If you want some context for the following post, you can find Part 1 of the series here:
While Bancor V2.1 was launched months before Olympus and well before the launch of Olympus Pro, Olympus Pro (and associated forks) are built off the pool 2 models of Uniswap and Sushiswap, so I thought it would be best to address it first.
For the sake of accuracy and partly for fear for my life at the hands of the Ohmies, I will try to steelman the very basics of their position using mainly their wording. Liquidity mining is a perpetual expense with no benefit at the end of the rainbow as your liquidity requires constant bribes to stay around. If liquidity mining could be equated to renting your liquidity, protocol owned liquidity would be owning your own liquidity. Renting proves to be cheaper at first but is more expensive in the long run. To avoid perpetual payouts to the extortionary mercenaries, protocols can choose to buy back their own liquidity using Olympus Pro, which aims to provide “Protocol Owned Liquidity as a Service”. To realize this, Ohm Pro utilizes bonds which allows protocols to buy up LP positions from users in exchange for discounted tokens. In theory, the treasury has now bought back a bunch of permanent income-producing “assets” and rewarded their “community” in exchange with discounted tokens.
To illustrate, let’s return to my hypothetical protocol. I realize I am issuing too much $Patrick to the $Patrick / $Eth pool just to induce sell pressure on my own token as the result of the 2 token staking model. I am paying way too much just to keep a liquid market for my own token at the cost of using my tokens to stimulate other kinds of growth. Using Olympus Pro, I choose to sell $Patrick at a discount of my choosing + pay 3.3% to Olympus for using their BATTLE TESTED smart contracts. First off, if I have $100,000 of $Patrick to spend on this buy back, why not just split half to $ETH and skip the discounts? In theory, the reason would be only if the users who got your discounted tokens were converted current liquidity providers to long term passive holders. However, this discount opens up an immediate arbitrage opportunity which renders this outcome unlikely but not impossible. Let’s say I chose a discount of 6.7% and Olympus takes their 3.3% fee and I get $90k of $Patrick / ETH for $100,000k of $Patrick. A mix of community members as well as savvy profit seeking market participants collectively get $100k of $Patrick for $90k. Both parties have an immediate incentive to sell half of the discounted $Patrick for $ETH and repeat the bonding process each week so long as the price has not dropped during the vesting period more than the discount, a process typically assisted initially by automated bots.
Alright so my protocol overpaid once with initial liquidity mining and again with Ohm Pro. I still don’t have the loyal community of token holders that are earning sustainable yield for holding my token and incented to be committed community members as a result, but at least I have this permanent “asset” and I don’t have the obscene liquidity mining expenses, so we’re not back at square one right? Right? Sort of, but this “asset” is far from ideal for several reasons.
The mandate of every protocol team should be to use their specific knowledge to deeply understand users and then build and communicate great projects to achieve product / market fit. No significant portion of the team's time in the earliest days should be actively managing an LP position on a DEX. By “solving” the yield farming expense problem with bonds, protocols have inadvertently given themselves a new responsibility of being a market maker for their own token. Should they choose to supply passively or without a corresponding hedge and they expect that their token will massively outperform ETH (or whatever Layer 1 pair token) they will be helpless and at the mercy of impermanent loss. If their token outperforms the pair massively, they would have just been better off just holding their token and the L1 token off the exchange and again, even better off holding the full amount in their own token. Passive managed protocol owned liquidity is short volatility and oddly a bet against your own token’s success.
Next, I see typical pool 2s to be clearly flawed from a protocol treasury portfolio construction approach. For every $1 of your own token that you want to supply to an AMM to earn yield for the treasury you have to take on $1 of exposure to another token. The other token, often another L1 token, might not be an asset that the protocol necessarily wants exposure to and if they do, they might not want to have linearly scaling forced exposure since every $1 of my own token that I want to make productive on an AMM needs a corresponding $1 of L1 token. A very practical reason for not wanting this much exposure is that L1s to date have been viewed more as a lazy man’s index and more than several have achieved their current valuations with little correlation to users, transactions and revenue fundamentals. With the defensibility of L1 value accrual in limbo as a result of the rapidly evolving multichain wars, burgeoning liquid staking derivatives adoption and questions regarding the extent of the parasitism of L2s, protocols may not want to be exposed heavily to neither heavy speculation or a major repricing event on a token whose price directly impacts the IL their protocol owned LP position might incur. Furthermore, if the model that the majority of DEXes use which creates a great deal of buying pressure for L1s is deemed suboptimal, there is a small but not insignificant chance that that form of buying pressure disappears.
Finally, we mentioned earlier that as a result of the 2 token staking UX, LPs that choose to compound yield farming rewards and those that choose to sell both contribute to sell pressure on the native token. Should the protocol continue to provide rewards to their Sushi or Uni pools (and associated forks) to incentivize and retain the liquidity they don’t own, they, in conjunction with the other liquidity providers, would now still be contributing to sell pressure on their own token and treasury LP assets as they too compound rewards and so the cycle repeats. Even if I try to buy back all my LP positions, I will likely just be selling to bots and end up at square with no community of users incentivized to share and contribute to my success. A lonely Mr. Potter Protocol with no community.
I hope to have highlighted two potential problems by now; 1. The dual token staking UX of traditional AMMs has systemic and (arguably insurmountable) flaws and 2. protocols and retail users are poorly equipped to properly manage their own liquidity on these exchanges. Tokemak set out to solve both of these issues.
While exchanges like Uniswap and Sushiswap have the highest trade volume and user numbers, their user experience is too complicated. Tokemak simplifies the staking process by building a layer on top of the exchange which allows two separate liquidity providers to each provide one side of the underlying pool. A third ecosystem participant, the $toke holders, decide how much liquidity each token is paired with as well as the exchange that the resulting pair gets staked to. Ideally, this allows passive market participants to earn yield on a single sided token without being involved in the complexities of market making or experiencing impermanent loss.
On the other side, the liquidity directors ($toke holders) provide impermanent loss protection and take on risk in exchange for a cut of the yield generated by the underlying position. Again, since the intention here is to make sure casual users won’t be crowded out of participation by cost and complexity and for the risk to be cleanly transferred from amateur to professional, the liquidity providers would ideally be a distinct user set from the liquidity directors who have specified market making knowledge.
In theory, the explanation seems perfectly reasonable, but in practice the implementation likely has undesirable second-order effects. The source of the inefficiency stems from the product decision to build upon other DEXes, which presents a more limited design space for Tokemak. One of these inefficiencies is that fees are not given out to LPs in the underlying token they put in. All the fees from the underlying exchanges on all pairs are held by a protocol owned pool to be restaked at the start of the next 7 day cycle. Liquidity providers and liquidity directors are instead rewarded with $toke, representing a potential proportional future claim on those protocol controlled assets (the LP positions and their rewards). As both LPs of both single sided pools are rewarded with $toke they are presented with two options. They can swap their $toke to retain full exposure to the asset they originally staked and the rewards they sought to collect, or they can re-stake the $toke to gain liquidity direction over a pool. Should LPs decide to retain true single asset exposure as advertised, reward sales would create immense sell pressure that would limit the ability for Tokemak to backstop impermanent loss risk with their own token’s value. However, if the LP restakes the $toke rewards, it creates a subtle but profound delayed second order effect : it erodes the line between the LPs and the LDs. The LP desires single sided staking (no involuntary exposure) and a clean transfer of risk from the LP to the LD in exchange for liquidity management expertise. The user, as a result of staking $toke, is now insuring themselves against IL and has forfeited compounding single token rewards for a new kind of two token exposure. The user now has the responsibility to contribute to liquidity direction, diluting the fundamental separation of labor that underpins the protocol’s design. Additionally, since the underlying pool structure has not changed, both the underlying pool of abc / eth and toke / eth pairs are now vectors for arbitrage and the creation of IL in the liquidity value chain.
The protocols who deposit capital inefficient reserve funds face a slightly different issue. For any treasury assets that are not part of the reserve fund, the effect can be evaluated just like any other user. However, in order to open a reactor for their token, protocols need a reserve of ⅓ of the assets in the pool. This creates a capital productivity inefficiency as that reserve can’t earn yield. In return, the protocol receives involuntary exposure to an equal amount of $toke and forthcoming $toke rewards, further eroding the line between LPs and LDs as protocols have to manage their own liquidity.
The final $toke holder profile, if we exclude the largest $toke holder - Tokemak itself, are destination exchanges that might desire deeper liquidity and stablecoins who might desire more “organic adoption”. In order to perpetuate this narrative, Tokemak has sought to draw parallels to the “Curve Wars” with the “Toke Wars” in their own communications. There is something that leaves a bad taste in my mouth when a pre-product company attempts to create FOMO for their token, especially when it undermines their ability to serve underlying LPs. Should an exchange acquire a sizable amount of $toke, they could direct liquidity to their product even if it doesn’t generate the best yield for LPs. Furthermore, the protocol has even advertised $toke as a “kingmaker” asset in the sense that decentralized stablecoins could purchase $toke to increase “organic” adoption. Yet stablecoins are a terrible reserve pair choice for volatile assets as divergence loss is amplified by one of the assets staying stable. Even though it’s a terrible outcome for LPs, power over “toke” controls their system and LPs have no say if their asset becomes “organically” paired with a stable. As Tokemak says themselves, “he who controls the $toke controls the world”.
The erosion of this line between LPs and LDs calls into question the fundamental assumption that democratized direction of liquidity itself would lead to an optimal destination for the LP’s liquidity to earn a return. This claim is dubious as $toke ownership is open to directors of diverse skills sets - protocols who exchange their reserve fund for $toke, free market participants (exchanges, stablecoins, protocols) that hoard $toke to serve their best interests, and finally average users, who earn all their rewards in $toke regardless of the reactor they are in and have limited market making expertise. The protocol’s construction and assorted rewards programs appear to create a fundamental principal / agent problem from the start and a dilution of the division of labor needed for competent liquidity management.
However, it’s not really accurate to say they are a liquidity manager as the liquidity direction voting process only determines the destination exchange and does not actively manage the liquidity while it’s there. Even if all Liquidity Directors were professional market makers, the protocol appears architecturally unable to engage in the informed and agile liquidity management necessary for a successful liquidity provisioning result. The average opinion of a hodgepodge of non-professional users and protocols with no experience in market making as well as self-interested market participants holding $toke is unlikely to lead to decisions that maximize yield and minimize IL for LPs.
The problem could be amplified if Tokemak chooses to integrate with UNI V3 as suggested by their documentation. If we can concede from earlier that Uniswap V3 is a professional’s playground and the cohort of $toke holders due to holder composition and architectural limitations is incapable of producing a good result for LPs, Tokemak might be forced to supply their liquidity to an intermediate liquidity manager/optimizer like Charm, Visr or Lixir (appendix). Now the fee stack could be 10% to Tokemak - the liquidity director, 10% to the liquidity manager and 16.6% to the underlying DEX which normally takes a 5 bip fee on .3% trading fees. None of these services have proven without a doubt that they can consistently remove risk and outpace IL for LPs, justifying their rake. As another participant in this form of a liquidity value chain, it is especially important to note that tokens like $sushi and $uni (if it too were to flip on a fee switch) provide no risk reduction to the LP in exchange for a rake (relevant also to the cohort of multichain copycats) . As there is much talk that liquidity is a commodity and we could see a price war eventually between DEXes, especially if more volume goes through aggregators, is it possible that a bloated fee stack like this could be a competitive disadvantage in staying price competitive?
In summary, the construction of Tokemak and associated rewards distribution undermines itself in one of two ways. Either sell pressure immediately undermines the protocol’s ability to underwrite IL with the value embedded in $toke. Or, should the LP restake $toke rewards, the protocol slowly undermines itself as the line between LPs and LDs is diluted creating a principal agent problem that compounds the already present risk given architectural limitations to active management of liquidity positions. When all is stripped bare of branding, the Tokemak protocol is a set of simple staking contracts governed by non-professionals whose output is creating a liquidity sandwich once a week. Put another way, it is the slowest and most costly autocompunder in existence.
Thanks for reading! If you want to support or contribute to our mission to reduce noise and create thoughtful DeFi research, please follow us on twitter @analystdao and join us in Discord @ https://discord.gg/eAkrN4XN.
Follow @patfitzgerald01 to get notified when Part 3 drops.
**
**