In prior posts, we’ve talked about the fundamentals of auctions and how they can be gamed or distorted by external factors. In this post, we’ll take a deeper look at how auctions are designed, and some of the ways that design choices directly shape outcomes.

Those past posts have focused on straightforward auction designs — English auctions (ascending bid), Dutch auctions (descending bid), and sealed (hidden bids) — which are the primordial templates for the much more sophisticated ones that have emerged over the past 20 years. The increasing complexity and power of auctions has been in no small part due to their incorporation into technology platforms: eBay made the use of auctions to buy and sell goods in a peer-to-peer market mainstream. Google, Bing and other search engines use auctions to sell advertising space in real-time, automated processes. And — although big-ticket NFT auctions have been dominating recent headlines — auctions have always been a part of blockchain on a much deeper level, with auctions incorporated into the fundamental protocols of crypto networks like Bitcoin and Ethereum to determine the priority order by which transactions are processed.

From simple to sophisticated

As we’ve discussed earlier, fundamentally, auctions have two basic purposes:

• They’re used to determine which buyer or buyers will receive the goods or services being sold in the auction
• They’re used to set the price paid by buyers to the sellers — particularly valuable in situations where valuations are uncertain and there is no standard pricing (because, for example, the item being sold is unique, new to offer or subject to volatile marketplace conditions)

While these are simple outcomes, the paths to arrive at them can be deceptively multidimensional.

Multiplying the challenge

Both real world and online auctions often have to contend with the fact that auctions aren’t always for single items — they often are held for many interchangeable units of the same good — and in most cases, auctions are not one-time-only events. Multiplying the number of items being auctioned and the frequency of auction transactions stretches the design challenge in significant ways.

Take as example the mechanism used to determine transaction costs on Ethereum, generally referred to as “gas.” Each user submitting transactions to be processed on the network also includes a bid reflecting her proposed payment for the service of processing. Block producers order incoming transactions, generally from the highest to lowest bid per available unit in their block, and fill their blocks in that order. Users whose transactions are selected for processing pay their proposed bid.

Based on what we’ve described in earlier posts, this is a first-price auction — which is to say, winning bidders pay what they submitted as their final valuation for the service of processing. But there are other variables here as well: Firstly, this a sealed bid auction — each bidder is submitting her bid with limited information on what the other participants are going to bid for that particular block. (While historical records of previous successful and unsuccessful bids may be public information, they may have limited applicability to the current block.) And secondly, this is a multi-unit sale, which is to say that there are many effectively interchangeable “instances” of the item being auctioned, in this case, processing time and inclusion in a block of processed transactions.

One key factor in multi-unit auctions is whether the ultimate price paid by winning bidders is the same (what’s known as a “uniform price auction,” in which the bids of buyers are ranked in order of bid price and their orders fulfilled at a common price, usually the lowest bid that clears all units of the auctioned item) or different (what’s known as a “discriminatory price auction,” in which buyers are ranked and awarded their requested units in order of bid price, paying the price they actually bid). Ethereum gas auctions are discriminatory.

Another relates to how units are distributed: Some auctions may apply quotas to how many units are assigned to each winner proportionately based on bid price, or may fulfill orders completely for higher bidders until they run out, or use other formulas to determine distribution to winners. Ethereum gas auctions usually fulfill orders for higher bidders first, although (as we’ve written about before) miners ultimately have discretion to fulfill orders however they choose, creating the potential for abuse.

There’s also another factor to consider in blockchain network auctions: Like many other real-world auctions, they’re sequential, which is to say, they take place on a recurring basis, often with pools that contain at least some of the same participants (on both the “sell” side, that is to say, block processors, and the “buy” side, that is to say, individuals seeking to have transactions processed).

Sequential auctions offer design challenges because each auction is not separate — because outcomes of one auction can impact buyer and seller decisions in subsequent auctions, some buyers may apply strategies that are dependent on sequences of transactions occurring in tandem, and others may use strategies that employ gamification (e.g., sabotaging competitors in one auction in order to win in a subsequent one). In fact, as we’ve discussed in prior posts, some of the vulnerabilities of EPNs to abuse relate directly to gas auctions and their sequential nature. And as we’ve also noted, in both the physical world and in blockchain, over the long term, sequential auctions can incentivize collusion among bidders or sellers, or both.

Optimizing auctions under these circumstances requires taking into account all of these factors and more — and each type of auction design has its own benefits and downsides. There’s rarely a single “best” or “worst” auction design, but rather, designs that are beneficial to different, and often mutually exclusive, sets of goals.

How the Ethereum Network is tweaking its gas auction design to realign incentives

In a first-price, sealed-bid, multi-unit auction setting such as the one behind Ethereum’s gas fee mechanism, the first transactions to be included in a block will almost inevitably end up paying more per unit of block space than the last transaction to be included. On the one hand, this allows users who highly value quick transaction processing to pay more, to ensure they’re put at the front of the line. On the other hand, this also can lead to users feeling the equivalent of the winner’s curse, and complaining that they’re paying “excess” fees that simply pad the pockets of miners.

These are some of the factors that are currently being debated by the proposed upgrade EIP 1559, due for review this summer. In particular, EIP 1559 notes that the current system doesn’t truly reflect the actual costs of network congestion, with gas fees sometimes outweighing the true network costs incurred by a factor of 10 or more times; produces needless delays in transaction processing because blocks are fixed in size; leads to overpayments due to the inefficiencies of fee estimation algorithms; and might create potential instability in the long term due to how it encourages selfish competition among miners. The EIP proposes an approach that establishes a base fee for processing that’s automatically adjusted by the protocol based on network congestion in a constrained and relatively stable fashion. While this fee would be manually adjustable, in most cases, the fee would be automatically set by user wallets in a fashion that’s both reliable and predictable. The base fee would be “burned” — e.g., it would be erased from the system rather than going into the pockets of the miner; miners would only keep the “priority fee” that is set on top of the base fee by those who want to uprank their transactions.

Among other things, this proposal helps to reduce the risks associated with miner extractable value, which we’ve written about in prior posts, while also decreasing miner incentive to manipulate fees in ways that are exploitative of users.

Beyond the bottom line

While revenue maximization is an important goal in any auction, designers may need to prioritize other objectives, such as simplicity and efficient allocation of resources. There’s no better illustration of this than the original FCC spectrum auction — what William Safire called “the greatest auction in history.” We’ve written about this auction in prior posts, but here’s a recap.

In 1993, the United States Congress passed a law allowing the Federal Communication Commission (FCC) to auction licenses for the Broadband Personal Communications Service (PCS) spectrum. For most of its history, the FCC had allocated spectrum licenses through hearings where would-be license owners would individually pitch their claims. As the cost and time required for these hearings increased, the FCC tried to allocate spectrum via simple lottery, which was a disaster. In theory, a well-designed auction could allocate licenses to the users that valued them most, in a timeframe of months rather than years.

The FCC turned to economists with expertise in auction design, Paul

Milgrom and Robert Wilson, to advise on both the design of the auction and optimal bidding strategies. They started by outlining four goals: The auction design should be simple and easy for first-time users to participate in. It should raise significant revenue for the government, and to fulfill diversity goals prescribed by the FCC. Finally, the outcome of the auction should be efficient: the users with the highest values for the licenses should obtain them.

For spectrum, unlike art or other common auctioned items, finding the efficient allocation of bandwidths and geographies to users was a complex combinatorial problem whose solution couldn’t be determined ahead of the auction.

By defining these goals, Milgrom and Wilson were able to narrow down the range of potential design options from an almost unlimited number of variations: Would the auctions run simultaneously or sequentially? Would bids be ascending or sealed? Would package bids (what’s known as a “combinatorial” auction) be allowed? Should bidder identities be concealed or revealed?

Through a mix of observation, experimentation, and finally, real-world pilot tests, they were able to develop a design that achieved the outcomes closest to their goals. The mechanism that the team designed, now known as the FCC auction, has since been used to allocate over $100 billion of spectrum worldwide — and the main economists behind the auction design won the Nobel Prize for their work..

The approaches used by economists to fine-tune this auction, and others that have been implemented by eBay, Amazon and Microsoft, among many others, can be applied to almost any setting — from blockchain to video games to estate sales. While in some of these settings, sophisticated auctions may be necessary, a basic understanding of auction theory can inform and improve design for a vast range of potential applications.

As we’ve noted, this is particularly the case in markets characterized by unique goods, idiosyncratic valuations, and limited comparable transaction histories. In our next post, exploring auctions for virtual real estate in games, we’ll show how auctions perform in a circumstance with some unique characteristics and extremely challenging dynamics.

Coinsmart. Beste Bitcoin-Börse in Europa
Source: https://medium.com/community-economics-by-forte/designing-auctions-theory-and-practice-1218c8427325?source=rss——-8—————–cryptocurrency