Hi everyone!
Today I’d like to get feedback on some thoughts that I have had in the last few days, after what has happened in the API3 token sale and the talks that have arised since.
In particular, I found interesting the move from Gnosis, labeling GPv1 as not suitable for IDOs and even discouraging new offerings to be held using the protocol itself.
This sparked the will in the community to build an IDO-specific dapp, leveraging easyAuction as the smart contract suite of choice to perform initial token offerings (see here).
I’ve had a look at easyAuction and I have some considerations on it, mainly about its fairness and “frontrunning” resistance. For those that haven’t read how easyAuction works, I suggest having a look here.
Basically speaking, anyone can set up an auction by specifying the amount of tokens to be sold, the amount of value to get back for them (in turn indirectly specifying the lowest possible price for each token) and the auction’s start date and duration.
Once the auction starts, anyone can place public bids that will be kept on-chain, ordered from the highest to the lowest in terms of price.
When the auction ends, in order to determine the clearing price, easyAuction just loops over the previously placed bids, from the highest to the lowest, and stops whenever it reaches a volume which satisfies the auction’s offered token amount.
At this point, the price for the lowest valid bid is taken and applied to all of the higher bids, so that all of the successful bidders end up paying the clearing price.
The interest of every investor participating in the auction is, generally speaking, not only to get the lowest possible price they’re willing to pay for each token they are buying, but to get the most bang for the buck. Higher bids are almost surely getting into the sale at the same price as the clearing price, but will most likely get less tokens related to bids closer to the clearing price itself, because more value has to be locked to confirm them.
If the orders are public (like in easyAuction, where they are encoded on-chain but easily analyzable and decodable by anyone), the bidders have all the information that is needed in order to place the best orders possible. They know when the auction ends, they know the volume for each bidded price etc.
Let’s make a practical example to see where the problem could lie:
Let’s image an initial offering of 10k TKN that are sold to raise at least 100k USDT. The minimum price that each token must have to reach the target is 100k USDT / 10k TKN = 10 USDT, so any bid below this threshold would just not be considered as valid. The auction lasts 5 minutes.
- Minute 1: a bidder comes in and is willing to pay 15 USDT per token to buy up 5k TKN, which would theoretically result in a net expense of 75k USDT.
- Minute 2: another bidder comes in and is willing to buy 2k TKN paying 12 USDT per token, with a theoretical expense of 24k USDT.
- Minute 3: a third bidder comes in and buys up the remaining 3k TKN for a price of 10 USDT per token, with a projected expense of 30k USDT.
At this point in time, any informed bidder can see that the clearing price has been reached and is 10 USDT. As long as the third bidder does not cancel their order, the only way to enter the sale right now is to place a bid higher than the one at 10 USD, for the same amount of tokens or less. In a way, any bidder would be incentivized to bid at the last second, at a just a fraction more price for the same volume (as long as the price they’d have to pay is low enough for them), to get the clearing order.
So a bidder could place a bid for 3k TKN at a price of 10.01 USDT per token, another one at a price of 10.1 USDT per token, and so on and so forth, all at the last second.
This should in general be good for the auctioneer, and an expected behavior in a “normal” auction, because an increased clearing price means more raised funds given the same sold volume, but not extremely fair or optimal in general, because at the end of the day, the one bidder that bids at the last second theoretically has all the information to place an order that will almost surely get filled and either determine the clearing price, or be very close to it as long as the current clearing price is low enough for him to slightly increase it in a subsequent order.
This happens even if the original idea for the bidder was to bid higher after the research they’ve done on the project at hand. Maybe a fourth bidder would have been more than willing to come in and buy up 3k TKN at 11.5 USDT, but why would they do it if they can just wait until the last second, see what price they’d need to pay to get into the sale and just pay it as long as it is less or equal than their original estimate?
It has to be said that there’s a risk in doing this, which is to not get the amount of tokens they initially wanted to buy up, but potentially less (because in the meantime other bidders could have placed higher bids for lower volumes than 3k TKN, not clearing up all the sold volume, and leaving just a fraction of it left in the sale.
I can see a kind of a more general “problem” here, and that is that the more information is available, the more these kinds of games can be played, unhealthy competition is sparked, and less focus is put on “healthy” price discovery, while more on how to “trick” the system.
In a way, we can consider this a kind of frontrunning on auctions. Quoting from this article:
Frontrunning is a form of information theft. It weakens the market’s competitive edge by decreasing the incentive to collect original information and base one’s trades on it—it depresses mass adoption by promoting unfair pricing mechanisms.
Which I think kinda summarizes the scenario.
So I thought about how to try and fix this (and by this, I mean the exploitation of market information to get an unfair advantage over other investors, mainly retail).
I reached the conclusion that sealed bid auctions might be a good solution, and could be implemented using some sort of commit/reveal scheme.
The general idea is to have a sort of commit/reveal scheme where bidders commit to buy a certain volume of token at a certain price, but the only public and visible thing in this whole process is the fact that the bidders transfers the amount of tokens they’d use to perform the buy to the auction contract. The amount that they want to get back in exchange for that (which in turns defines the price) is kept secret, and passed in as an opaque value (for example it might be the hash of the actual value plus a message signed by the user to increase entropy and avoid attacks trying to find out the final value).
All of the commitments would be taken in in the first phase of the auction, on which it is possible to place sealed bids. Once the first phase ends, the second one starts, in which every bidder has to fully reveal their original bid. They call a function giving in the amount they wanted to buy and the private, signed message used in the previous step to commit, and a hash of the two values is calculated on-chain. If the calculated value matches with the one that was previously given, the order is indeed confirmed. In this scenario, order checks on prices etc would be done right now, so any order below the minimum price could just be ignored for example.
This second phase, just like the first one, would have a time limit, at the end of which the orders would be public, transparent, and guaranteed to be genuine by cryptography.
At this point the clearing price determination algorithm can just be run.
There’s the risk that for whatever reason a bidder that had previously committed to a certain price might not want to reveal it after seeing the other bidders’ votes revealed, but at that point he could simply get slashed, because of the amount that he locked in in the first phase (the auction contract still has it as a “hostage”, so things could be done).
My hope is that an auction where bidders cannot be influenced by other bids due to lack of information would bring the bidders themselves to a deeper research on the project/market at hand, and in turn in “better-placed” bids (that hopefully translate in a healthier price discovery and determination altogether).
This is the basic idea. I’d really like to get some feedback by people that know a lot more than me about this, because it’s like the first time that I’ve really thought about these things. Hope this can be helpful or even just spark a good discussion!