Evolutions in Liquidity Sourcing

Let’s start with this: why do so many platforms nowadays have Liquidity Mining programs? It wasn’t always done this way; most of the older tokens such as MKR and LINK have never done this. So why did this catch on? Let’s consider this from the perspective of a DAO. The basic answer is so many modern Defi applications need liquidity at a certain depth for their platform to function. Manipulating interest rates with emissions attracts liquidity locusts and those locusts serve some purpose. There is something they need liquidity for. Stablecoin protocols such as Alchemix, Synthetix, Liquity, and Abracadabra use liquidity to protect their peg. Expiring derivative platforms such as UMA, and rate speculation platforms need liquidity for price discovery prior to asset expiry. XToken has CLR pools to grant exit liquidity on staked assets with lock times. The list is endless and growing all the time. Acquiring liquidity at the necessary depth is costly. To add insult to injury, when a DAO incentivizes liquidity they do so with their governance token. This in turn necessitates sourcing even more liquidity in the form of a “Pool 2” farm (e.g. ALCX-ETH). Without this the price would just plummet, the emissions would be worthless, and the system would cease to function. This can more then double the required emissions. It’s a bit of an uncomfortable Ouroboros situation where emissions of a token are used to incentivize buying and holding the token. In the end, hodlers are paying a heavy inflation tax for the privilege of the liquidity.

A further uncomfortable truth is emissions only draw liquidity as long as they last. You have to keep emitting to retain the liquidity. The locusts are extraordinarily mercenary. I call traditional liquidity mining renting liquidity. This is effective in the short term and expensive in the long term. Usually, when structured as I outlined above the base emissions gradually overwhelm the demand for the governance token and drive down the price of the governance token on a long enough time frame. You have to eventually answer how the protocol is going to pay for the emissions and run a balanced budget. To show price growth under heavy emissions the protocol has to also be aggressively growing at the same time (e.g. SPELL). Case in point compare the price of DPI (Defi Pulse Index) against the TVL shown in Defi Pulse. TVL goes up and to the right, DPI price… doesn’t. Why is the success of protocols not reflected in their token price? The answer is often liquidity mining.

I measure the effectiveness rented liquidity by how many $’s of liquidity you draw for $1 of emissions. The effectiveness certainly varies by asset class. Renting stablecoins like for the alUSD generally requires 10-20% APR. That means it takes $1 of emissions to rent $5-$10 of liquidity for a stablecoin pool for a year . Renting ETH has a ratio of about 10-20:1 in its pure form. Renting liquidity for your Pool-2 is usually the most expensive type of liquidity. Many protocols aren’t even getting 1:1. By the end of one year, they could have just bought the liquidity they rented.

Protocol Owned Assets

Recently some DAO’s such as Alchemix have started doing exactly this using Olympus Pro. You can see Scoopy’s reasoning in this tweetstorm. Olympus DAO writes:

Liquidity mining, unsurprisingly, faces the same drawback as PoW mining: it is a perpetual expense with no lasting benefit. Liquidity mining can be equated to renting — it may be cheaper than buying at first, but stop paying rent and you no longer have it. As DeFi matures, it is becoming increasingly clear that incentives are not a viable long-term strategy for networks. The goal should always be to bootstrap and accrue long-term defensible value, rather than perpetually pay high interest on mercenary capital.

Basically they are saying Proof of Work is to Proof of Stake as Liquidity Mining is to Bonds. I agree with that much. Where I disagree with them is that it’s more economical to buy LP’s via bonds than to buy LP’s directly. At a basic level, as a bond buyer you shove $1 into the bond contract and you get $1+X dollars worth of governance tokens after some delay. What is the advantage to the protocol of buying $1 of LP for $1+X instead of just buying $1 of LP? They are buying the LPs; they are just adding some conditions on the sale via this bond program. But why pay Olympus DAO a 3.3% fee for the privilege of buying an LP? Clearly this isn’t more capital efficient than just buying the LP.

What the bonds actually do is add a useful delay which adds risk to buyers. Imagine you are a DAO treasury with a bunch of tokens. Your pool-2 is highly illiquid because you don’t want to do liquidity mining. So, you set the bond to like a 1 year time and offer a steep discount of something like 50%. For reference, a 50% discount would be equivalent to the 100% APR XTK is having to offer LP’s for their XTK/ETH pool right now. The difference is that when the bond matures the protocol owns the liquidity as opposed to having rented it. At a minimum, the treasury didn’t market sell their governance token into a highly illiquid pool when they sold the bond unlike if they had to buy the LP using their governance tokens. In the short term it actually goes the other way; people with ETH had to buy the governance token to mint the LP to buy the bond. In other words it bootstraps more efficiently. Further, the bonds can be streamed out at a variable market rate unlike standard emissions which tend to be a constant rate over time. As a result, the token can maintain its value in the short term while the protocol becomes established and by the time the bill is due there is hopefully a deeper liquidity pool to absorb the bond maturing and hopefully there is revenue to create base demand on the token somehow. Furthermore it changes some game theory dynamics in important ways when hodlers are sure there is price support. This is a win for the DAO if investors are willing to pay for the bonds and risk their capital being locked for the delay.

Liquidity Direction Rights

The last option available is a bit more abstract: Liquidity Direction Rights (LDRs). Basically it is the right to decide where liquidity is deployed. As a simplistic example, imagine I was giving away $10k. It’s a foregone conclusion that I’m going to donate it somewhere. However, you can bribe me to influence my decision on where it goes. How much is the power to influence that decision worth? If one of the places I am willing to donate it to is also a place you were independently going to donate to then securing my donation is worth up to $10k. Any amount you bribe me for less than that is basically a leveraged donation by you. If you can secure $10k to your cause by bribing me $1 then bribing me is 10,000x more effective than donating yourself.

This might sound insane but this is literally what is going on with Curve. Their DAO has decided on an emission schedule for CRV for LP’s who stake in a gauge. What they didn’t do however is lock in the rates per pool. Instead CRV voters vote on which pools get what share of the emissions. Then the DAO explicitly enabled a bribe system to pay CRV voters who vote for a pool you incentivize. In so far as directing CRV emissions attracts liquidity locusts having control over CRV emissions indirectly grants you liquidity direction rights for all assets on the Curve exchange. Locking CRV and voting with it is owning LDRs. Bribing CRV voters is renting LDRs.

Sitting upon this system is Convex. Locking CRV provides multiple benefits which increase the longer your lock is. First, your voting weight is increased. Second, crvLP’s you lock in gauges receive amplified CRV emissions if your address has more CRV backing it. Convex has engineered a way to decouple these benefits and secure a monotonically growing share of Curve voting power. The Convex platform provides a market-based liquidity to CRV pairing system that gives both CRV holders without assets and asset holders without CRV extra income but the real magic is that they tokenized locked CRV. You see, on the Curve platform no token is returned when you lock CRV. Convex created cvxCRV to change this. cvxCRV decouples the asset value of the underlying CRV from the governance rights it otherwise provides. All the voting power of the cvxCRV goes to CVX voters while cvxCRV retains the asset value of the underlying CRV. This makes CVX a pure LDR token whose value should approach the value of the CRV emissions for the same reason that the 10k donation example above should approach 10k in value in bribes. Yunt capital wrote an excellent CVX explainer if you want to dive deeper.

From a DAO perspective this must sound promising. If they need to incentivize Curve LP’s anyway and CRV or CVX is undervalued compared to the value of CRV emissions then how much of a multiplier do they get for investing in LDR’s compared to Liquidity Mining or Bonds? How much CRV issuance do you control with $1 of CVX? What is the multiplier of using your emissions to bribe/rent LRDs compared to paying LPs?

At a baseline 1 CVX currently has the voting power of about 7.5 locked CRV. Over the next year the CRV emissions subject to Convex voting is about 225M CRV. Now about 35% of the voting weight of Curve is owned by Convex. That gives Convex control over about 80M CRV emissions. At current prices the CVX market cap is $840M while those Curve emissions would currently be worth about $400M. That means buying CVX should be about as effective at renting liquidity over a year as half that amount of capital in liquidity mining. Over 2 years, it pays for itself and then you own the LRD forever.

Regarding CRV bribes we have a more direct empirical measure. Here’s a quote from a recent experiment of Alchemix. From the Alchemix Discord (they recently started Curve bribes).

for the record, we went from 95m TVL on Saddle to now 48m on Saddle and 265m on Curve, so 300m TVL total

This was for an ETH pool. That 95M on Saddle was sourced using an alETH staking pool that was paying ~10% APR at the time. So Liquidity mining was allowing them to rent ETH at a 10:1 efficiency. Now they added a 400 ALCX bribe. Bribes are every 2 weeks. So let’s call that 200M in rented liquidity for a cost of about $150k every 2 weeks ($4M a year). That’s a 50:1 efficiency. That’s 5x more liquidity per dollar spent by the DAO compared to Liquidity Mining. It’s an incredible force multiplier which just means that bribes are currently underpriced by the market. Here’s another example. 9k invested for 22M liquidity. That’s closer to 100:1. Just incredible.

Tokemak

The most significant drawback of Convex is just that it only applies to the Curve exchange. Now maybe all Pool 2’s are destined to migrate to Curve following their TriCrypto model but I have no evidence to speculate that yet. How might DAO’s get the same efficiency boost that Convex provides for Curve for their Sushiswap and Uniswap Pool 2’s? Enter Tokemak. I think Tokemak is the least documented/understood of these projects mostly because it’s newest and literally still undefined in places. Here’s how they describe themselves:

Tokemak creates sustainable DeFi liquidity and capital efficient markets through a convenient decentralized market making protocol.

TOKE can be thought of as tokenized liquidity.

If I was to recast this is a more familiar light Tokemak has invented the ‘Modular DEX’. This is akin to the recent meme of a modular blockchain. Here are the modules:

In a traditional exchange capital is mostly owned by market makers and if IL insurance exists at all it is provided as a secondary product through tranching systems. We see attempts at such systems through Saffron finance for example. Just like modular blockchains did to L1’s, Tokemak challenges many assumptions we take for granted about dex’s today. Tokemak is agnostic of any particular trading technology. It decouples asset ownership from asset management by explicitly creating a role for market makers and granting them liquidity direction rights (but not ownership) over decentralized liquidity. Unlike traditional tranching systems the first line of defense against IL isn’t a single concentrated actor but socialized market-wide risk. Lastly, it adds extra liquidity incentives just like Convex has managed to without being exchange specific like CRV or SUSHI.

I learned several useful things from their Gitbook:

  1. The system goes to extreme lengths to cover LP’s IL. They take from the profit on a particular token first, then socialize the loss from the profit of other coins, then they drain the emission rewards for that reactor, then the staked TOKE of the liquidity directors is sacrificed, and lastly they draw from protocol controlled assets such as the DAO treasury. It’s extremely robust. They are serious on this point. The result is safer assets which means less incentivization is required. Most one-sided staking pools offer less than half the APR of a pool 2. This implies DAO’s can spend half as much incentivizing liquidity to join a Tokemak reactor as they would a Pool-2.
  2. $1 of TOKE can deploy up to $1.5 of tokens from the reactor, which is paired with another $1.5 from the genesis pool, leading to up to $3 of liquidity per $1 invested. This makes TOKE a LDR token like CVX. It is literally a liquidity amplifier. Just like with CVX, the LDR isn’t lost but reused in perpetuity. If the system is in profit from trading fees after accounting for IL the liquidity directors (DAO’s in this case) even get paid. Optimistically the DAO won’t need to front all the TOKE, this is only a worst case.
  3. Deposited assets return a 1:1 tToken. This is useful because it’s a Defi lego that allows DAO’s to convert their existing Pool 2 staking pools into one-sided tAsset pools. Combined with point 2, they can directly incentivize assets to deposit to the reactor. Furthermore, TOKE emissions go to tAsset holders and a staking pool is the tAsset holder. This gives LP’s the choice between farming TOKE and farming the DAO emissions. In the latter case this turns the staking pool into a bond-like system where DAO emissions go to LP’s but the DAO is building a reserve of TOKE in exchange.

So what does creating a Pool-2 look like from the DAO perspective? Since they already have their governance token they can deposit that directly to the reactor for one half of the eventual LP. They can back this with $1 of TOKE for deployments sake as their stake against IL. Next they can deploy another $1 of TOKE to the ETH genesis pool and vote to send the ETH to their Pool-2. Now with $2 of TOKE, they have deployed $3 of liquidity. Where do they get the TOKE for all of this? Bonds. As long as the bond overhead is not 50% this is more capital efficient than bonds that purchase LP’s directly.

My conclusion is that Tokemak is a more capital efficient mechanism to source liquidity for DAO’s, especially for pool 2’s. If you believe that protocols will source their liquidity in the most capital efficient way they can and you believe my conclusion that Tokemak is more capital efficient for Pool 2’sthan you can conclude that the future of liquidity sourcing is using LRD’s. Further, if you believe that the aggregate liquidity depth is going to continue to grow exponentially then these LRD tokens are going to need to grow proportionally to the demand for liquidity.

Unsolicited Advice

Lastly, here is my well-intentioned and unsolicited advice to DAO’s as Defi undergoes this inevitable transformation to sustainably sourcing liquidity.

First. Stop overpaying for liquidity. Stop emitting a flat rate to staking pools. Have an explicit numeric target for your liquidity depth. Actively measure the elasticity of liquidity to incentive and pay only what you need. Adjust your emissions frequently. More emissions draw more liquidity certainly but they also suppress your token price which makes future emissions less effective. A well-reasoned emission structure therefore rents only as much as it needs to for its goals and uses the remaining emissions to grow the protocol in a more sustainable way.

Bribes are currently very underpriced but the market for them is rapidly heating up. In the short term you can join the arms race and get a massive multiplier to direct Liquidity Mining by renting LDRs. In the 2+ year term you will be better off buying CVX. Use that short term liquidity boost to free up emissions to buy LDRs and PCAs. Bonds are an effective way to do this without absolutely tanking your token price in the short term. You can use bonds to buy CVX as easily as Pool 2 LPs.

It won’t be long until someone adds a bribe layer on top of Tokemak. I’ll call this protocol Concave Finance until it actually arrives. It’s inevitable though.

The net result of this advice is meeting your liquidity needs in the short term while building sustainable long term value for the protocol. This is good for everyone in the ecosystem.