Blockchain

Game Theory in Blockchain

In 2016, The DAO on Ethereum raised $150 million - and was exploited. The code was cryptographically flawless: hashes, signatures, consensus - everything worked. The problem wasn't in the cryptography, but in game theory: the voting mechanism allowed a rational attacker to drain funds without violating a single cryptographic rule. Since then, the industry learned a hard lesson: a blockchain is only as secure as its economic incentives are well-designed. Cryptography answers the question of "who can." Game theory answers the question of "who will want to."

  • **EIP-1559 (Ethereum)** - mechanism design fee market: base fee is burned, users are incentivized to bid their true price rather than play a blind auction
  • **Polymarket** - a $1B+ prediction market where Schelling points determine the outcome of events: "Who will win the election?" is resolved by economically-incentivized voting
  • **Flashbots MEV-Boost** - PBS on Ethereum: 90%+ of blocks are built via MEV-Boost, redistributing >$300M MEV per year through mechanism design instead of chaotic extraction

Предварительные знания

  • The Consensus Problem: Why and How
  • Proof of Work: security through energy

Nash Equilibrium

In 1950, John Nash proved a theorem for which he received the Nobel Prize 44 years later: in any finite game there exists at least one state from which **no participant has an incentive to deviate unilaterally**. This state is called a Nash Equilibrium. In blockchain, this idea is the foundation of everything: protocols are designed so that honest behavior is exactly this equilibrium.

0

1

Sign In

Prisoner's Dilemma for Miners

Imagine two large miners controlling a significant share of hashrate. Each faces a strategic choice: **mine honestly** (follow the protocol) or **attack** (double spend, selfish mining, transaction censorship). Let's write out the payoff matrix - the table of payoffs for each combination of strategies.

Why exactly these payoffs? With both miners mining honestly, both receive stable block rewards. If one attacks while the other is honest - the attacker gains little, because they're burning resources on an attack that undermines trust in the network and drives down the BTC price. If both attack - the network loses trust entirely, the price collapses, and both lose their equipment investments.

Why Honest Mining Is a Nash Equilibrium in Bitcoin

The key argument: **a miner with a large hashrate is the network's largest stakeholder**. They've invested millions in ASICs, power contracts, and infrastructure. If an attack crashes the BTC price, that miner loses the most. This isn't altruism - it's pure rational self-interest, aligned with the network's interests.

Mixed Strategies and Security Thresholds

In pure strategies, a player chooses a single action. But Nash Equilibrium also exists for **mixed strategies** - where a player randomly alternates between strategies with certain probabilities. In the Bitcoin context this means: theoretically, a miner could attack with probability p and mine honestly with probability (1-p). Eyal and Sirer showed in 2013 that **selfish mining** becomes profitable at as little as ~33% hashrate, not 51%. This means the security threshold is lower than commonly assumed, and the protocol relies not only on pure mathematics but also on the economic irrationality of an attack.

**A key distinction:** the mathematical feasibility of an attack (from 33% hashrate) and the economic rationality of an attack are different things. Bitcoin is secure not because an attack is impossible, but because it is economically absurd for any sufficiently large participant.

Why is honest mining a Nash Equilibrium in Bitcoin, assuming no single participant controls more than 50% of hashrate?

Mechanism Design

Game theory studies strategic behavior within given rules. **Mechanism design** reverses the problem: given a desired outcome, **design the rules so that rational self-interested actors inevitably arrive at that outcome**. This is "reverse game theory." Leonid Hurwicz, Roger Myerson, and Eric Maskin received the Nobel Prize in Economics in 2007 precisely for mechanism design.

Vickrey Auctions and EIP-1559

A classic example of mechanism design is the **Vickrey auction**: the participant with the highest bid wins, but pays the second-highest price. This is brilliant: participants are incentivized to bid their true valuation, because overbidding doesn't help (you pay the second price) and underbidding risks losing. The mechanism is **incentive-compatible** - honesty is optimal.

EIP-1559 is a textbook example of mechanism design in blockchain. Before it, Ethereum used a first-price auction like Bitcoin: users blindly picked a gas price, overpaying or getting stuck. EIP-1559 split the fee into **base fee** (burned, determined by algorithm) and **priority fee** (tip to the validator). Result: users are incentivized to bid their actual willingness to pay, not play a guessing game.

Token Incentives and Punishment Mechanisms

Mechanism design in blockchain extends far beyond auctions. **Tokenomics** is an entire toolkit of mechanisms: staking rewards incentivize validation, slashing punishes malicious behavior, vesting schedules prevent dumps.

**Why is mechanism design critical for blockchain?** In centralized systems, rules can be changed and enforced. In decentralized protocols there is no arbiter - the mechanism must be **self-enforcing**: participants follow the rules because it's in their interest, not because they're compelled to.

Why is EIP-1559 considered an example of mechanism design, while Bitcoin's first-price auction is not?

Schelling Point

In 1960, economist Thomas Schelling ran an experiment: "You need to meet a stranger in New York. You can't agree on a place or time. Where and when do you show up?" Most people answered: **noon, Grand Central Station**. Without any coordination, people converge on the same point. Schelling called this a **focal point** - a point that feels "natural" to each participant. In blockchain, Schelling points solve a problem that code alone cannot: **how to bring subjective real-world information on-chain**.

SchellingCoin and Blockchain Oracles

In 2014, Vitalik Buterin proposed the concept of **SchellingCoin**: a mechanism where participants vote on a value (e.g., the ETH/USD price) without seeing each other's votes. Those whose votes land within the median receive a reward. Those far from the median lose their stake. The Schelling point here is **the true value**, because every rational participant expects everyone else to also vote for the truth.

UMA and the Optimistic Oracle

The **UMA (Universal Market Access)** protocol developed the Schelling point idea into an **Optimistic Oracle**. The principle: data is assumed to be correct until someone disputes it (optimistic). If a dispute arises, the **Data Verification Mechanism (DVM)** kicks in: all UMA token holders vote on the correct value using a stake-weighted median. The Schelling point is truth, because lying is economically counterproductive.

  • **Polymarket** - a prediction market using UMA for dispute resolution: "Who will win the election?", "Will the ETF be approved?"
  • **Augur v2** - a decentralized prediction market with a reporter system: reporters stake REP on an outcome, the minority loses their stake
  • **Chainlink** - aggregates data from many oracles; if an oracle deviates from the median, it loses reputation and future revenue

Coordination Without Communication

Why do Schelling points work at all? Because **truth is a natural focal point**. When each participant knows that everyone else also wants to land near the median, and the median most likely aligns with reality, the rational strategy is to vote for the truth. This is self-reinforcing: the more participants vote honestly, the more reliable the median, the more profitable it is to vote honestly.

**Attacks on Schelling points:** with a low number of participants or a highly concentrated stake, a coordinated attack is possible - a colluding group votes for a false value and shifts the median. Defenses: high **quorum** (minimum number of voters), **escalation mechanisms** (UMA allows disputing results at a higher level), and **reputation systems**.

Why do SchellingCoin participants try to vote for the true price of an asset rather than an arbitrary value?

Incentive Compatibility

A protocol is **incentive-compatible** if every participant's best move is to be honest - to reveal true preferences, follow the rules, and not try to game the system. This is the key property that unifies Nash Equilibrium, mechanism design, and Schelling points into a single architectural idea: **the rules of the game are such that each participant's rational self-interest aligns with what's good for everyone**.

Revelation Principle

The **Revelation Principle** (Myerson, 1981) states: for any mechanism where participants play complex strategies, there exists an equivalent **direct mechanism** in which the optimal strategy is simply to tell the truth. This is a powerful tool for the designer: no need to anticipate clever participant strategies - just build a mechanism where truth is more profitable than lying.

MEV as a Violation of Incentive Compatibility

**MEV (Maximal Extractable Value)** is the profit a validator can extract by manipulating the order of transactions in a block. Front-running, sandwich attacks, liquidation sniping - all of these are examples of broken incentive compatibility: **the validator is better off reordering transactions than following honest FIFO order**.

Solutions: PBS, Fair Ordering, MEV Burn

How to restore incentive compatibility? The community is developing several approaches:

  • **PBS (Proposer-Builder Separation)** - role separation: the proposer (validator) proposes a block, the builder assembles transactions. Builders compete for the right to build the block, passing the profit to the proposer. MEV is redistributed but doesn't disappear
  • **Fair Ordering (Chainlink FSS, Themis)** - protocols guaranteeing transactions are processed in arrival order (FIFO). Encrypting transaction contents before block inclusion (commit-reveal) makes front-running impossible
  • **MEV Burn (EIP-7762)** - burning MEV profit instead of passing it to the validator. Analogous to EIP-1559 for MEV: if extracted value is burned, the incentive to manipulate drops
  • **MEV Share (Flashbots)** - redistributing MEV profit back to users whose transactions were used in the extraction. Partial restoration of incentive compatibility

**Impossibility result:** the Myerson-Satterthwaite theorem proves that an ideal mechanism (simultaneously strategy-proof, individually rational, and budget-balanced) is impossible for bilateral trade. This means **any blockchain protocol contains trade-offs**. PBS sacrifices decentralization (builder centralization). Fair ordering sacrifices throughput. MEV burn sacrifices validator revenue. Choosing the trade-off is the engineering art of mechanism design.

Blockchain protocols work because participants are altruistic and want to support the network for the common good

Blockchain protocols work because they are incentive-compatible: the rules are designed so that each participant's rational self-interest (maximizing their own profit) aligns with behavior that benefits the network - without requiring trust or altruism

This misconception is dangerous because it implies that having "good people" is sufficient for network security. In practice, protocols must be robust to the worst case: every participant is rational, selfish, and ready to cheat if it's profitable. Mechanism design ensures that cheating is unprofitable. When this principle breaks down (as with MEV), the protocol is vulnerable - regardless of participants' intentions.

What makes MEV a fundamental incentive compatibility problem rather than just a technical bug?

Key Takeaways

  • **Nash Equilibrium** - a state from which no participant benefits from deviating unilaterally; honest mining in Bitcoin is an example of Nash Equilibrium, because an attack devalues the attacker's own assets
  • **Mechanism design** - "reverse game theory": designing rules so that the desired behavior becomes the equilibrium; EIP-1559 is the textbook example of mechanism design in blockchain
  • **Schelling points** - focal points that allow participants to coordinate without communication; the foundation of blockchain oracles (UMA, SchellingCoin), through which subjective data enters on-chain
  • **Incentive compatibility** - the foundation of protocol security: rules must make honesty the optimal strategy; MEV is an example of violating this principle, and exactly such violations were warned against by The DAO story in our introduction - when economic incentives diverge from developer intentions, no cryptography will save you

Related Topics

Game theory permeates every layer of the blockchain stack - from consensus to financial protocols:

  • Bitcoin Economics — Nash Equilibrium and incentive compatibility directly determine Bitcoin's security: miner payoff matrices, fee market design, and security budget are all applications of game theory
  • Proof of Stake — Slashing, correlated penalties, and validator rewards are mechanism design for PoS, turning stake into skin in the game and making honest validation a Nash Equilibrium
  • MEV — Maximal Extractable Value is the leading example of broken incentive compatibility in modern blockchains; PBS, fair ordering, and MEV burn are mechanism design solutions
  • Tokenomics — Token incentive design is a direct application of mechanism design: vesting, staking rewards, and burn mechanics are designed to create incentive-compatible systems

Вопросы для размышления

  • The DAO was exploited not through cryptography, but through game theory. Can you imagine a protocol that is mathematically provably incentive-compatible under all conditions? Or does the Myerson-Satterthwaite theorem make that impossible?
  • MEV redistributes $300M+ per year from users to validators/builders. If PBS legalizes MEV (making it transparent and structured), is that a solution to the problem or a normalization of it?
  • Schelling points work when participants coordinate around truth. But what if "truth" is subjective (e.g., "Is this a work of art?")? Can Schelling point mechanisms be used for subjective assessments?

Связанные уроки

  • gt-01
Game Theory in Blockchain