Weighted fictitious play

Weighted fictitious play is a learning process in Game Theory where you predict opponents by tracking past actions, but you give more weight to recent moves. It models boundedly rational players who learn and adapt over time.

What is weighted fictitious play?

Weighted fictitious play is a Game Theory learning model where each player forms beliefs about what opponents will do by watching their past actions, then puts extra weight on the most recent observations. Instead of treating every old move as equally informative, the model says newer behavior should matter more because strategies can change.

It starts from standard fictitious play. In plain fictitious play, you estimate an opponent’s strategy from the frequency of what they have done before, then respond with your best reply to that estimate. Weighted fictitious play changes the update rule, so the belief is not just a raw count of past moves. Recent choices might count more than earlier ones, which makes the learning process more responsive.

That difference matters when the game is not perfectly stable. If an opponent experiments, adapts, or switches tactics, equal-weight memory can be too slow. Weighted fictitious play gives the model a short memory that still uses history, but with a tilt toward what has been happening lately. In effect, it captures the idea that players do not have perfect foresight, they infer patterns from limited experience.

This is a bounded rationality model, so it fits situations where people are trying to do reasonably well without solving the whole game from scratch every round. You can think of a repeated pricing interaction, an arms race in strategy, or any repeated choice setting where each side watches the other and updates.

The weights can be chosen in different ways. Sometimes they decay over time, so older actions fade out. Other versions let the analyst pick the weighting scheme to reflect how quickly players think opponents adapt. Under certain conditions, weighted fictitious play can converge to a Nash equilibrium, but it does not promise that in every game or every setup. The main point is the learning rule, not a guarantee of one final outcome.

Why weighted fictitious play matters in Game Theory

Weighted fictitious play matters because it shows how strategic behavior can evolve when players do not know the other side’s plan in advance. A lot of Game Theory is about equilibrium as a static endpoint, but many real games are repeated, noisy, and full of adjustment. This model bridges that gap by showing one way players might get there, or miss it, through learning.

It also gives you a cleaner way to talk about recency bias in strategy. If a player only remembers long-run averages, they can miss a genuine shift in behavior. If they overreact to the last move, they may chase noise. Weighted fictitious play sits between those extremes, which makes it useful for describing more realistic decision-making.

In class, it often comes up when comparing rationality assumptions. Standard best response assumes you know what others are likely to do. Bounded rationality says you may need a rule for learning that is simpler than full optimization. Weighted fictitious play is one of the clearest examples of that kind of rule because you can trace exactly how beliefs change after each round.

It also helps explain when a strategic interaction settles down and when it keeps shifting. If the environment is stable, the model may drift toward a stable pattern. If the environment keeps changing, the weighting keeps the player responsive. That makes it a strong tool for interpreting repeated games, classroom simulations, and problem-set examples where actions unfold over time.

How weighted fictitious play connects across the course

Fictitious Play

Weighted fictitious play is a modified version of fictitious play. In standard fictitious play, each past action is usually counted the same way when forming beliefs about an opponent. The weighted version changes the update rule so recent actions matter more, which makes the learning process faster to react to changing strategies.

Bounded Rationality

This term sits inside bounded rationality because it models limited memory, limited information, and rule-based learning. Instead of assuming a player solves the whole game perfectly, weighted fictitious play shows how someone can update from experience and still make decent strategic choices. It is a realistic shortcut, not a fully omniscient solution.

Nash Equilibrium

Weighted fictitious play is often discussed with Nash equilibrium because the learning process may converge to one under certain conditions. The connection is not that the two ideas are the same, but that one describes a possible stable outcome and the other describes a process that can lead there. In repeated games, that process is often what you analyze.

quantal response equilibrium

Both weighted fictitious play and quantal response equilibrium relax the idea that players always pick a perfect best response. They handle imperfect decision-making in different ways, though. Weighted fictitious play focuses on how beliefs change over time, while quantal response equilibrium focuses more on probabilistic choice when players are making errors or tradeoffs.

Is weighted fictitious play on the Game Theory exam?

A quiz or problem-set question might give you a repeated game and ask how a player updates beliefs after seeing a sequence of opponent moves. Your job is to show that weighted fictitious play uses historical frequencies, but gives more influence to recent actions. If the prompt asks why the model is more realistic than standard fictitious play, mention bounded rationality and changing behavior.

You may also be asked to compare learning rules. In that case, explain whether the model reacts slowly or quickly to new information, and whether the situation looks stable enough for convergence. If a homework problem gives weights, use them to build the belief estimate step by step, then describe the best reply based on that weighted belief rather than on raw counts alone.

Weighted fictitious play vs Fictitious Play

These are easy to mix up because both update beliefs from past opponent moves. The difference is that weighted fictitious play does not treat every past move equally. It emphasizes recent observations, so it reacts faster when the opponent’s strategy changes.

Key things to remember about weighted fictitious play

• Weighted fictitious play is a learning rule in Game Theory where players predict opponents by giving more weight to recent actions.

• It is a version of fictitious play, but the belief update is not a simple equal-count average of the past.

• The model fits bounded rationality because it describes players who learn from experience instead of solving the whole game perfectly.

• It is useful in repeated or changing strategic settings, especially when opponents may shift strategies over time.

• Under some conditions, the learning process can converge to Nash equilibrium, but that is a result, not a guarantee.

Frequently asked questions about weighted fictitious play

What is weighted fictitious play in Game Theory?

Weighted fictitious play is a learning process where a player estimates an opponent’s strategy from past actions, but recent moves count more than older ones. It is used to model repeated strategic interaction when players adjust based on what they have seen lately. That makes it a more flexible version of standard fictitious play.

How is weighted fictitious play different from fictitious play?

In fictitious play, past actions are usually treated as equally informative when you form beliefs. In weighted fictitious play, the belief update gives extra importance to recent observations. That means the player changes course more quickly when the opponent starts behaving differently.

Why does weighted fictitious play count as bounded rationality?

It assumes players do not solve the entire game with perfect information and unlimited computation. Instead, they use a simple learning rule based on experience. That makes it a classic bounded rationality model because it captures a realistic shortcut for strategic decision-making.

Can weighted fictitious play reach Nash equilibrium?

Yes, in some games and under certain conditions, the learning process can converge to a Nash equilibrium. But that depends on the game structure and the weighting scheme. The main idea is that repeated updating can lead to stable behavior, not that it always will.