39.040
paid for by a third party. And that number, let's say, has to be between one and ten. It's an integer.
00:31:47.460
So, if we both pick four, we both get four. If we both pick five, we both get five. If you pick four
00:31:55.140
and I pick six, we both get zero. So, yeah, we both pick ten, if we're cooperating.
00:32:00.580
Okay. Well, now you've ruined it because you gave me a sense of what it is you're going to do before
00:32:05.720
we played. Oh, okay. Darn it. But that's okay. So, one of the things this game illustrates is that
00:32:12.920
there's a lot of Nash equilibrium. If I'm going to pick six, what is it that you're going to do in
00:32:19.000
this game? What do you want to do? Well, I'd pick six, too, if I knew that. Exactly. Now, you might say,
00:32:24.740
okay, well, six, six is not such a great outcome. We could both pick ten. I got it. But you have to
00:32:35.020
be sufficiently confident that I'm going to pick ten in order for you to pick ten. And in fact,
00:32:43.360
this is a game where there's multiple Nash equilibria. You might decide, well, it's not that
00:32:48.560
hard to pick between them. Because isn't it obvious that everyone should pick ten? Okay. But
00:32:54.760
actually, this game helps explain a lot of why we see some countries developing faster than others.
00:33:02.220
So, in much of the world, there's corruption. And corruption creates problems with police,
00:33:08.500
with doing business. And you can say, you know, I think a world in which there's no corruption
00:33:14.660
is a better world. But if I think you're going to be corrupt, then I have to be corrupt. And if you
00:33:20.320
think I'm going to be corrupt, then you're going to be corrupt. And so, in such a case, we end up both
00:33:26.840
picking two, if you'd like, and we each get two rather than ten and ten. And so, if you're scared
00:33:33.100
that I'm going to pick two, even if you want to pick ten and you know I want to pick ten, if I'm scared
00:33:40.720
that you think I'm scared, then you might pick two because you think I'm going to pick two.
00:33:45.400
And we both know there's a better answer. But neither of us
00:33:48.500
has the confidence that we're willing to go there.
00:33:51.820
Okay. Well, and so that's one example of a game. What's another
00:33:54.660
example? You said there was another example of a game?
00:34:18.240
Well, we both get a third party who's going to pay
00:34:46.640
that we want to be uncoordinated, but it's not quite