Research Report | Algorand Auction and Transaction Deduction Analysis

This article produced by the Institute of Fire currency block chain, this report Published June 22, 2019, Author: Yuan Yuming, Hu Zhiwei, Weng Yi-ming

Summary

Algorand and its Dutch auctions have received a lot of attention recently. The auction results not only enabled the team to receive $60 million in investment funds, but also allowed early-stage private equity investors to see many times the price increase.

Algorand uses the PoS mechanism, but no additional penalties are required because it ensures that the consensus is valid through VRF. The process of consensus is divided into two major stages. The first stage randomly determines each round of voting users in the form of encrypted lottery; the second stage runs a binary Byzantine agreement to determine the final consensus output block. Algorand includes two types of nodes in the system design: the relay node and the non-relays node.

Algorand is a Dutch auction that will be returned to the project at a price of up to 90% after one year. It should also be noted that this auction is not a pure algo, but an algo + put option.

The reasons for designing the auction mechanism are as follows: (1) the team believes that the price should be discovered by the market; (2) it can make as many bidders as possible to make the algo, making the PoS network more secure (3) making the participants have a well-known The common "psychological reserve price" manages the expectations of traders in a transparent manner.

There is no fixed optimal bidding strategy for each participant. Because the Dutch auction has a Nash equilibrium solution: it is Bayesian equilibrium, so the bid for each participant must consider the distribution of bids from other participants.

From the point of view of the secondary market transaction price after the completion of the first auction, algo has actually completed a credit-filling market after the auction, and may have caused a partial premium due to reasons such as thresholds.

There are many trading strategies for bidders, for example, they can sell immediately after the auction and continue to participate in the auction and the secondary market linkage transaction; or sell and fall and re-buy for arbitrage even when the conditions permit. The account option is sold for profit.

In addition, there are some situations that may lead to losses, including the auction bid is too early or too high, the bidder is difficult to avoid losses; in the down market, if the bidder's arbitrage funds are not enough to support the price, it will cause the price to fall and cause losses. .

Report body

Event review

In addition to the white paper on the Libra project in the blockchain world this week, the most popular focus may be Algorand's auction and main online line.

The reason why the Algorand project has received much attention is due to the strength of its team, led by Turing Award winner Silvio Micali, and its uniqueness in terms of technology, especially cryptography. On the other hand, it is online. The eve of the Dutch auction mechanism has also made many users who are used to participating in fundraising at a fixed price feel more novel.

Algorand designed an auction plan with a time span of five years: two auctions per month, with an annual auction target of 50 million Algo. On June 19th, the first auction of the project started with an auction of up to $10, gradually reducing the price, and finally completed the first auction of 25 million Algo for $2.40 after about three and a half hours.

The result of this auction not only enabled the team to receive $60 million in investment funds, but also made early-stage private equity investors see many times the price increase. So, how does Algorand's auction mechanism, especially after the start of the transaction? Here we start with the Algorand project itself, introduce the Dutch auction process and analyze the future trading situation.

2. Introduction to Algorand

2.1. Pure PoS

Similar to many current popular public chains, Algorand also uses the PoS mechanism. However, many PoS currently need to be pledged at the verification node for a period of time to obtain revenue, namely Bonded Proof of Stake. Algorand has a slight difference in that it uses "Pure PoS" (hereinafter referred to as "PPoS"): the user can decide whether to take or not at any time; when determining the stake, the corresponding token does not need to be locked for a while.

At the same time, the two characteristics of the past PoS consensus, liveness and security , often need to guarantee a certain online time (uptime) by the node and detect whether there are multiple signatures to verify the same block. When the discovery does not meet the requirements, the corresponding node Or the user will be punished (Slash) to keep the entire system up and running.

Instead of using a penalty mechanism, Algorand uses cryptography to ensure that as long as more than two-thirds of the token's holders are honest, the consensus process can be effective.

2.2. VRF

One of the key issues in consensus design is how to choose a fair or open way to choose a blocker or verifier. A verifiable random function (VRF) can be considered to solve this problem: a node that is an outlier or verifier can be selected in a fair and open manner. Therefore, in the public chain platform, the combination with the Nakamoto Satoshi consensus can accommodate many participants and avoid excessive concentration.

Algorand mainly uses technologies such as VRF and divides the process of consensus into two major phases:

First, Algorand uses VRF to randomly determine each round of voting users in the form of an encrypted lottery. In addition, Algorand proposes and uses a new Byzantine consensus protocol BA★ to share the consensus in a weighted manner. Is the BFT class consensus + PoS or PoWeight architecture.

The main steps of the consensus include:

1. Determine participants based on VRF lottery. The randomness of the lottery process can be proven and unpredictable to ensure fairness and security. In addition, the lottery mechanism can also hide the true identity of these participants, and only the users themselves know that they are eligible to vote at some point. The signature key used for voting is temporary and expired, which protects the privacy of the user and also increases security (because it is difficult for the perpetrator to attack or corrupt it).

2. Select a candidate block with the most "verifier" consensus through a hierarchical consensus (Graded Consensus) and complete the validity verification of the candidate block.

3. Run a binary Byzantine protocol BBA ★ (accept block or generate empty blocks), equivalent to the PBFT submission phase, determine the final consensus output block.

Algorand can be independent of miners, all users have the opportunity to block out; as long as the number of tokens owned or controlled by the attacker is less than 1/3 of the total, the probability of forking in the network is negligible, and the transaction confirmation time is with the user. The increase in the number has not changed much.

The principle of VRF can also be seen from the four functions contained in the verifiable random function: 1. Generate a key, generate a public key private key pair; 2. Generate a random number output; 3. Calculate a zero knowledge certificate; Verify the random number output. The specific process is to combine the previous random number (the original random number is given by the protocol) and some variable representing the height and the round, and sign it with a certain private key (or first signature and then combine ), and finally hash to get the latest random number. The random number generated by this is easy to verify that it conforms to the algorithm and implements "V"; and the hash return value is randomly distributed to achieve "R".

2.3. System structure

Algorand includes two types of nodes in the system design: the relay node and the non-relays node.

The non-relays node can be run by anyone. The node holds the valid participation private key of one or more online accounts and participates in the network election.

The relay node needs to have public IP and open ports, so that other nodes can link and communicate, and also maintain all the accounting data, which is equivalent to the responsibility of the network "relay".

3. Auction mechanism analysis and transaction derivation prospects

3.1. Dutch auction + 90% repurchase mechanism

In addition to technical innovations, this may be more interesting. Algorand has designed an auction plan with a time span of five years: two auctions per month, with a monthly auction target of 50 million. Algo. The auction is conducted in the form of a Dutch auction, and the auctioned algo can be returned to the project party at a maximum price of 90% after one year.

3.2. Analysis of the reasons for designing this mechanism

Before we begin to analyze Algorand's design of this bidding mechanism, we need to clarify the various auction mechanisms:

Auction auctions that are common in everyday life can be divided into four categories. They are the Sealed first price auction, the Sealed second price auction, and the open Dutch auction and the English auction.

1. In the Sealed first price auction, each participant submits an auction that is unknown to other participants, and the highest bidder pays the highest price.

2. In the Sealed second price auction, each participant submits a bid that is unknown to other participants, and the highest bidder pays the next highest price.

3. In the English auction, all participants start from the lowest price and continue to increase the price until no one has a higher price.

4. In the Dutch auction, all participants can also see other people's quotations, the starting price is the opposite of the Dutch auction: the auction price starts from the irrational high price, and continues to lower the price until a buyer is willing to bid. This is the auction of this Algorand.

An example of a Dutch auction is: assuming the Algorand Foundation is willing to sell 200 algos, the following quotes are collected in the market:

1. A is willing to bid $3.5 to buy 50 algo

2. B is willing to bid $3 to buy 50 algo

3. C is willing to bid $2.4 to buy 200 algo

4. D is willing to bid 2 dollars to buy 50 algo

Their bids began to clear from the highest price, A, B each bought 50 algo in 2.4 dollars, C bought 100 remaining algo in 2.4 dollars, D bid too low to get 0 algo. From this example, it can be seen that this mechanism encourages buyers to quote high prices because the purchase price of A and B is actually lower than their psychological price, which means that for most buyers, the final transaction price is lower. Its estimated price. It is worth mentioning that US Treasury bonds are also sold in this way.

After understanding the auction mechanism, we then analyzed the reasons why Algorand designed this mechanism.

(1) Why not directly priced but auction?

The Algorand team has also stated that prices should be discovered by the market rather than by the project party itself. This is also the decentralized, democratized financial model that the Algorand team has said.

At the same time, project-side pricing is not a simple task. Due to various considerations, the team tends to overestimate the price, and often it is impossible to raise enough funds.

(2) Why not use traditional auction methods?

A very interesting fact is that the desending price auction (the Dutch auction) and the first price auction are strategically equivalent. That is to say, the strategy in the Dutch auction will have the same result as the strategy in the first price auction. Because although the Dutch auction is an open auction, the price in the process does not contain any valid information, so the strategy adopted in the Dutch auction will be as effective as the strategy in the first price auction. Specifically, we do not expand in detail here, interested readers can refer to the relevant literature [2].

So since the strategy is equivalent, why not use the traditional auction method?

This may be due to the PoS mechanism itself. As stated in the previous consensus section, Algorand can ensure that the consensus is valid through cryptography, provided that more than two-thirds of the algo is held by honest users. Naturally, the more dispersed the token holder, the better the security of the entire network.

In a Dutch auction, you can get algo as long as the bid is greater than the final auction result. So Dutch auctions can make as many bidders as possible to make algo, making the network more secure.

In addition, the Dutch auction mechanism will allow auctioneers to buy at the same price.

(3) Why design a 90% price repurchase for 1 year?

On the one hand, the purpose of the design is probably for the ultimate stability of the price. In the actual transaction process, it is true that the participants have a well-known common "psychological reserve price" to manage the trader's expectations in a transparent manner.

On the other hand, the team dared to design the repurchase mechanism, also because the team had already obtained a lot of financing through private placement, etc., with the back-end funds as the backing. Of course, in the future, there are risks that cannot be finally redeemed due to poor team management, extreme market conditions, etc., and participants need to pay attention.

3.3. Is there an optimal strategy for auction?

So, do participants in Dutch auctions have the best strategy to participate?

It is worth noting that since the auction's algo can be repurchased by the project party at a price of 90% one year later, this auction is not a simple algo, but an algo + put option . The pricing method for put options is discussed in detail in the article [3] published by Amber AI before the auction begins. In this article we review the structure of the auction and the impact of this structure on the bidder's strategy.

From an economic point of view, the best strategy question for an auction is equivalent to whether there is a dominant strategy for a Dutch auction. The good news is that there is a Nash equilibrium solution. The bad news is that this Nash equilibrium solution is a Bayesian equilibrium, which means that there is no fixed optimal strategy for each participant. Each player's strategy is based on an understanding of other players. In other words, in such an auction, there is no single bidding strategy that is fixed by all parties. Each participant’s bid must take into account the distribution of bids from other participants . From the perspective of life experience, the more you get the bidding opponent's pricing for algo, the greater the advantage in this game.

But in addition to focusing on the auction opponents, there will be some changes in the next auction: algo has started trading in some digital asset markets. Focusing on the trading price of the secondary market before the auction may make the auction more efficient.

It should be noted that the put options obtained in the auction cannot be resold. Only the account created in the auction has the right to sell the algo back to the project party. When the option is about to exercise, the customer of the ordinary account can't directly ask the algo project to repurchase his own algo, that is, only the money of the algo is traded in the secondary market, and there is no quota for the put option.

As can be seen from the above discussion, since the auction contains options, a simple formula can be formed as follows:

Auction Price = Current Token Price (p) + Option Price

Then, we quote the assumptions and conclusions of the Amber AI report. The auction price is 1.8p. We can inversely calculate the price of the auction as 1.33usd for the auction and 1.06usd for the option.

Before each auction, participating auctioneers can use this method to re-estimate a theoretical price.

According to the above formula, the transaction price should be less than the auction price within a short time before and after the auction. But why is the current transaction price (when this report was written) higher than the auction price?

There are two possible reasons for this:

1. From the transaction price point of view, algo actually finished a credit-filling market after the auction.

2. Since not all individuals and institutions optimistic about Algorand are eligible to participate in the auction, there is a certain threshold, which forms a partial premium.

3.4. Bidder's trading strategy

Although there is no best auction strategy, after starting the transaction, the successful bidders can still design a reasonable strategy based on the actual situation.

The more straightforward situation is that when the secondary market transaction price is higher than the auction price, the auctioneer can directly sell the profit. In addition, there are some cases specific to Algorand:

Linkage with the secondary market

Algo's holders continue to participate in the auction and participate in the auction at a relatively high price. This can also drive the secondary market price to a certain extent to sell profit.

Current arbitrage opportunities

For auction participants, when the market falls, you can buy in the band and wait for future buybacks.

For example, an algo currently auctioned for $2.40 is assumed to be trading at $3. An effective strategy for auction participants is to choose to sell algo while looking for a third party to sell their own auction account. The $3 algo costs only $1.33 for auction participants, and has been profitable 1.5 times in two days. At the same time, there is a valuable one-year option in the hands of the auction participants. If conditions permit , the auction participant may attempt to choose to split the account with the option to obtain an additional premium. Assuming that the bearish option price falls to 0.5p at $3, each option can also earn 0.66usd.

3.5. Situations that may result in losses

From the above analysis we can see the exquisite design of algorand in the release. But will it be sure to make a profit without using the above strategy? The result is obviously not the case. There is a certain selling pressure when the overall valuation of the project itself is considered, and the factors such as the early investment and unlocking in the secondary market, or the algo without the 90% repurchase mechanism generated by the Stake are combined. But these selling pressures only threaten buyers in the secondary market. For auction participants, their main risk is the premium of options.

We can do some scene deductions.

Auction bid is too early or too high

Due to the psychological expectations of the previous auctioneers, many participants may bid at the beginning of the auction. Due to the large number of participants, the final transaction price is higher, even in some extreme cases, the price is much higher than the secondary market, but the price of the secondary market is still above the interval of p but less than 1.8p. Concussion, then because the auction price is too high, participants will likely be "quilted", as time goes by, the price of the option returns to zero, and the auctioneer loses all option costs.

Falling market trend

If the transaction price continues to fall, the auction price is likely to decrease. Any algo buyer in a secondary market should always remember that only the account participating in the auction contains a put option. The token traded on the secondary market is just a token, regardless of the option. If algo is in a downtrend, the first option after the auction participant gets algo is to choose to sell the algo and buy it back below the strike price to get a risk-free return. Participants in the auction can be in a favorable position in the transaction as long as they can buy the put option at a reasonable price.

However, these funds, which are potentially used to support prices, are limited in size and can be estimated by the amount of algo sold at auction. Therefore, when the price repeatedly fluctuates and falls, the amount of the early auctioneer may not be enough to support the price and cause losses.

Reference material

[1] Orange Book. Third-generation public chain in advance, Algorand's atypical decentralization [EB/OL]. [2019-06-22]. https://mp.weixin.qq.com /s/7vRkR1QaLg-strTqCXRWJA.

[2] KRISHNA V. Auction Theory [M]. 1st edition. Elsevier, 2002.

[3] AMBER AI GROUP. ALGO Auctions Cheatsheet – Amber AI Group – Medium[EB/OL]. [2019-06-22]. https://medium.com/@amberaigroup/algo-auctions-cheatsheet-aab001d277bd.

Author: Fire Academy currency block chain