Guests **Kshitij Kulkarni, Matheus V. X. Ferreira and Tarun**

• Matheus
  • Postdoc @Harvard
  • Works at the intersection of economics and cryptography on ideas like credible auctions (esp auctioneer’s incentives), equilibria of staking, EIP-1559, etc
• Kshitij
  • PhD on dynamic incentive or mechanism design
  • Nerdsniped on MEV and GDAs by Tarun
• Tarun
  • Interacted with opening auctions while working in trading, got interested in how well the theory of algorithmic game theory matched with how people actually behaved

Auctions outside of crypto

• Can’t experiment with legacy systems like opening auctions for CME because of entrenched interests of the participants and legal risks.
• Ad auctions - high-dimension but sparse. For example, as an advertiser, you can bid on specific tags like “female”, “lives in London”, “buys expensive shoes”, etc. All these tags have their own auctions.
• MEV is a mix of dense trading auctions and sparse ad auctions

History of auctions in blockchain

• Bitcoin paper proposed first-price auctions for inclusion which are called transaction fees. Off-chain auctions were also settled on-chain in the early days like for the early Pepas NFT collection mints. What are we pricing here? blockspace on bitcoin
• Starting new markets around new tokens in ICOs and more recently NFTs. What are we pricing here? batch of fungible tokens or single non-fungible tokens

Types of auctions

• Private or public. public when auctioneer or bidder performs “open outcry” (or publicing revealing price) or “sealed-bid” where you can only communicate in a private channel between the auctioneer and the bidder
• English auction (ascending price)
  • Seller ask if the bidders want to remain in the auction every round
  • Public-outcry, communicationally expensive
• Sealed auctions
  • Everyone seals and submits their bid directly to the auctioneer.
  • First-price - Highest bidder wins (allocation rule) and has to pay their bidding price (payment rule)
  • Second-price - Highest bidder pays the second bidder’s bid price
• Dutch auction (descending)
  • Flipped version of English - seller offer bidders a price and if no one accepts, lowers it and asks again.
  • Better communication complexity than English (especially in an on-chain environment where communication is crucial)
• Gradual Dutch Auctions (sequential auctions)
  • Sell your items at decaying prices over time. Find the unknown “average” price what people are willing to buy at.
  • Example - paradigm’s Art Gobbler
  • Paper - here
• Revelation principle (how the participants reveal their preferences) has a property that even the more complicated communication can be distilled down to direct revelation or requesting bids from bidders for purpose of analysing auction design.

Credibility and incentives

• Incentives for sellers
  • Do they have an incentive to deviate from what was promised? Can they lie or participate in the auctions?
  • They can set the price curve - what the decay rate is and what is the initial price?
  • If they can participate in the auction, they can make a profit (value - cost) by buying their own supply(cost) with the expectation that bidders will pay a higher price later(value).
• Incentivesfor buyers
  • Do they have an incentive to bid their private valuation (truthfulness)? or can they buy now and sell at a higher price later or vice versa?
  • Can be modelled as ephemeral where she comes in during one round with a “take it or leave it” preference or perpetual where her utility is dependent on price history from the prior sequence of auctions or has a “watch now and buy later” preference
• Collusion can take place through off-chain agreements. Seller and buyer can collude or buyers can collude with each other.
• Secondary markets dramatically increases the complexity for IC. Instead of having a simple preference of just bidding, you can have strategies like collusion, signalling, direct bribes, etc. You’ll need to construct an extended strategy game with joint equilbrium of the auction and the external market. (way harder to reason)
• For sequential auctions
  • You’ll lose some information because people can’t bid on aribtrary sequence of auctions or buy back to a previous item but it’s easier to reason about for the seller.
  • Not studied in a low communication complexity environment

MEV and auctions