Another new money god: Introducing H2O

TLDR; this post introduces H2O, an authorized or “friendly fork” of RAI, as the newest member of the money god league to join the likes of VOLT. Where RAI seeks stability and collateral in assets like ETH, H2O will be backed by OCEAN and data tokens within the OCEAN ecosystem, the data standard. The (un)governing token of the protocol is AQUA, which will be used to govern the initial parameters of the protocol, bootstrap ecosystem development, and incentivize the liquidity of the protocol. AQUA token emissions will be done fairly, distributing a portion of the tokens to the Reflexer community.

Money God League?

The Money God League is an initiative meant to bring non-pegged stable assets (stablecoins) into the mainstream. The League is also devoted toward making control theory a common framework used in a wide variety of DeFi related projects. Through community coordination, we have explored the multiple paths we can take as a protocol to expand RAIs reach and create new use cases for it.

The time for the money god league is upon us as the Reflexer protocol begins to be able to stand on its own two feet. This week, we have already seen one emerge, VOLT, proposing a multi collateral RAI. We believe the time is now to propose H2O.

Why H2O?

Firstly, we have to understand how data can be used as an asset in DeFi, we can compare it with other seemingly unorthodox DeFi-accessible assets. NFTs are a good example here as they show us how even unproductive assets can be employed in various forms in DeFi through collateralization and fractionalization. There are two ways to integrate data into a decentralized financial system. Either take each dataset as separate fungible assets (e.g., bitcoin and ether) or cluster them under the name datasets —like NFTs. The first option is generally more suitable for financial activities since we are already witnessing how NFT fractionalization projects emerge to integrate NFTs to DeFi. Nonetheless, the importance of data exclusivity will be a significant parameter when deciding if a dataset should be issued as ERC-20 and open to the public or ERC-721.

Data has intrinsic value manifesting through real-world usage, and datasets are less prone to speculative price action due to low penetration to trading markets. Therefore, one possible use case for datasets would be to employ them as collaterals for stablecoins. We hope to manifest a vision of the data standard hand in hand with the Reflexer community and build a unique implementation of the RAI protocol which grows in value as more data begins to flow into DeFi through other protocols like Ocean.

Our initial implementation is the identical codebase as RAI applied to the OCEAN governance token, quickly followed by data tokens. Unlike most other stablecoins, we will be launching with a use-case for the token that promotes natural volume by using H2O as the native medium-of-exchange in the Ocean V4 data marketplace.

Our protocol is a proud friendly fork working cooperatively with the Reflexer team for almost half a year on this launch. To help ensure the Reflexer community feels the benefits of this expansion of Reflexer, we will allocate 7% of total AQUA supply to the Reflexer community.

The distribution is planned to occur over 18 months via staking contracts. Subject to change based on the Reflexers future liquidity plans, we will initially be distributing2.45% to single sided FLX staking, 2.45% to FLX/ETH Uniswap V2 LP, 2.1% to RAI Curve LP

An initial analysis

We did some simulation analysis using a cadCAD model of the Reflexer RAI system to understand its performance under several “shock” scenarios and under different parameter choices.

30% OCEAN price drop - Sets of PID controller parameters :

Kp Ki alpha
Set 1 2e-7 5e-9 0
Set 2 5e-14
Set 3 0
Set 4 5e-8 5e-9
Set 5 5e-14
Set 6 0

Note: alpha is a so-called leaky integral parameter (a parameter to tweak the integral error)

Results:

As we can see clearly with ki=5e-9, the system reacts too aggressively to the price shock and sets the target price too low.

The sets with Ki = 5e-14 or 0 sets the system quite stable in this shock scenario

Next, the same sets of PID controller parameters have been examined again with 20% price shock.

Results:

The parameters sets actually behave quite similarly in case of 20% price drop

Clearly the choice of parameters is very important to the stable functioning of the system.

Choosing Ki:

  • According to previous work from BlockScience team, if the system includes Ki, then it has to have alpha, a parameter to tweak the integral error term of the PID controller.

  • The reason is that in the scenario of a malicious whale buys up a large portion of the RAI supply and uses it to forcibly hold the market price of RAI at a constant.

  • In case of the absence of alpha with negative Ki, redemption price explodes in the presence of an attacking whale, making this a clearly unacceptable parameter choice.

Shock simulations with Ki and alpha

  • The controller is examined again with sweeping parameter alpha, in case of 30% impulse shock
Kp Ki alpha
Set 1 5e-8 1e-10 0.001
Set 2 0.999

Results :

Clearly that PID controller higher alpha adjust the target price more aggressively in this shock scenario.

So introducing Ki in the controller would make it more complicated to tune because in addition of choosing Ki itself, we have to choose its leaky factor alpha.

Conclusion on controller parameters:

Based on the above arguments and the previous works of BlockScience team, here are the conclusions on choosing the controller parameters

  • P controller was the simplest and safest configuration for launching the network and we can always update params post launch.
  • However, since proportional controllers are known to suffer from steady state errors, an integral term may also be desirable to include.
  • But it has to have the leak term which satisfies the condition:
    Kp > -Ki /(1 — 𝛼)
    Where alpha is a leaky integral parameter.

Who is behind H2O?

The H2O project will be overseen and executed by the New Order DAO. New Order is a new permissionless incubation DAO, generating long-term value by building and launching early-stage financial protocols, applications, tools, and infrastructure. The DAO will support and launch projects across all layer one ecosystems and explore DeFi markets surrounding new digital asset classes, including NFTs, data tokens and more. With keeping all protocols best interest at heart, we will be working directly with the Ocean, Reflexer and our own community to develop and sustain operations for this project.

Wen launch?

Our team has been working closely with the Reflexer team for the last few months to make sure we can give you a smooth and safe launch. We are currently in the final stages of the initial deployment and already looking towards the gigabrain data-token backing with our token engineers. To achieve its initial use-case of being the official currency of the Ocean data marketplace, we want to line up the official mainnet deployment with the release of the Ocean V4 upgrade, while V4 does not have an official date, we expect it to arrive early Q1 2021.

As the protocols rollout is still in its infancy, there is not yet an official website or gitbook up for the project. We have a discord for our DAO where we will be coordinating the final stages of deployment with our community and funnelling in new core contributors for future developments and protocol maintenance.

4 Likes

Thank you for the post frens, love the technical details for the controller! :heart_eyes:

2 Likes

Please be mindful that the diagrams and statements in sections titled “Results” are in fact Hypotheses. These models are not anything like physical models of systems with known laws of physics. I think the actor models in blockscience’s cadCAD model are likely flawed and possibly fatally flawed, even one small wrong assumption can make the emergent behavior totally different. It is questionable if PID is even the right mental framework, these are not physical processes with polynomial relationships. The actor models should be built from first principles and then also shown to actually reflect reality before you try using them to inform your decisions about anything.

One way to interpret this is “just try it”, that’s the point of having many of these systems. Excited to see what happens.

1 Like

Totally agree.
All models are wrong, some are useful, some are not :slight_smile: