[FIP - 103] Development of Liquidity Risk Profile Dashboard for Frax Finance

Summary

Development of Liquidity Risk Profile Dashboard for Frax Finance
This dashboard includes:

  • LCR (Liquidity Coverage Ratio) 1day, 1w, 1mo etc.* LVaR (Liquidity-adjusted VaR)
  • Quick and Current Ratios, capital position and collateral ratios
  • Structural liquidity metrics and stickiness (Tenor and behavior of collateral investors and redemption activity)
  • Liquidity exposure on Ethereum and Fantom by protocol (covering 85+% of the existing liquidity, future upgrades could include other chains and protocols)
  • Exposures across all liquidity pools by crypto asset such as USDT, USDC, DAI, and ETH
  • Structural correlation analysis between liquidity pools
  • Develop contingency funding plan and measures under various pre-identified liquidity stress scenarios
  • Scenario Analysis: Withdrawal of deposits, loss of exchange liquidity and volumes

Development Cost: $80,000 USDC
Ongoing Cost: $1,500 USD per month (Credmark fees)

Proposal

We propose to develop DeFi risk models and build summary risk dashboards quantifying Frax risk exposures and measures. While there are a few dashboards on activity and summary analytics of Frax, there are currently no risk models or risk frameworks that exist for any stablecoin, anywhere. These models and frameworks we propose to create will quantify all the various risk exposures so they can be properly identified, managed, and mitigated and open source to give further validation and confidence in Frax.

The analysis and on-chain risk modeling will be implemented with Credmark protocol. Credmark is a full stack Defi financial modeling platform that runs natively in Python which allows us to move beyond simple analytics and develop sophisticated forward looking financial risk models. We run our own database of indexed Ethereum data, python based dev-tooling, and cloud infrastructure for building and executing complex financial models. This enables DAOs and financial institutions to use comprehensive and accurate on-chain financial metrics.

The Credmark team has published numerous research reports including a detailed post-mortem of the Terra ecosystem collapse Additionally, Debt DAO leverages Credmark’s technology and expertise in order to evaluate counterparties cash flow and treasury in order to determine appropriate loan terms. One such loan proposal was recently passed by the Frax community via governance vote.

Overview

We will develop comprehensive beginning to end construction, tooling, and ongoing risk modeling for the Frax ecosystem, setting the standard for other stablecoins, and drastically reducing emerging regulatory risk exposure.

The skills and backgrounds necessary to build proper DeFi risk models and risk management framework requires the intersection of many fields including: financial risk modeling, on-chain analytics, technology development, treasury function including asset liability duration management, economic capital modeling and capital risk bucketing, liquidity risk identification and modeling, scenario analysis and stress testing, investment and collateral risk, monte carlo simulation, and sentiment analysis. The team has extensive experience in all of these areas and is best in class in the intersection of financial risk management and blockchain technology development.

Here is an example Frax analytics and trading dashboard recently completed with Credmark:
Example (Frax/FXS/FPIS distribution and inflows/outflows).

Functional Collateralization Risk Dashboard Metrics

The Frax Collateralization Metrics uses Credmark on-chain data and risk modeling framework. The dashboard below summarizes the collateralization metrics and runs a VaR model which estimates the probability of Frax being able to fully meet its liabilities at any point in time for any forward looking horizon.

The initial implementation also uses simple stress tests which quantify if there is less than a 1% chance of Frax able to fully meet its liability over a 1 and 3 month time horizons assuming no changes in liability or collateral ratio (which are big assumptions) but can be refined by incorporating a scenario analysis framework in the future. Some of the key risk measures are the overall Frax Collateralization, the Max % Safe Price drop of FXS, and the VaR stress tests which quantify a 99% confidence for the maximum likely price drop of FXS to meet all Frax liabilities over a set time horizon. For these measures, FXS market cap would cover the non-collateralized portion of Frax as part of Overall Frax collateralization. As below the 90day Horizon stress test fails because the VaR of -88% is greater than the max Safe Price drop of -76.6%.

As a comparison and validation, the collateralization metrics and VaR models as of Apr 20, 2022 (before the Terra UST collapse) are detailed below to determine how the risk exposures have changed.

As can be seen from the collateralization metrics comparing July to April, there was a significant deterioration in the Overall Frax Collateralization from 175% to 124%, and the Frax treasury as well which was dominated by FXS. The other change is the decrease in Max % Safe Price drop of FXS from 85% to 76%, and the massive increase in VaR in July driven by the 85% price drop of FXS from April into May. In response the collateral ratio was also increased.

In summary overall risk exposures have increased, although Frax is still in a very healthy risk position with many levers to manage risk.

Another potential area of further risk analysis is the high Beta of FXS relative to the Frax market cap (overall liability) and the high percentage of Frax treasury dominated in FXS. The overall Frax market capitalization declined from $2.6b to $1.4b which is almost a 50% drop, yet the FXS price declined from $35 to $6 which is about an 80% drop, indicating a Beta of around 1.6.

As part of a contingency funding plan framework for market stress scenarios that might be needed to cover the liability, diversifying some portion of treasury into other assets that are not correlated to the Frax liability (market cap) would strengthen the overall risk profile of Frax if it’s ever needed to intervene during times of market stress.

This sort of risk modeling framework can help guide decisions for the protocol and manage and monitor risk exposures. As an example, it could help determine the Required Collateral Ratio by setting risk tolerance limits and having a data driven approach to strategic decisions and managing overall risk exposures depending on market conditions and risk premia. This is especially important in Crypto given the incredibly high volatility, and the recent blowups of other stable coin chains and protocols which were not implementing proper risk modeling and management. The urgency of implementing sound risk modeling frameworks across DeFi protocols is critical.

This is only a proof of concept dashboard just for collateralization. Our proposed risk dashboards would include many more risk models and measures.

Proposed Engagement Activities and Services

We propose to develop custom risk models and summary risk dashboards for Frax. These will quantify and model various risk exposures to allow for management and risk mitigation.

A summary of the risk dashboards and example metrics along with detailed descriptions follows. These measures are quantified as point in time with customizable lookbacks (1hr, 1day, 1w, 1mo, 1q etc.) where appropriate, and will provide time series charting for the metrics. All the reports and dashboards will be designed to be persistent with little to no ongoing maintenance and allow user customization and risk reporting. Additional requested risk measures can be added interactively as requested.

Detailed Risk Dashboards & Example Risk Metrics

Liquidity Risk Profile Dashboard

  • LCR (Liquidity Coverage Ratio) 1day, 1w, 1mo etc.
  • LVaR (Liquidity-adjusted VaR)
  • Quick and Current Ratios, capital position and collateral ratios
  • Structural liquidity metrics and stickiness (Tenor and behavior of collateral investors and redemption activity)
  • Liquidity exposure on Ethereum and Fantom by protocol (covering 85+% of the existing liquidity, future upgrades could include other chains and protocols)
  • Exposures across all liquidity pools by crypto asset such as USDT, USDC, DAI, and ETH
  • Structural correlation analysis between liquidity pools
  • Develop contingency funding plan and measures under various pre-identified liquidity stress scenarios
  • Scenario Analysis: Withdrawal of deposits, loss of exchange liquidity and volumes

Development Cost: $80,000 USDC
Ongoing Cost: $1,500 USD per month (Credmark fees)

Asset Liability Mismatch Dashboard

  • Summary of Assets vs Liability exposures & leverage ratios
  • Long Term Funding Ratio (LTFR) (Contractual maturity mismatch)
  • Term structure of funding assets and liabilities & duration gaps
  • Average asset duration, average liability tenor
  • Surplus funding capacity

Collateralization Risk Dashboard

  • Collateralization measures
  • VaR models
  • Counterparty exposures and measures
  • AMO risk exposures
  • De-Centralization ratio
  • Pool exposures and trading activity

Scenario Analysis & Stress Test Dashboard

  • Identification of several stress test scenarios and modeling their impact on risk measures and exposures
  • Ability to tweak and customize the scenarios and run user defined stress testing
  • De-pegging scenario analysis impact across all aggregated Frax Liquidity pools
  • Analogous UST like de-degging scenario
  • Individual & aggregate Liquidity pool stress testing
  • Funding withdrawal redemptions
  • Tolerance Stress tests - Identification of limits that would de-peg Frax and how far away those are from current market conditions.
  • In all of these scenarios, quantification of how all the Liquidity, ALM and other risk metrics change

All the tooling will be persistent, open source, and adjustable by the Frax community.

Additional Detail

The goal of liquidity stress testing is to analyze if Frax can withstand unexpected market disruptions.

A contingency funding plan (CFP) is needed to address liquidity needs under stress and incorporate quantitative information generated during the liquidity stress testing process.

Work with Frax governance and develop customized, forward-looking scenarios to accurately reflect real market conditions, and incorporate custom financial, behavioral, and economic variables according to the funding and liquidity profile.

Surplus funding capacity: Amount of capacity that exists after taking into account the headroom required to survive a stress event (whether specific to Frax or market-wide).

Accelerated construction of the technology enabling first to market development of a fractional-algorithmic risk management framework.

A holistic real-time dashboard quantifying and monitoring potential risks of Frax to help manage and mitigate risk exposures and ancillary factors.

Risk reports for the market and users of Frax. Potentially develop new markets to mitigate emerging exposures for Frax that are currently unable or hard to mitigate and hedge.

Analogous Modeling

A good, but incomplete analogy to Frax stable coin is fractional reserve banking. How much and what kind of economic capital and asset liability management do banks need to perform to prevent a run on the bank? Although not a perfect analogy as there are obviously different types of risks when comparing fractional reserve lending to fractional-algorithmic stable coins built on blockchain technology, there are significant overlaps.

The team in aggregate has combined decades of experience building and running quantitative financial risk models in traditional finance, and we will use many of the tools, modeling techniques, and insights from that background.

Regulatory risk is a new emerging and growing concern especially since the collapse of UST. By employing similar first principles model frameworks used in fractional reserve banking according to Basel III and other regulatory frameworks, Frax could also be the leader and ahead of any competition if there is eventually a regulatory crackdown on stable coins. By implementing a thorough risk management framework that is held to as high of or even a higher standard than traditional fractional reserve banking and regulators have familiarity with, and also open sourced to be validated by the Frax community.

Process and Capabilities

  • Identify and quantify risks and exposures
  • Breakdown the risks from quantitative to qualitative
  • Develop models for quantitative risks and exposures, rank and order the qualitative and operational risks
  • Develop a comprehensive scenario analysis framework including but not limited to simulating transaction and gas fee increases, decreases in liquidity by the protocol and pool level, and collateral risks
  • Identification of market risks and sentiment risk factor that could cause a degradation of the confidence of the un-collateralized portion of Frax and cause a potential de-pegging event
  • Simulation tooling allowing users to run their own custom historical, pre-defined, or monte carlo simulations
  • Quantification and modeling of a Frax reserve and when to intervene with the reserve depending on market conditions, volatility, and Frax confidence
  • Quantification and understanding of how the risks change as Frax scales in size and is incorporated into new protocols and exchanges including potential market impact modeling
  • Develop real time identification and monitoring of potential de-pegging threats and events through on chain analytics
  • Quantify the market for Frax according to different risk levels such as normal, medium, high and critical and have scenarios prepared and identified ahead of time to mitigate any potential de-pegging events; documenting risk mitigation steps ahead of time for each scenario
  • Emerging threat identification and risk mitigation strategies such as a market collusion for a run on the peg event
  • Catastrophic risk Insurance: Could de-pegging catastrophe insurance be part of the solution and risk mitigation strategy (Potentially even creating the catastrophe insurance market itself)
  • Systemic and event risk identification.
  • Liquidizing and market impact modeling - As Frax scales to new exchanges or is integrated in new protocols, quantify the delta and potential market risk impact changes.

Team Background and Qualifications

David Fake, Chief Data Scientist

Experience

  • Data Scientist in Quantitative Finance, 14+ years in Investment Management
  • Masters of Financial Mathematics, University of Minnesota
  • Three bachelor degrees in Math, Physics & Statistics, University of Minnesota
  • Certified FRM (Financial Risk Manager)
  • Passed Level II of the CFA Program
  • Adjunct Lecturer in Experiential Learning Class for Master of Finance program at Carlson School of Management

Work History

  • Chief Data Scientist - Irulast
  • Data Scientist Advanced Analytics, Global Research: $510 billion AUM - Columbia Threadneedle Investments
  • Director Investment Risk $250 billion of Fixed Income Portfolios - Columbia Threadneedle Investments
  • Equity Quantitative Portfolio Analyst - $10b Global High Dividend Yield Equity Strategy - Columbia Threadneedle Investments
  • Quantitative Analyst - RBC Global Asset Management
  • Quantitative Analyst - Galliard Capital Management

Kunlun Yang, Senior Quantitative Analyst

Experience

  • 14 years experience as a Senior Quantitative Analyst covering risk modeling, derivative pricing, quantitative trading and system development.

Work History

  • Credmark
  • Standard Chartered Bank
  • Exxon Mobil
  • Cargill
  • Louis Dreyfus

Credentials

  • BEng (Mechanical),
  • PhD (Structural Biology with Molecular Simulation)

Grant Cermak, Senior Blockchain Developer

Experience

  • 20 years experience developing software and leading blockchain based technology companies
  • 7 years experience developing smart contracts on Chia, Ethereum, Terra, and EOS
  • Bachelor of Science, Computer Science; Mathematics Minor, University of Minnesota
  • Graduate Studies in Artificial Intelligence and Computer Security, University of Minnesota

Work History

  • CTO, Otomo.ai
  • CSO, Exodus
  • CISO, Nucleus Health

Adam J. Weigold, Senior Blockchain Developer

Experience

  • 20 years experience developing software and leading blockchain based technology companies
  • 8 years experience using blockchain technology, DeFi, and smart contracts on Ethereum, Bitcoin, Chia, and Terra
  • Extensive experience owning and maintaining open source software

Work History

  • CTO, Irulast
  • Platform Architect, Watermark Insights
  • VP Engineering & Chief Architect, TrustedChoice.com
  • Senior Software Engineer, Vital Images
4 Likes

Hi there @grantosan! This is very timely. We are looking into doing similar but with a focus on understanding the collateralization ratio, the growth factor, and how they’re affected by the market. In other words, we’re trying to understand how the Frax system reacts to different pressures given that there are inherent reflexive properties.

Are you interested in collaborating? Feel free to reach out to me on twitter @cinjoncin. I’m a CS PhD in ML and MIT math; my collaborator is MIT math and Google eng.

1 Like

Hi there, this is interesting, and I think each dashboard can be a good tool for the community. But I did not see any pricing or scope of work for the first phase. Could you please a little bit elaborate on that?

The first phase is the Liquidity Risk Profile Dashboard. I anticipate that this could be done in roughly 2-3 months for $80k USDC.

1 Like

I followed you on Twitter. Please follow me back so that we can DM. My Twitter handle is @grantosan.

We have the Liquidity Risk Profile Dashboard functional on Credmark. We’re prepared to demonstrate this to you, Sam, and Travis. Is there anyone else you think should see it in action?

to avoid having a vote every few months for new funding can you please break down the overall costs for this proposal and any ongoing cost to maintain it.

1 Like

I’d prefer that we just stay focused on the first part which is the Liquidity Risk Profile Dashboard with the bullet points listed under it. If the community likes the work I don’t have a problem voting on the next part or the rest in its entirety. We’re also prepared to demo what’s done so far town hall style if that would help.

You must understand that we dont want to pay $80k to build something just to find we are gonna get charged silly amounts to keep it running.

Even if we just keep this proposal about building the dashboard, it will still need someone to maintain the dashboard going forward as things change, so this will have a cost.

i should point out im in favor of building something like this and i have already built something like this for myself.

im also very keen on avoiding risk on a protocol level.

Hi @sparkes25. Credmark core team member here.

Access to the full Credmark API is currently $1500/mo. This allows access to all public models, including the ones proposed.

Note that we’re planning to introduce a token that can be used to provide discounted access. Staking a large enough amount of the token will potentially cover monthly fees thereby providing “free” access while supporting our project.

OK, FRAX pays you to do the work of building it, then your selling this to the public afterwards.

So will we be able to pay the $1500 a month and then just put all the data on the FRAX app?

If we do this, then all the information will be free to the public.

Once we’ve delivered this dashboard it will be up to the Frax community to determine the visibility. Probably similar to the way that other dashboards are controlled. The code sits in the public Github repository and hosting is done the same way that existing dashboards work. The only difference here is that this dashboard is dependent on a datastream that lives in Credmark.

I dont think FRAX gets to control the visibility as anyone whos paid the $1500 would have access to the data source it seems.

so overall the ongoing cost will be $1.5k a month after its been built.

Hi @sparkes25. You’re correct. If the models are public, any Credmark customer will be able to run them. This is by design. Modelers have the ability to leverage existing models, many of which have been commissioned. I come from the traditional open source world, where this is very common. I’ve sponsored many projects that others benefited from and I’ve benefited from a lot of corporate sponsorship of open source software.

Having said that, we recognize that sometimes people want to keep models private and we can accommodate that by bringing up dedicated infrastructure. This is more expensive ($5K more per month) because we have to replicate a lot of internal services.

Hope that helps.

I’ve always been in favor of having the option to not contribute back to the community (the old MIT/Berkeley vs. GNU licensing issue). In open source

This proposal is up for voting here: Snapshot