📌 “A rational player looks ahead and reasons back.” Backward induction and subgame perfect equilibrium are the twin pillars of solving sequential games in game theory. This article breaks down how they work and when to use each.

In game theory, when players make decisions one after another, we call it a sequential game. To find rational outcomes, we often use two main tools: backward induction and subgame perfect equilibrium (SPE). Both aim to eliminate non-credible threats by thinking from the end of the game backwards to the start.

What is Backward Induction?

Backward induction is a step-by-step method. You start at the last decision point of the game, figure out what the player there would choose, then work backwards to the first move. This process "solves" the game by removing moves that would not be chosen rationally later on.

Example 1 The Simple Take-or-Leave Game

Game: Player A moves first, choosing between Left (payoff 2 for A, 2 for B) or Right. If A chooses Right, Player B moves next, choosing between Up (payoff 3 for A, 1 for B) or Down (payoff 1 for A, 3 for B).

🔍 Explanation: We solve backwards. At B's turn (if A chose Right), B gets 1 from choosing Up and 3 from choosing Down. A rational B picks Down. Knowing this, A now compares: Left gives A 2. Right leads to B choosing Down, which gives A only 1. So A chooses Left. The backward induction outcome is (Left).
Example 2 The Centipede Game (Short Version)

Game: Two players alternate. Each turn, a player can Take the pot (ending the game) or Pass (doubling the pot and letting the other player move). The pot starts at $1. Player 1 moves first.

🔍 Explanation: Work from the last possible turn. If the game reaches the final turn, the player there will always Take the pot (it's the highest payoff). Knowing this, the player before them will also Take, and so on. Backward induction predicts Player 1 will Take on the very first move, ending the game with $1, even though cooperation could yield more.

What is Subgame Perfect Equilibrium (SPE)?

A Subgame Perfect Equilibrium is a stronger concept. It is a strategy profile (a plan for each player) that forms a Nash Equilibrium not just for the whole game, but for every subgame within it. A subgame is any part of the game tree that starts from a single decision node and includes everything that follows. SPE ensures players' strategies are rational at every possible point in the game, not just on the predicted path.

Example 1 The Ultimatum Game

Game: Player 1 proposes how to split $10 (e.g., "I get $7, you get $3"). Player 2 can then Accept (the split happens) or Reject (both get $0).

🔍 Explanation: Consider a strategy where Player 1 offers ($9 for self, $1 for other) and Player 2's strategy is "Reject any offer where I get less than $2." This is a Nash Equilibrium for the whole game (Player 1 wouldn't deviate, Player 2 is playing a best response). But is it SPE? No. Look at the subgame where Player 2 faces the ($9,$1) offer. Rejecting gives $0, Accepting gives $1. A rational Player 2 in that subgame would Accept. Therefore, the "reject low offers" threat is not credible. The SPE requires Player 2 to Accept any positive offer, leading Player 1 to offer the smallest positive amount.
Example 2 Entry Deterrence Game

Game: An Entrant decides to Enter or Stay Out of a market. If the Entrant Enters, the Incumbent can then Fight (costly for both) or Accommodate (share the market).

🔍 Explanation: One Nash Equilibrium is: Entrant Stays Out; Incumbent's plan is "Fight if Entry happens." The Entrant stays out to avoid a fight. But is this SPE? Check the subgame after Entry. In that subgame, Fighting gives the Incumbent a low payoff, Accommodating gives a higher one. A rational Incumbent would Accommodate. So the threat to Fight is not credible. Backward induction (and SPE) predicts the outcome: Entrant Enters, Incumbent Accommodates.

Key Differences and Relationship

While closely related, backward induction and SPE are not identical. Backward induction is a solution method. SPE is a solution concept (a type of equilibrium). You often use backward induction to find SPEs in games of perfect information.

Backward Induction vs. Subgame Perfect Equilibrium
AspectBackward InductionSubgame Perfect Equilibrium
NatureA step-by-step solving algorithm.A refinement of Nash Equilibrium.
FocusFinds one rational path by pruning irrational moves.Requires rationality in every part (subgame) of the game.
ApplicationBest for finite games with perfect information (like chess).Applies to any sequential game, including those with imperfect information.
ResultProduces a specific outcome or path of play.Produces a complete strategy profile for all players.
RelationshipIn finite games of perfect information, the outcome found by backward induction is always part of a (and often the unique) Subgame Perfect Equilibrium.

⚠️ Common Pitfalls and Clarifications

  • Not All Nash Equilibria are SPE: Many Nash Equilibria rely on non-credible threats off the equilibrium path. SPE filters these out. The Ultimatum Game example shows a Nash Equilibrium that is not SPE.
  • Backward Induction Requires Perfect Information: You can't apply standard backward induction if players don't know previous moves (imperfect information). SPE can still be defined but is harder to find.
  • SPE is Stronger: Every Subgame Perfect Equilibrium is a Nash Equilibrium, but not every Nash Equilibrium is Subgame Perfect. SPE is a subset of Nash Equilibria.

Conclusion

Backward induction and subgame perfect equilibrium are essential for analyzing strategic interactions where timing matters. Backward induction is the practical tool you use to solve a game by reasoning from the end to the beginning. Subgame perfect equilibrium is the theoretical standard that ensures every part of a player's plan is rational, eliminating empty threats. For most practical problems in finite sequential games, applying backward induction will lead you directly to a subgame perfect equilibrium. Mastering both concepts allows you to predict outcomes in negotiations, business competition, and any scenario where players move in sequence.