The words people actually use at work and in the news.
The field guide
Explained in full
twelve ideas, and you have the working vocabulary
This is the whole subject’s working vocabulary. Everything else is a special case, a proof, or a name for a particular arrangement of these twelve.
Reading at
01 · also called a strategic situation
A game
A situation with more than one decision-maker, where each one's best move depends on what the others do.
The word is unfortunate, because it suggests either entertainment or manipulation, and it means neither. A game is any situation with interdependence: your outcome depends on choices other people are making at the same time or in response to you. That is the only requirement. It covers a price war, a custody negotiation, a queue, a treaty, an argument about the dishwasher and the entire advertising industry. The moment you accept that, most of the subject is just careful bookkeeping about who is deciding what.
The mistake almost everybody makes
Assuming a game means somebody is being manipulative. Most games have no villain in them at all, and the ones that hurt the most usually have nobody to blame.
What you do differently
Before you decide anything with another person in it, say out loud: this is a game, not a decision. That one sentence stops you from planning as if the world will hold still while you act.
02
A player
Any actor whose decision affects the outcome — whether or not they are in the room.
Counting the players honestly is the single highest-return habit in the whole subject, and the failure is always the same shape: the list contains the people in the meeting. The people who are not in the meeting still have strategies. A regulator who has not acted yet is still a player, because your move changes what they will do. An audience that never speaks is a player, because both sides are performing for it. Implementers are players, because a decision they will not carry out is not a decision. And there is nearly always a player whose interests are represented by nobody present — the future, the customer, the child, the person the policy is about.
The mistake almost everybody makes
Listing only the people in the room, then being surprised by the reaction of somebody who was never in it.
What you do differently
Write the player list before anything else, then deliberately add three: who enforces this, who watches this, and who has to live with it afterwards.
03
A strategy
A complete plan of action: what a player does, including what they do in response to each thing the others might do.
Two things separate this from ordinary planning. First, a strategy is complete — it includes your response to their response, not just your opening move. Second, it is defined by what is available, not what is intended, which means somebody's strategy set can include things they would never choose but which change the game by existing. A threat you would never carry out is still in your strategy set, and that is exactly why credibility becomes a separate problem. Most real-world option sets are wider than either party has admitted, and enumerating them honestly is the cheapest unstick there is.
The mistake almost everybody makes
Confusing a strategy with an intention. What matters is what somebody CAN do, because that is what the other side is reasoning about.
What you do differently
List every option each player has, including the ones you consider unthinkable. Unthinkable options change other people's arithmetic whether or not you would ever pick them.
04
A payoff
The value of an outcome to a player, in whatever currency that player is actually counting.
Payoffs are the most common place a good analysis goes wrong, because the analyst substitutes their own scoreboard for the other side's. Reputation, autonomy, the avoidance of humiliation, keeping an option open, and not setting a precedent are all real payoffs with real magnitudes, and in negotiation they routinely dominate the number on the table. The classic demonstration is the ultimatum game: people reliably reject free money to avoid being treated as somebody who can be squeezed, in every country it has been run in, and calling that irrational is simply refusing to write down the right payoff.
The mistake almost everybody makes
Assuming everyone is counting money. The most decisive payoffs in ordinary life are dignity, time, and not having to admit you were wrong.
What you do differently
When someone acts 'irrationally', do not diagnose them. Write down the payoff that would make their choice correct, and then check whether that payoff is real.
05
An incentive
The marginal pull toward one action over another, given the payoffs and the probabilities.
The operative distinction is between what people are told to do and what they are paid for, where paid means any currency from the payoff list. When a target is set, behaviour moves toward the target, not toward the thing the target was standing in for — response times improve while care worsens, arrest numbers rise while safety does not, engagement climbs while trust falls. Goodhart's law is the compressed version: a measure that becomes a target stops being a good measure. It is not cynicism, it is mechanism, and the fix is structural rather than moral.
The mistake almost everybody makes
Trying to fix an incentive problem with a values conversation. Sincerity and incentive point in different directions constantly, in perfectly honest people.
What you do differently
When behaviour is not what you asked for, stop asking what is wrong with the people and ask what the arrangement is actually paying for.
06
Information
The distribution of knowledge across players, including beliefs about others' knowledge.
Asymmetric information changes what is possible, not merely what is comfortable. When one side knows something the other cannot verify, the market for the good can collapse entirely — that is the lemons result, and it explains warranties, deposits, probation periods and qualifications far better than trust does. Meanwhile common knowledge, the condition where everyone knows, everyone knows that everyone knows, and so on without end, is a much stronger and rarer condition than shared knowledge, and coordination often requires it. A public announcement is not valuable because it informs people. It is valuable because it makes the thing common knowledge.
The mistake almost everybody makes
Treating 'they know' and 'they can prove it' as the same fact. Enormous numbers of stand-offs are between two people who both know and neither can show.
What you do differently
In any stuck situation, split the question three ways: what do they know, what do they think I know, and what could either of us actually demonstrate?
07 · also called Nash equilibrium
Equilibrium
A stable point where no single player gains by unilaterally changing strategy.
The concept's power is descriptive before it is prescriptive: it tells you which situations will persist. A practice that is individually rational and collectively awful is not a scandal waiting to be exposed, it is a stable state, and it will regenerate after any intervention that does not change somebody's payoffs. That is why exposure alone so rarely fixes anything, and why 'everyone knows this is bad' is perfectly consistent with it continuing for decades. Note also that equilibria are not unique — many games have several, some far better than others, and the whole practical question becomes which one a group lands in and whether it can be moved.
The mistake almost everybody makes
Hearing 'equilibrium' and thinking 'balance', 'fair' or 'settled by agreement'. It means none of those. It means nobody can move alone.
What you do differently
When something is persistently bad and nobody is obviously at fault, name it as an equilibrium out loud. It redirects the entire conversation from blame to structure.
08
A dominant move
A strategy that yields a higher payoff than an alternative against every possible play by the others.
Checking for dominance is the fastest useful diagnostic in the subject and takes about ten seconds: look down each column and ask whether one of your rows beats the other every time. If it does, you have found the arrow, and the situation is a trap rather than a negotiation. Its absence matters just as much — a game with no dominant move is one where the answer genuinely depends on your read of the other side, and those are the games where reputation, signalling and nerve start to do real work.
The mistake almost everybody makes
Believing that a bad outcome with a dominant move behind it proves somebody was foolish or malicious. It proves the opposite: everyone was correct, individually.
What you do differently
Ten-second test on any two-sided decision — is one of my options better whatever they do? If yes, stop strategising and start changing the payoffs.
09 · also called commitment
Credibility
The difference between a threat or promise you would carry out and one you would abandon when the moment came.
The classical image is burning your own boats after landing, and the point is not courage, it is communication: negotiating with me about retreat is now a waste of your time. Real-world commitment devices are dull and everywhere — a published price list, a policy with no exception process, an automatic transfer on payday, a constitution, a contract with a penalty clause. Every one of them is a decision made once, in the calm, so that it does not have to be made again under pressure. They are rare for an obvious reason: everybody wants the credibility and nobody wants to give up the exit.
The mistake almost everybody makes
Announcing a hard line during the pressure. A commitment made in the room is just something you said; it has to be made before, and visibly.
What you do differently
If you keep losing the same argument at the moment of pressure, stop trying to win it there. Move the decision earlier, make it binding, and tell people it is binding.
10 · also called the shadow of the future
Repetition
Whether the interaction is one-shot or ongoing, and whether either side knows when it ends.
Two conditions matter and both are load-bearing. The future has to be worth enough relative to today, and the end has to be unknown. If everyone knows which round is the last, cooperation unravels backwards from it — the last round is a one-shot, so defect; knowing that, the second-last is effectively a one-shot too, and so on to the first. That is why fixed-term arrangements behave worse near the end, why outgoing officials behave differently, and why 'we're closing this branch anyway' is such a reliable predictor of how the last month goes.
The mistake almost everybody makes
Treating a relationship as one-shot because this particular transaction is. Almost nothing is actually one-shot once you count reputation and the internet.
What you do differently
If you want people to behave better, do not appeal to their character — lengthen the shadow of the future. Make the relationship repeat, make the record visible, make the next round matter.
11
A signal
A costly, observable action that reveals private information because faking it would not pay.
This is why credentials, warranties, notice periods and probationary terms exist, and it is a better explanation of them than trust is. None of them are about honesty; each is a cost structured so that only somebody telling the truth can comfortably bear it. It also explains a large amount of otherwise baffling behaviour — conspicuous expenditure, elaborate qualifications for jobs that do not need them, and the entire practice of making people prove commitment by inconvenience. Cheap talk, by contrast, conveys information only when interests are already aligned, which is exactly when you needed it least.
The mistake almost everybody makes
Reading sincerity. Sincere people and convincing liars produce identical words; only the costs differ.
What you do differently
When you cannot verify a claim, stop trying. Design the offer so the two kinds of person have to behave differently, then watch which one you are looking at.
12 · also called a mixed strategy
Mixing
Randomising across options at a set frequency, so no opponent can exploit your tendency.
Two practical points. First, mixing only helps in the genuinely adversarial case where the other side's gain is your loss — reaching for unpredictability in a coordination problem actively harms you, because there both sides need to be readable. Second, the equilibrium mixing probability is often counter-intuitive: it is set so as to make the OPPONENT indifferent, not to make your own options equally attractive, which means changing your own payoffs may not change your own mixing rate at all. That result surprises people every time and is one of the sharper things the formalism gives you.
The mistake almost everybody makes
Using unpredictability where you needed to be predictable. In coordination problems, being readable is the whole job.
What you do differently
First ask whether the other side profits by predicting you. If yes, randomise properly. If no, the opposite applies — become as readable as you can.
