7QQMM200 Chap.7 Backward Induction and Credible Play
Backward Induction and Credible Play
Solve continuations first
Backward induction selects optimal actions at the final decision nodes, carries their payoffs to earlier nodes and repeats until the root is solved. The method relies on perfect information at the decisions being optimised.
Credibility lives off path
A Nash profile can depend on an action that would be inferior if its information set were reached.
Sequential rationality asks for a best response at every information set. Complete strategies therefore matter even when two profiles generate the same observed outcome.
Work from terminal decisions toward the root
At each final decision node, compare the payoff component belonging to the mover. Mark every best response and replace that decision with the induced terminal payoff.
Earlier players then compare continuation values created by rational later behaviour. This order prevents the root player from choosing against a response that the later player would never make.
Report strategies as well as outcomes
The selected branches identify one realised path, but the equilibrium profile must prescribe actions at every information set.
Two profiles can reach the same terminal node while differing off path. Those differences affect the beliefs and incentives supporting the root action, so a terminal payoff pair alone is not a complete equilibrium answer.
Test threats at the node where they operate
Assume the threatened information set has been reached and compare the mover's actions there.
If the prescribed response gives a lower payoff, it is not credible. Correct the local continuation before returning to the earlier decision.
This diagnostic exposes Nash equilibria that rely on behaviour a rational player would reject when actually called to move.
Respect the information boundary
Simple backward induction cannot optimise separately across nodes in one non-singleton information set because the mover cannot identify the current history. Beliefs or a continuation equilibrium are then needed.
Do not sever the information link merely to force a solution; that changes the game. When later payoffs tie, retain all best responses and acknowledge possible multiple induction solutions.
Carry continuation values without changing payoff order
At a final node controlled by player B, compare the second component of each payoff tuple when the convention is A first and B second.
Circle all maximising branches, then copy the entire resulting tuple back to the predecessor. At player A's decision, compare the first components of those continuation tuples.
This mechanical discipline prevents a common error: choosing the branch with the largest first number at every node regardless of who moves.
Handle ties as branches in the solution set
If a later mover is indifferent, backward induction does not justify discarding one best response. Carry each tied continuation backward because it can change the earlier player's comparison.
The game may consequently support several complete strategy profiles or outcomes. Present the alternatives explicitly and distinguish multiplicity caused by a genuine payoff tie from multiplicity created by leaving an off-path instruction unspecified. A complete profile must still state the action at every information set.
What this chapter covers
- 01
Continuation payoff
- 02
Sequential rationality
- 03
Backward induction
- 04
Equilibrium path
- 05
Off-path action
- 06
Credible threat
Worked example · free
Trace a chosen continuation path
- 3Choose the responder's larger second payoff at each node.
- 2Carry the selected pairs to the root.
- 3Choose the root branch using the first payoff.
Key terms
- Backward Induction
- Backward Induction solves a finite tree from later choices toward the root.
- Equilibrium Path
- The Equilibrium Path is the history reached with positive probability under the equilibrium strategies.
- Credible Threat
- A Credible Threat prescribes an action that remains optimal if its decision point is reached.
Backward Induction and Credible Play FAQ
What is the order of a backward-induction solution?
Start at the latest decision nodes, select each mover's best action, replace the node with its continuation payoff and work carefully toward the initial choice.
Why can two strategies produce one outcome?
They may prescribe the same on-path actions but different responses at information sets that are never reached. Those off-path differences matter for credibility and equilibrium refinement.
When does simple backward induction fail?
It cannot optimise a mover separately across nodes that belong to one non-singleton information set. The player cannot identify the node and needs beliefs about which history occurred.
Exam move
Cover the root of a game tree and solve only the final decisions. Reveal one stage at a time, recording whose payoff is maximised. Then write the full strategies and mark every off-path instruction before naming the outcome.
Working through Backward Induction and Credible Play in 7QQMM200? Sia is AskSia’s AI Economics tutor — ask any 7QQMM200 Backward Induction and Credible Play question and get a clear, step-by-step explanation grounded in how 7QQMM200 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.