Decentralized finance (DeFi) allows strangers on the internet to earn money much the same way bankers do, by earning fees on financial services.
In most cases, people do that by providing liquidity. Those investors “stake,” or lend, digital currencies, not dollars or euros or yen, and they are lending the digital currencies to apps rather than to people or companies. One set of apps with the most need for liquidity and the heaviest flow of transactions is decentralized exchanges (DEXs), particularly automated market makers (AMMs) led by Uniswap.
But how does anyone decide which AMM is the right one for their funds?
This post is a starting point for each investor in the almost $60 billion DeFi market to begin doing their own research on which exchange to take part in. Joining a particular DeFi community can mean much more than just putting up funds. It can also mean participating in discussions, helping out new users, voting on governance decisions and even writing suggested improvements for a software protocol's code base.
And yet here we are. Decentralized applications that many CoinDesk readers will have heard of, such as Uniswap, Curve and SushiSwap, are all now in the top 10 decentralized applications as ranked by DeFi Pulse. Just on its own, Uniswap's worst week so far in 2021 was $5.8 billion in volume, usually coming in somewhere like $8 billion a week since mid-February, according to its own stats.
Once a sector is putting up numbers like that, there's no need to compare it with similar fields any longer. AMMs have become a respectable business on their own. Even if they never grew from here, it would be worthwhile to a lot of people to be part of such a market.
But they do need to entice investors to join.
Facebook is terrible, but we all keep opening the app because everyone is there. That's also why no one leaves, even though we all wish we could. One could say Facebook has social liquidity.
DEXs work much the same way. The exchanges work better the more liquidity they have on hand, and so they compete as much on providing attractive deals for liquidity providers (LPs) as they do on being affordable and easy for traders.
So, for those considering putting some of their crypto assets to work by becoming an LP, what follows is something like a map of the various automated market makers of significance out there so far. It's not a comprehensive guide, and it's definitely not meant to do anything like help investors decide which place will make them more money.
This map, again, is a starting point for deciding which community an investor might want to get involved in. Because this is DeFi; once you are in, you are part of it. It's not like buying some Enron stock and seeing where it goes. Backers of these platforms have a voice, and they work best when a lot of them use it.
Each of these applications is part of a specific ecosystem, and each represents a specific vision of the future. Potential backers of such protocols should understand that vision and decide whether to buy it. That's where, we hope, this post can help.
Why it matters
DEXs struggled as long as the market was led by exchanges that attempted to imitate the experience of a centralized crypto exchange like Binance or Kraken. It was the advent of the AMM that allowed DEXs to gain a foothold.
For a great overall explainer that gets into the mechanics of how AMMs work, Haseeb Qureshi did a deep dive on his blog last year that really gets into some of the harder-to-grasp concepts.
Here's the core insight of the AMM: Users don't trade against another person – they trade with the smart contract, a type of software program. An AMM is always willing to buy and always willing to sell at a stated price.
The basic AMM works as a series of pools of two assets (say, ETH, the foundational currency of the Ethereum blockchain, and DAI, an Ethereum token designed to hold its value against the U.S. dollar). The price offered by the AMM does not reference the external world. In the simplest form, it's just a function of the ratio of the two assets. So if there were 1 ETH in the pool and 2,000 DAI, 1 DAI would cost 0.0005 ETH and 1 ETH would cost 2,000 DAI.
However if someone bought 0.01 ETH for 20 DAI, then the price would change; now a DAI would be 0.000495 ETH. A tiny change, but the shift gets bigger with bigger trades. Each trade changes the underlying price. A big trade would knock a pool out of line with other markets, but arbitrageurs can be relied on to trade it so it's back in line; that's not a bug, but a feature.
Furthermore, an AMM with enough token pairs can trade between any two tokens listed, even if those two tokens aren't in the same pool. A user can trade BAT for ZRX, for example, by trading BAT for ETH and then ETH for ZRX, all in one transaction in the Metamask software wallet. Very smooth.
Understanding those iterations can help a potential liquidity provider understand which DEX squares with their worldview.
"A lot of these are functionally the same from a tech perspective," said Will Price, a DeFi investor active in this area. "It's just a difference in community and a difference in what they choose to incentivize about yield farming"
In what follows, we describe each protocol for a liquidity provider. They all follow the same pattern, describing each of the following where applicable:
- First, we describe how the protocol works with the intent of illuminating what niche it aims to fill,
- Second, we deal with any standout features offered to traders that might make it attractive,
- Third, we cover any less obvious sources of yield, such as liquidity mining opportunities
- Finally, we describe its approach to building community, mainly focusing on governance.
Uniswap remains the clear market leader. As of this writing, it has $6 billion worth of liquidity committed in the protocol according to DeFi Pulse, making it the third-largest app in DeFi.
Uniswap just announced version 3.0, which considerably refines the notion of an LP in an AMM. Big picture,with this new version, Uniswap signaled that it's laser focused on being the best AMM, full stop. It's not trying to expand into other areas of DeFi. Instead, it wants to continue to lead in its niche.
The new version is slated to launch May 5.
The new version solves the problem of so-called wasted liquidity. This can be a little tough to grasp, but let's give it a shot.
Remember how we wrote above that each trade shifts the price in an AMM pool? More accurately, the price shifts even within a trade.
Most of the time that doesn't really matter, but if someone trades a significant amount of funds in one pool, an amount that has a meaningful impact on the price of the pool, their final price for the trade will actually be an average of a series of prices.
So in our example above, imagine a pool that started with 1 ETH and 2,000 DAI. If someone wanted to buy 0.5 ETH, the trade would throw the price of ETH and DAI wildly out of whack with the larger market. In fact, it would be so severe that it's unlikely that anyone would ever make such a large trade.
So, if no one is ever willing to make trades that use most of the liquidity on one side of any pool, that also means that no one would ever buy nearly as much ETH as sits in the whole pool nor nearly as much DAI. Only a portion of it is ever likely to be traded, and so there is wasted liquidity.
"How capital-efficient do you want your position to be?” Price said. “Most tokens in a Uniswap pool don't actually get touched, because you're providing liquidity from negative infinity to positive infinity."
So Uniswap's next version will create a different system, one in which a liquidity provider can define a range of the AMM curve at which they will participate. If the price falls outside of that range, their liquidity just sits it out. That might sound like wasted liquidity, but people are going to load up in the part of the curve where trades are most likely. So it actually means there will be a lot more liquidity.
That has the impact of allowing traders to make much larger trades within the normal expected range of a given pair.
Stablecoins provide the easiest example of why this is powerful. One USDC should generally trade for 1 DAI, because both are pegged to the dollar. Yet in a simple AMM, a big trade can knock the prices for the two widely off the mark. With Uniswap version 3.0, though, a trader will be able to define their liquidity as only participating in trades where DAI sells for no more than 1.01 and no less than 0.99 USDC. If they put in $1 million of liquidity with that parameter, then a trader could be guaranteed a trade of, say, $500,000, with slippage no greater than a penny.
It should be noted that this arrangement will very strongly favor the most sophisticated market makers. For small investors looking to make passive income, Uniswap could be tough, but the change would be good for all traders, including small ones.
The new version could also have spillover effects on other platforms.
"The existence of these concentrated liquidity positions means the best execution of the trades is going to wind up on Uniswap a bigger portion of the time," Price said. "The returns to passive LPs in other systems are going to go down."
That said, several market sources said they expect startups to come along that take retail investors’ assets and actively deploy them on Uniswap, so that normal people can once again set it and forget it on the leading DEX.
If Ethereum were comparable to personal computing, then Uniswap up to now has been like the DOS era, where any decently adept hacker could code up something others might use. The new Uniswap version seems tantamount to the Mac/Windows era: easier for everyone to make use of but a tougher environment for individuals and tiny teams to build in.
The upshot of these customizable pools is that it also would enable clever traders to put in orders to buy at certain prices if they get hit. So, for example, a trader could plan ahead to "buy the dip." Uniswap is calling these "ranged pools."
Lastly, Uniswap will allow LPs to customize fees somewhat on pools in version three, so they could, for example, charge a higher fee on less frequently traded pairs.
Fun fact about v3: Pool accounting will no longer be done with ERC-20 tokens, a common type of asset created on Ethereum. Instead, each stake will be unique and tracked using Ethereum's non-fungible tokens standard, which spawned the recent NFT craze. That will make the LP tokens less composable in the immediate term, meaning it will be harder to mix and match them like Legos to build innovative money applications. But no doubt entrepreneurs will soon find clever ways to make use of the new NFTs.
Uniswap was the last of the big AMMs to release a so-called governance token. But when it did, “airdropping” UNI as a sort of participation award to every past user, it was probably one of the biggest moments in DeFi for 2020, which was already an historic year.
While Uniswap briefly ran a liquidity mining program to further distribute UNI as a bonus for liquidity providers in key pools, it was discontinued as planned and has not been renewed. Smart money would probably bet on a new distribution coming along to encourage migration from v2 to v3 pools.
Uniswap is a venture capital-backed company, one that appears designed to enable its core team to retain control of it for quite a while longer. It requires at least 4% of UNI to vote in any governance decision for the final decision to be effective. Failure to reach such a quorum is what doomed the two unsuccessful measures.
That desire for control for a time is also reflected in Uniswap’s decision not to immediately make version 3.0 of its software open-source, or free for anyone to copy and modify.
SushiSwap has a reputation as the community-powered AMM.
Whereas Uniswap is the AMM for insiders, SushiSwap has opened itself up to the wider world, and never stopped shelling out governance tokens (ticker symbol: SUSHI) to its liquidity providers.
SushiSwap took its place in DeFi by force, stealing the limelight from Uniswap just long enough to create what seems to be a lasting place for itself in DeFi.
If the “DeFi Summer of 2020” had been a novel, SushiSwap's debut would have been the final act, covering the last few chapters. Using a clever construct that's come to be referred to as "vampire mining," the Sushi chefs convinced Uniswap LPs to turn over their tokens to its control and let the protocol move all their assets to SushiSwap.
Here was a weird part, though: After SushiSwap ended the extraction of funds from Uniswap's pools, Uniswap actually had more assets under its control than it had when the struggle started.
When the dust settled, Uniswap was back on top, SushiSwap had different leadership and had fallen considerably in the rankings. The drop in SUSHI rewards and then the debut of UNI caused funds to come rushing back to the O.G. AMM.
For a short time, it looked like SushiSwap would become an interesting footnote in DeFi history, but it wasn't so. It's often said that "community is everything" in crypto, though it usually comes off as rhetoric.
SushiSwap has since distinguished itself in a few ways, and now has almost $4 billion in total value locked (TVL) in it, as of this writing.
"If you're reading their Medium posts and their updates, the team is building in a very interesting direction," said Santiago Roel, a partner at ParaFi Capital. "It's definitely innovating and moving in a direction different than Uniswap."
SushiSwap has been actively helping new, small token projects get traction by promoting them through its Onsen system. Onsen pays out extra rewards in SUSHI for those who stake to it. This helps attract new users and builds loyalty to SushiSwap as certain projects find their footing.
"Some coins are only initiated on SushiSwap," said Jason Choi, general partner at Spartan Capital.
To be fair, coins also launch on Uniswap, but they don't get a special deal out of it. Onsen has been a business development strategy for SushiSwap, and it appears to have been fruitful.
On an obvious level, marrying a lending facility with an exchange creates options for leveraged long positions or shorts, but we may see even more creative trades before long.
"I think their niche is, they ship periphery products very quickly. SushiSwap is more focused on building products around the core products," Choi said. Version 3.0 is well underway.
"I find it interesting when teams deploy on other chains," Roel said, but "I don't think they've found meaningful traction"
Uniswap and SushiSwap have the same trading fees, but crucially, SushiSwap never stopped its liquidity mining program. Users who stake their LP tokens on SushiSwap's "Sushi Bar" earn SUSHI continuously. The ability to drive additional earnings makes SUSHI attractive above and beyond its governance powers. By contrast, for now, UNI does not earn anything.
The fact that SUSHI earns revenue isn't its only feature. It also has been driving a decent amount of governance activity.
SushiSwap moves through quite a few proposals, as can be seen on its voting page.
BentoBox is a good example: It was an idea that came from a community member. Once it was greenlit, he built it and now that it's up, he will earn a small piece of all its fees.
Curve is the protocol that beat Uniswap on stablecoins.
It's the second biggest AMM, measured by TVL on Defi Pulse. It's conducting $286 million in trading volume per day.
Curve Finance constructed a new formula for its AMM, specifically for pools with pairs of tokens that should travel basically in sync with each other. The price of the stablecoins, USDC and DAI and USDT and TUSD, should be generally 1:1, with tiny variations, because they are all designed to follow the price of the U.S. dollar. Similarly, WBTC and renBTC also should travel pretty much in lockstep, because they both ape bitcoin.
With that in mind, Curve was built so that the price curve really only started moving out of step on the far outer edges of the liquidity bands.
It is also worth noting that Curve's early days coincided with Yearn Finance's early days.
Cronje liked that there were lots of opportunities for yield on Ethereum, but he didn't want to mess with worrying about token volatility. So he stuck to playing with stablecoins in the early days. That made Yearn and Curve naturally simpatico.
One key difference with Curve's design versus Uniswap or SusiSwap is that it handles pools with multiple tokens more easily (Balancer, the DeFi portfolio manager, also does that). That made Curve useful to Yearn, because it had strategies for users to earn returns on all the major stablecoins, DAI, USDC, USDT and TUSD. Receipts for deposits in those strategies yielded tokens with names like yDAI and yUSDC.
So, Curve created a pool to make it easy to switch between these four pools. Additionally, users who intend to HODL a given Yearn token can still deposit it in Curve's liquidity and earn transaction fees.
Cronje likes to call himself "Batman for yield," and in some ways, Curve has been his Robin. In the comics, the original Robin went on to become a superhero in his own right. To that end, though, Curve has established a niche in trading tokens that are non-volatile (relative to each other). “We have pools for volatile pairs finishing the last audit," said founder Michael Egorov.
Which means Curve is ready to face off with the other AMMs directly.
As we spoke to Egorov, about Curve recently, he explained that this is one of Curve's big ideas: meta-pools. Curve likes to run pools-of-pools (it can really get quite dizzying).
It has also shown itself willing to chase opportunities beyond Ethereum. Curve just launched a version on the Ethereum L2 known as Matic. In February, it announced plans to go on Polkadot, one of the largest blockchains by market capitalization but still quite new and close but not quite fully functional.
Similar to SushiSwap, Curve users can boost yield by earning its governance token, CRV. Curve had a surprise launch for its liquidity mining program last August.
Liquidity providers in Curve pools can earn CRV, but they have to stake their LP tokens in Curve's "gauge." Curve calls it a gauge because it gives more CRV to people who have staked longer. This reporter tried the gauge once, and the transaction fees to do so were so high, we never staked.
"In general, if you're deploying a few hundred dollars, it might not be worth it for a lot of farms," Choi said. "A lot of the smaller retail guys are basically priced out of this game. If you look on-chain, you see a lot of very chunky deposits."
So, it might not make sense to go this route casually. Try to research in advance how much it will cost in fees before going in. It takes several transactions, and so it's not enough just to check Metamask.
The main governance activity for CRV holders is deciding how much CRV is earned by LPs for different pools. That is one of the ways the community entices new liquidity onto the platform. For example, when UST was becoming popular on Ethereum, the returns on the pool early on were extremely high as they tried to get the pool filled up. As it did, though, and the CRV spread out over more holdings, the returns became more reasonable.
Curve's governance system is complex. It's not the CRV itself that votes, for example. By staking, accounts accrue veCRV, which has no value except that it can be used to vote.
If that isn't already clear, there are a lot of fiddly details with Curve, but some folks are really into that.
Bancor has made a big comeback.
No one disputes that Bancor pioneered the basic AMM model, but through all of 2020, when a Uniswap fork, or splinter project, was able to break multiple billions in liquidity, Bancor had numbers stuck in the low millions, as tracked by DeFi Pulse. That turned around late in the year though, and now Bancor sits just outside of DeFi's top 10.
The new version now has some real advantages for LPs.
First of all, Bancor offers single-sided liquidity, which is unique in the sector. "Single-side liquidity provision is pretty good for user experience," Choi said.
Most automated market makers require a user to enter with two cryptocurrencies, equal in dollar value. Some AMMs will automate swapping half of a deposit for the other cryptocurrency on the fly, but that costs the LP in on-chain computation, or “gas,” fees.
Bancor doesn't do that. Instead, Bancor will simply mint the required amount of the other currency, because the other currency is always its Bancor Network Token (BNT).
When an AMM pairs every currency with the same other currency, it makes it easy to make transfers between any two listed tokens. If a user wants to trade WBTC for BAT, Bancor can just swap WBTC for BNT and then BNT for BAT. To the user, it looks like one transaction. The common token is like a conductor.
Bancor's approach to single-sided liquidity is worth dwelling on for a moment. When Bancor's first token sale was announced, Cornell University professor Emin Gün Sirer argued that there was no good reason for Bancor to use a new token when it could have just used ETH as the common token. Indeed, that's what Hayden Adams did when he created v1 of Uniswap and it worked amazingly well. That approach made Uniswap the dominant AMM.
(With version 2.0, Uniswap opened it up to allow any two tokens to be paired, knowing there would be enough pairs that a path could always be found.)
But here's the twist: Bancor can opt to make changes to its own token. It can't change ETH. So now the Bancor protocol can mint BNT as needed to take the other half of any deposit in the pool. Bancor refers to this as co-investing with LPs.
Basically, all BNT holders pay a little as the supply expands for each new LP, but it's probably worth it because more LPs make Bancor work better, which means more demand for BNT. That should be good for the price in the long term.
The most obvious concern here would be the impact all that minting would have on BNT’s price for its holders in the short-term, but BNT has done great over the last year. The real test will be what happens when the overall market is a little calmer.
Bancor's also got a program for creating deflationary pressure on the token with a portion of fees.
While single-sided liquidity seems savvy, Bancor's claim to fame in some corners is its approach to the hobgoblin of AMM LPs: impermanent loss.
Impermanent loss is one of the most head-scratchy terms in DeFi. It's a particular danger in AMMs.
"You can't get rid of impermanent loss. You can mitigate it," Roel cautioned.
In simple language, AMMs prices are typically expressed through a simple ratio of the two assets. That means that when one asset appreciates dramatically and another doesn't, that pool will end up with less of the asset that appreciated and more of the one that's staying flat (because it needs more of the less valuable asset to equal the value of the appreciating asset on the other side).
We saw that a lot in the early days of the bull run, with BTC appreciating and ETH not really moving. So a WBTC (wrapped Bitcoin)/ETH pool in Uniswap would lose WBTC to ETH as more ETH would be needed to equal the value of WBTC on the other side.
The problem is a lot of times that means the value of your portion of a pool will lose in dollar value even though it's been earning trading fees. When redeeming LP tokens on an AMM, an LP doesn't necessarily get the same quantities of tokens back. They get their share of the pool back.
In a two-stablecoin pool, say DAI/USDT, the value of that share is unlikely to really change (except in trading fee yield). In a WBTC/ETH pool, that can change a lot, and potentially in a direction unfavorable to the LP.
Impermanent loss is called impermanent loss because a lot of times it evaporates if the LP just waits long enough. But sometimes people need to get out and it can be painful.
To go through this with numbers and moving graphs, the Qureshi post mentioned above is good.
Bancor built an insurance policy. Users who suffer impermanent loss on Bancor are guaranteed to be made whole in terms of the value of their initial deposit as long as they stay in for at least 100 days.
The difference is made up in BNT (Thorchain, below, does the same thing, with its native token, RUNE).
One could argue that users are just trading the cost of impermanent loss with the cost of lost flexibility, but for some projects that go into pools with the intention of sticking around over the long haul, that guarantee can be appealing.
"What I've observed is some teams, like Yearn, find that interesting. From that standpoint, that becomes compelling," Roel said.
Circling back to Uniswap, impermanent loss is also part of the motivation behind allowing LPs to define their trading band for a pool. A user won't suffer impermanent loss if their funds go outside a band that would go against them.
Bancor also just launched limit orders. Basically, a user can call a trade in advance if he thinks something is going to happen. For example: "Buy ETH if it drops below 1800 DAI in the next 30 days." That kind of thing. It's an extremely common feature on centralized exchanges but not built directly into many AMMs, mainly because traders sophisticated enough to use limit orders are using trade aggregators, such as the aforementioned Matcha and 1inch, that handle that kind of issue for them.
Bancor is also a multichain project. Its decentralized application (dapp), known as BancorX, allows swaps between the EOS and Ethereum networks.
LPs that commit their trading fees back into the pools can earn increased rewards over time, encouraging LPs to stick around longer.
The governance model's latest iteration only kicked in at the end of March. (Founders were not exactly rushing to start DAOs in 2017 when Bancor ran its record-breaking initial coin offering).
Thorchain is the one of only two AMMs on this list that isn't on Ethereum.
There's not a lot of AMM activity outside of Ethereum. There's not a lot of anything but hodling outside of Ethereum. It remains to be seen what will happen with Thorchain. It is very, very new.
"I wouldn't be comfortable putting liquidity into it yet," Price cautioned.
Thorchain is worth mentioning because it takes a completely different approach to settlement than any of the blockchains above; it trades between blockchains. Uniswap, Curve, Sushi and Bancor are all on blockchains. They trade around tokens within that blockchain. Some can execute trades between separate blockchains, but they use workarounds and hacks to do it.
Thorchain has its own blockchain, natively designed to operate across multiple architectures. Like Bancor, it uses a common token (RUNE) in order to make trades, for example, between BTC and ETH, but unlike Bancor it doesn't need to offer users a wrapped (synthetic) version of either. A trader can put BTC into Thorchain from their wallet and get ETH paid out to an ETH wallet.
Like Bancor, Thorchain offers an impermanent loss guarantee for deposits held more than 100 days. When we spoke to Thorchain’s team before its launch, they said their models showed that there's hardly any such losses for people who hold that long, but you know the old saying: Past performance is no guarantee of future returns.
RUNE is a governance token as well as a utility token on the network. Users earn newly emitted RUNE for serving as LPs.
Those governance tokens will be important sooner than folks might be ready, because the mostly anonymous team behind Thorchain has committed to completely stepping down from the project during the next year, turning further developments over to RUNE token holders.
The project has a considerable treasury as RUNE has appreciated dramatically in price. It could go any number of directions, but the first order of business is adding cryptocurrencies to its exchange as fast as it safely can. Even as it announced its first handful (bitcoin, ETH, litecoin, bitcoin cash and Binance's BNB), many others were well under way in development.
"One good thing about crypto transactions is they cost the same whether it's $10 or $10 million. For hedge funds and whales, these transactions are negligible," Price said.
There's another way to make fees fairly negligible, though: Leave Ethereum.
PancakeSwap, like SushiSwap, is a fork of Uniswap, built for Binance Smart Chain (BSC), which uses the BNB token much in the same way Ethereum uses ETH.
"Currently they have well over 100 pairs of assets," Choi said. "There's quite a lot of new projects launching on Binance Smart Chain."
BSC is more centralized, with only 21 validators leading the chain, using a system much like EOS. That makes BSC faster and cheaper to use, but considerably more vulnerable to tampering by third parties.
According to Choi, that tradeoff is just fine for many users.
"It's a faster chain,” he said. “The retail [investors] don't actually care that much about decentralization.”
BSC is compatible with Ethereum's code logic, making it easy for forks of successful Ethereum projects to be run on BSC. PancakeSwap is a case in point.
But the AMM also ground to a halt in mid-April. Transactions were too much of a good thing:
As with the other AMMs, PancakeSwap liquidity providers earn fees on trades within pools they have participated in. Since volumes on PancakeSwap are very high and transaction fees are much lower, that should create more activity that will enable participants to generate more yield.
Like on SushiSwap, LPs who stake their positions will earn a governance token, CAKE.
PancakeSwap's governance is quite active via its snapshot page. It is largely oriented around approving boosted CAKE rewards for new pools, as a way of generating early liquidity and interest from users, much like SushiSwap's Onsen program.
It also just let CAKE holders decide how to change the transaction fees and how to allocate them.
Launching projects with what are known as initial exchange offerings is still quite popular outside the U.S. and with every dapp on Ethereum available for duplication, there's plenty of new dapps to offer.
A final thought
If this was a lot to take in, get ready. It's only going to get bigger and more complicated. This is an exciting time for someone ready to do some research and thinking, because getting involved in all of this craziness makes it make sense faster.
That will be very advantageous, because it will prepare LPs to better understand (and perhaps profit from) the new wrinkles as they come. And they will come. They will come fast.