In preparation for our upcoming launch, we have been testing our economic incentive system using simulation to make sure it is economically viable. It is critical to get the crypto-economics correct, that’s why we’re performing these tests. While we can’t ever be certain everything will play out as expected, testing allows to understand where the boundaries are and is an important plank of our crypto-economic testing framework. In summary we’ve performed the following tests:
- In depth deterministic modelling (spreadsheet based).
- using (simulation approach using AI agents who can bribe each other out-of-band).
- Stress testing simulation techniques (using MATLAB).
From our understanding many projects stop after #1, but we believe much more is required. We previously conducted #2 with Incentivai and are now providing our results under #3.
To demonstrate how important this testing is; we used the results to define changes to the bonding curve formula to remove the ‘number of ’ element. ‘Number of ’ was a proxy for but resulted in some non-intuitive behaviour in some of the simulations. As a result we replaced it with a more concrete metric for our particular situation, the minimum capital requirement, which represents the amount of capital needed to back the covers written.
We also amended our capital levels to create a more viable working system. Our previous capital model was too conservative, held too much capital, and couldn’t handle expected business growth. So we removed excess conservatism, but kept our model more conservative than the existing UK and EU regulations (Solvency II).
In summary, our simulation testing demonstrated sufficient capital ratios under a range of outcomes. The biggest threat to capital sufficiency is growing too fast. This is a good problem to have because there are governance actions that can be taken to alleviate it if necessary.
Overall, this testing gives us a much higher level of confidence that the system will hold up under the scrutiny of real world use and that we have actions available to handle the worst outcomes. A full copy of the report can be found here:
If you want to come and talk to us about our poisson distributions, why we chose the base values for each parameter we tested or about the Nexus Mutual turtle, join our or tweet us .
Published at Tue, 21 May 2019 09:28:33 +0000
