Your next move depends on theirs.

In 1979, political scientist Robert Axelrod invited specialists to submit computer programmes for a tournament. Each programme would face the others in a repeated game with a peculiar problem: betraying the other player offered an immediate advantage, but if both players kept betraying one another, both ended up worse off.

Fourteen strategies arrived. Some tried to recognise and exploit weaknesses. Others retaliated severely. A few took a more accommodating approach. There were elaborate rules for anticipating what an opponent might do next.

The winner, submitted by mathematical psychologist Anatol Rapoport, was almost embarrassingly simple.

Its name was Tit for Tat. It began by cooperating. Thereafter, it did whatever its opponent had done on the previous turn. Cooperate with it, and it cooperated. Betray it, and it retaliated. Return to cooperation, and it did too.

Under the tournament’s rules, Tit for Tat could not outscore an opponent in an individual match. Its best possible result against any one rival was a draw. Yet when the points from its encounters with the whole field were added up, it finished first. In a second tournament, with more entrants who already knew the earlier result, it won again.

That sounds like an agreeable moral lesson. For anyone responsible for a business, it raises a harder question. How often do we mistake getting the better of someone for creating a better outcome?

The deal does not end when you make it

Imagine the founder of a specialist technology company deciding to cut prices to win an important customer. On paper, the calculation may be convincing: the new price clears a purchasing obstacle, and the additional volume appears to justify the smaller margin.

But the customer gets another decision. They can buy now, or wait to see whether the discount becomes permanent. A distributor may question its own margin. Existing customers may ask for equivalent terms at renewal. A competitor may respond with a different offer, or decide to compete on service instead.

None of these reactions is inevitable. Their possibility is what makes the decision strategic.

Game theory examines situations in which the consequences of one person’s choice depend on what others choose. It is concerned with incentives, expectations and responses. Its more useful question for a leader is rarely, What can we get away with? It is, What will our decision give other people a reason to do?

That is easy to overlook in a business case. We estimate what a move will cost and what we hope it will produce. The responses it invites can become a footnote, even though they may determine whether the plan works at all.

A company can make an attractive offer that teaches buyers to postpone their purchases. It can negotiate a contract so aggressively that a promising supplier stops investing in the relationship. It can protect its intellectual property so completely that potential partners decide integration is not worth the risk.

The immediate transaction may look successful. The next transaction takes place in a world the first one helped create.

Why the cleverest player did not win

The game Axelrod used was the prisoner’s dilemma. In its familiar form, two players choose independently whether to cooperate or defect. Both earn a good reward by cooperating. Either can earn a larger individual reward by defecting while the other cooperates. But when both defect, each earns less than they would have earned through mutual cooperation.

Played only once, the incentives pull both players towards defection, even though the result is worse for each. Repeat the encounter and something changes: today’s choice can affect how the other player treats you tomorrow.

Tit for Tat exploited that possibility in an unusual way. It did not attempt to predict an opponent’s private intentions or trick them into an irreversible mistake. It made its behaviour legible. Cooperation was rewarded immediately. Defection brought a response. One cooperative move was enough to reopen the relationship.

Axelrod described the successful qualities as being willing to cooperate first, able to retaliate, ready to forgive and easy to understand. The precise labels are less important than their interaction. Trust without any consequences can invite exploitation. Retaliation without a way back can turn a small dispute into a permanent one. Unpredictable behaviour makes even a willing partner cautious.

The tournament was a mathematical model, not a study of executive character. Its players had restricted choices, defined payoffs and no uncertainty about what the opponent had actually done. Real organisations have lawyers, technicians, incomplete information, disagreements about value and people who interpret the same incident differently.

That difference becomes important when we try to apply the lesson.

The most convincing promise changes your own options

Taiwan’s TSMC has built its semiconductor business around a commitment that can seem like a missed opportunity.

It manufactures chips designed by other companies, while choosing not to sell semiconductor products under its own name. In its 2025 annual report, TSMC says it pioneered the pure-play foundry model, and that because it does not design, make or sell chips under its own brand, it “never competes with its customers”.

For a chip designer, that distinction matters. Working with a foundry involves revealing designs, discussing technical difficulties and planning future products. A manufacturer that also competes in those products creates a different set of risks, however persuasive its assurances of confidentiality may be.

TSMC has chosen to organise its business around the absence of that particular conflict. The commitment is more substantial than a sentence about partnership on a website: it shapes what the company does and, importantly, what it has chosen not to do.

That does not eliminate dependence, concerns about intellectual property, manufacturing risk or geopolitical exposure. Customers still have to evaluate those issues. But the business model changes one part of the calculation they make before trusting the supplier.

This is a useful distinction for founders who speak about building ecosystems. A potential partner does not decide solely on the strength of your product or the warmth of your intentions. They also consider what you might rationally do once you have their data, their customers, their technical knowledge or their dependence.

A credible commitment makes some feared future behaviour less likely or more costly. A documented interoperability policy, enforceable service obligations or clear contractual limits on data use can matter for that reason. Their value depends on the detail and whether the other side can rely on them.

When competitors need the same future

In 2012, Dutch lithography manufacturer ASML announced an unusual arrangement. Three major customers (Intel, TSMC and Samsung) would help finance the development of the next generation of chipmaking equipment.

According to ASML’s October 2012 announcement, the three agreed to contribute EUR 1.38 billion to research and development over five years. Separately, they paid EUR 3.85 billion for equity interests. The programme focused on extreme ultraviolet lithography, among other technologies. The two financial amounts were distinct parts of the arrangement.

These were substantial companies with their own commercial interests, different market positions and reasons to outperform others. Yet they were also exposed to a common technological bottleneck: the equipment needed to manufacture future generations of chips was extraordinarily difficult and expensive to develop.

ASML’s office tower in Veldhoven, 2011. Johfrael, CC BY 3.0. Source

One company making progress on its own products would have limited value if the manufacturing technology on which those products depended failed to advance. Helping to fund a supplier could therefore serve each participant’s own interests, even while they continued to pursue their separate commercial strategies.

This is where the idea that every competitive situation must have a winner and a loser begins to break down. Some advantages are created by preventing someone else from succeeding. Others depend on several parties making a difficult thing possible in the first place.

The ASML arrangement was neither a guarantee of technical success nor evidence that competitors should always collaborate. Research can fail. Shared projects can produce new dependencies. And cooperation among competitors has legal boundaries: the European Commission’s Horizontal Cooperation Guidelines distinguish potentially beneficial research cooperation from conduct that restricts competition. Decisions about pricing, bids and customers must remain independent, and any proposed arrangement requires proper legal assessment.

The more useful point is strategic. Before asking how to capture a larger share of an opportunity, a leader might ask whether the opportunity can exist without contributions from people whose interests are not identical to theirs.

Sometimes a rival is also a customer, a supplier, a source of expertise or another party exposed to the same constraint.

One misunderstanding can change the game

There is a flaw in Tit for Tat that becomes especially revealing when we leave the clean world of the original tournament.

The rule responds to the last move it observes. Suppose two parties intend to cooperate, but one action is misunderstood as a betrayal. The other retaliates. That retaliation is then taken as deliberate hostility, prompting a response in return. A single error can produce a sequence of entirely sincere retaliations.

Real business relationships are full of opportunities for such mistakes. A shipment is late because of a problem neither party expected. A contract clause is interpreted differently. A technical team describes an integration risk, and the commercial team hears an attempt to renegotiate.

The sensible response is not always instant forgiveness. There may be a genuine breach, a repeated pattern or a risk that must be contained. But an organisation that treats every ambiguous signal as proof of hostile intent can manufacture the behaviour it fears.

Subsequent game-theory research, including a 2018 study in Nature Communications, has explored how error and interference can undermine simple reciprocity. The practical implication is narrower than a universal instruction to be generous: when information is imperfect, leave room to verify what happened and make recovery possible where recovery is warranted.

Leaders often spend great effort designing what happens when everyone performs as expected, and less on what happens when somebody does not.

There is no strategy that always wins

It would be convenient to turn Axelrod’s result into four rules for becoming a successful executive. It would also miss an important part of the research.

In 2015, Amnon Rapoport, Darryl Seale and Andrew Colman re-examined the tournament results. They argued that the apparent superiority of Tit for Tat depended on the design of the tournament, including whether points or wins counted as success, and suggested that a different payoff matrix might remove its advantage. A strategy that performs exceptionally well when total points matter may look very different when the objective is to win individual matches.

The surprise was real, but it does not justify prescribing the same behaviour in every setting. As the Stanford Encyclopedia of Philosophy notes, no strategy is best in every environment, and how well Tit for Tat does depends on which other strategies are present.

The distinction matters in business. A one-off tender does not have the same incentives as a twenty-year supply relationship. A partner whose promises cannot be verified presents a different problem from one whose performance is observable. A buyer who needs a validated, reliable instrument is making a different decision from someone shopping for a commodity.

Game theory cannot tell a founder what a competitor secretly intends. It cannot assign reliable numbers to outcomes that have never been studied. And it cannot make a technically unsuitable product attractive through clever incentives.

Its value is more modest, and perhaps more demanding. It forces us to be explicit about who has a choice, what each party is trying to achieve and how our own action changes their options.

The question to bring into the boardroom

Consider the next consequential decision in your organisation: entering a market, changing a pricing model, choosing an exclusive partner, opening a platform or negotiating a long-term supply agreement.

Most leadership teams can explain why the proposed move looks attractive from their own position. They have estimated the costs, assessed the immediate benefits and prepared a recommendation.

A more revealing discussion begins after that recommendation.

Who gets a move because we make ours? What becomes more attractive for them, and what becomes less attractive? If they respond as we hope, what would keep them doing so? If they respond badly, have we left ourselves a way to repair the relationship, change the offer or withdraw without making everything worse?

You may find that the original plan remains the right one. A lower price may genuinely open a valuable market. A hard contractual boundary may be necessary. A partnership may be worth entering despite imperfect alignment. Game theory does not privilege agreement over competition.

It challenges the habit of evaluating an idea as though everyone else will remain still while we carry it out.

In Axelrod’s tournament, the winning programme did not need to defeat every other programme. It did well because of the conduct it made worthwhile for the players it met.

A business decision can look brilliant at the moment it is announced. Its quality becomes clearer when customers, suppliers, partners and competitors have had their turn.