# Introduction

The BlockScience team ran 4800 monte carlo simulations across a wide variety of both parameter constellations as well as behavioral signals to stress test aspects of the lending program. Below is the distillation of the much larger report that can be found [here](https://gitlab.com/thorchain/misc/cadcad-thorchain/-/blob/main/research/20230731%20Report.ipynb?ref_type=heads).

# KPIs

The following KPIs were utilized in evaluation of the system success for different simulations.

## Circuit Breaker KPI

The circuit breaker KPI is marked as true if at no point during the simulation there is a period where the circuitbreaker is hit.

## Liability Management KPI

The liability management KPI finds all the period liabilities (based on what value of RUNE would be if a loan were to close minus the cost to close) and determines if at any point during the simulation this value is 110% of the burned RUNE available. If the swap fees came in at 10% or less, the circuit breaker would be triggered from these loans closing. Of course it is unlikely all closes at once, but it is meant to be a flag for potential danger especially if conditions worsen.

A value of 1 means that at no point were liabilities at this level or greater.

## Minimum Haircut KPI

The minimum haircut KPI is meant to represent the importance of the borrower stakeholders. It finds the average of the collateral amount received on loan close divided by the collateral amount given at loan initiation and tests whether this number is less than 10%.

A value of 1 means that on the average borrowers are taking no less than a 10% haircut on the collateral they put into the system.

## Net Burn KPI

The net burn KPI measures whether or not the inflation KPI indicator is negative at the end of the simulation timeframe. It is important to note that this version does NOT include the liability on the other side from unclosed loans.

A value of 1 means there was a negative inflation value.

## Net Rune Burn w/ Liability

This KPI is the same as the net burn KPI except it includes the current liability.

A value of 1 means there is a negative value for inflation + current liability on loans.

## Liability Management 2

This KPI takes the original liability management KPI a step further. It checks if at any point the net burned rune plus the liability were ever greater than 2/3 of the starting net burned rune.

A value of 1 means that at no point was the liability + net burned rune greater than or equal to 2/3 of the starting burned rune.

## Success Criteria 1

The first success criteria is the less conservative one which is a success if all the following KPIs are also a success:

1. Circuit Breaker KPI
2. Liability Management KPI
3. Minimum Haircut KPI
4. Net Burn KPI

## Success Criteria 2

The second success cirteria is a success if all the following KPIs are also a success:

1. Circuit Breaker KPI
2. Liability Management KPI
3. Minimum Haircut KPI
4. Net Burn KPI
5. Net Rune Burn w/ Liability
6. Liability Management 2

# Conclusion 1: The Influence of Behaviors & Price Changes

From a KPI perspective, the biggest driver of success or failure comes from outside influences on the system. But it is not just the directionality of signals such as price changes, but more importantly the volatility of the signals that drives much of the impacts.

It is important to remember that the loans share some similar behaviors with an option, including one-sided risk (there is a finite amount of deflation that could happen from a loan open, but an infinite amount of inflation based on how collateral outperforms RUNE). Thus volatility has a negative correlation with system success in many ways.

As well, worst case scenarios or black swan events are very hard to prepare for. Put simply, if collateral goes 10X up and RUNE goes to 1/3 its starting value over a period, the lending lever needed to contain that doomsday loss is so small it would be unreasonable.

To summarize, outside factors will have a very large impact on system success, and specifically the volatility of outside factors such as price changes will have large impacts.

# Conclusion 2: Derived Pool Parameter Tradeoff

There is an interesting tradeoff that occurs for the derived pool parameters. The more conservative versions favor most of the KPIs and have a large influence on them, but the minimum haircut KPI has a negative correlation to more conservative parameters because it means that more often the implied fees are higher.

The maximum anchor slip is the parameter that is important to discuss in this context. When it is lower, the haircuts that borrowers take become larger. When it is higher, there is a much less pronounced steepness with which haircuts begin to become large.

One can view these parameters as the counter-balance to much of the risk that the lending program brings. Similar to insurance policies, premiums are taken out on each trade in anticipation of potential loan closes that go against the lending protocol's deflation.

Because of the both the positive and the negative influences on the KPIs for derived pool parameters, decisions on the level of conservatism to use must be preceded by defining the importance to assign to each KPI.

# Conclusion 3: Impacts of the Lending Lever and Collateralization Ratio

The lending lever can be seen to amplify total volatility for both the inflation and inflation with liability metric. As it becomes larger and relaxes more the ability to open loans, the volatility of simulation inflation/liabilities grows with it.

The collateralization ratio has effects that might be of interest for more studying. Overall, however, the likely explanation is that in these scenarios, the higher maximum collateralization ratio is leading to a less volatile but higher liability.

The lower volatility makes sense because as the "strike price" on closing the loan goes down, it makes the liability more and more often in the money. It begins to approach modeling exactly price risk without the optionality component.

The higher overall liability makes sense because the average price increase in these simulations is positive. A higher collateralization ratio means there is a larger overall exposure to increase.

# Parameter Recommendations

| Parameter | Value |
| -------- | -------- |
| Lending Lever| 1/4-1/3 |
|Minimum Collateralization Ratio | 2 |
| Maximum Collateralization Ratio| 3-5 |
|Derived Depth bp| 10000 |
|Max Anchor Blocks| 300 |
| Max Anchor Slip| 6000 |
| Derived Minimum Depth| 1000 |

Overall, the parameter recommendations are meant to balance different stakeholder groups in the ecosystem. The lending lever provides a range where larger dislocations between collateral price appreciation and RUNE price appreciation won't be catastrophic because the system is not highly leveraged. The maximum collateralization ratio can be set higher or lower depending on preferences... a lower one provides more stability against price movements (the parameter exploration notebook details why this may be), but a higher one provides more stability in reducing demand when the program is near capacity. The derived depth parameters came out of discussions about how to ensure high enough haircuts were applied on collateral to avoid bank run scenarios + earn premiums for risk while still allowing for users to receive a fair amount of their collateral back during normal times.

All of these recommendations are meant to be just that, recommendations. In any cases where the general community or 9R team disagrees, it may be sensible to ignore in favor of other parameter values. All of these parameters hold within their own tradeoffs... on a pure risk perspective, having a lending lever close to 0 would reduce almost all risk, but that ignores the very real consideration of program success and revenues that a higher lending lever allows for.