game-theory¶
modelingprivate (curator-owned)formal-modelingCurator-private skill — copy text from 100xOS/shared/skills/modeling/game-theory.md.
Game Theory for Platform Economics & IS Research¶
Scope¶
This skill covers game theory as used in IS and platform economics research: strategic interaction in platform markets, mechanism design for digital systems, auction theory, and computational methods for verifying equilibria.
Core Concepts¶
Normal Form Game¶
- Players N, strategy sets S_i, payoff functions u_i(s_1, ..., s_n)
- Nash equilibrium: no player can unilaterally improve payoff
- Mixed strategy NE: players randomize; indifference across support
Solution Concepts (strongest to weakest)¶
- Dominant strategy equilibrium
- Iterated elimination of dominated strategies
- Nash equilibrium (pure or mixed)
- Subgame perfect equilibrium (sequential games)
- Perfect Bayesian equilibrium (incomplete information)
Platform-Relevant Models¶
Two-Sided Markets¶
- Platform sets fees (f_b, f_s) to buyers and sellers
- Network effects: utility increases with participation on the other side
- Key trade-off: subsidize one side to attract the other (Rochet-Tirole 2003)
- Equilibrium: solve for participation thresholds on each side given fees
Platform Competition¶
- Cournot/Bertrand adapted for platforms: compete on fees, features, or quality
- Multi-homing vs single-homing affects competitive dynamics
- Winner-take-all vs market sharing depends on differentiation and multi-homing
Token Mechanism Design¶
- Token as coordination device: participation thresholds, staking equilibria
- ICO/token sale as mechanism: price discovery, adverse selection
- Governance tokens: voting games, delegation, whale capture
Mechanism Design¶
- Social choice function f: type profiles -> outcomes
- Revelation principle: any implementable outcome achievable by direct truthful mechanism
- IC (incentive compatibility): truth-telling is equilibrium
- DSIC: dominant strategy IC (strongest)
- BIC: Bayesian IC
- IR (individual rationality): participation constraint
- VCG mechanism: DSIC for efficient allocation; each agent pays externality imposed on others
Key Impossibility Results¶
- Gibbard-Satterthwaite: with 3+ alternatives, only DSIC+onto mechanism is dictatorship
- Myerson-Satterthwaite: no efficient+IC+IR+budget-balanced bilateral trade
Auction Theory¶
Standard Formats¶
- First-price sealed-bid: b(v) = v - integral [F(t)/F(v)]^(n-1) dt
- Second-price (Vickrey): b(v) = v (dominant strategy)
- All-pay: b(v) = integral t(n-1)F(t)^(n-2)*f(t) dt
- Revenue equivalence: same expected revenue across standard formats (IPV)
Relevance for IS¶
- NFT auctions, DeFi liquidation auctions, ad auctions
- Combinatorial auctions for spectrum/cloud resources
- Dynamic pricing as mechanism
Computational Methods¶
Nash Equilibrium Computation¶
Support enumeration (2-player, small games): Enumerate support pairs, solve indifference conditions, verify no profitable deviations. Complexity O(2^(n+m)) -- only for small games.
import nashpy
game = nashpy.Game(A, B) # payoff matrices
equilibria = list(game.support_enumeration())
Fictitious play: Converges for 2x2, zero-sum, potential games, strategic complements. Does NOT converge for all games.
Backward induction: Recursive solution for finite extensive-form games with perfect information.
Verification Checklist¶
For any computed equilibrium, verify: 1. Best response: no unilateral deviation improves payoff 2. Probability constraints: sigma >= 0, sum = 1 3. Support condition: strategies in support yield equal expected payoffs 4. Indifference: strategies outside support yield weakly lower payoffs
Common Pitfalls¶
- Multiple equilibria: always search for all NE
- Numerical precision: use tolerance (1e-8) for equality checks
- Mixed strategies: ensure probabilities sum to 1
- Dynamic games: verify subgame perfection in ALL subgames