Weekly Gauge #18: Frax Saga Part 2: Bribe Marginal Contribution
Last week, we introduced the major role of FXS incentives within the Curve bribe market and analyzed the evolution of its distribution, from its birth to the bootstrapping phase of the Basepool. Frax is now diversifying its incentives in multiple gauges.
This transition in Frax’s incentive program will ultimately impact the distribution of votes between gauges and voting layers. The increasing complexity to schedule each gauge's incentives according to their capital efficiency might lead the protocol to misalign with its own and Curve’s ecosystem sustainability.
In this article, we will introduce, step by step, a model-agnostic method to predict and quantify the impact of every bribe compared to various other parameters on Frax’s cumulated revenue.
Plot representation of every parameter's contribution to the prediction (not accurate)
In order to anticipate the impact of each agent on the value captured by the protocol, it is possible to calculate the Shapley Value of several parameters responsible for Frax’s capital efficiency. The shapley value of a feature is its average contribution to the outcome of a system - calculating it for every pool would allow to optimize the incentive distribution framework.
Let’s break it down through an example on the FRAX-3CRV pool:
We can take into account every governance layers (veCRV ; vlCVX ; veFXS) and say there are 4 features which rules the distribution of votes on this pool, respectively:
Organic veCRV
Organic vlCVX
Bribe
Delegation
We will calculate the marginal contribution of bribes to the revenue generated by this pool and captured by FRAX.
https://dao.curve.fi/gaugeweight
First, we can look for the total amount of veCRV captured by the gauge, here it is 8.76% of 533M veCRV which corresponds to 47M veCRV. This will be the sum of every feature’s contributions.
https://dune.com/paladin/gauge-weekly
The easiest way to continue our research is to determine how many delegated votes have been directed to this gauge. Here we know that delegation addresses only votes for bribed pools and that the only available bribe for the FRAX-3CRV gauge is on the Convex layer of Votium. So we can calculate 12.1% of 20M vlCVX which equate to 2,42M vlCVX, converted in veCRV this is 13M votes.
https://llama.airforce/#/bribes/rounds/votium/cvx-crv/29
Thanks to the information in the picture above we can affirm that there is no vlCVX directed to this gauge which isn’t incentivized (0 organic vlCVX) - because the total amount of vote in the gauge is equal to the total acquired through the bribe. Moreover, we can calculate the total contribution of bribe excluding the delegated market maker by subtracting the 2.42M from votium.eth to the total vlCVX votes, and multiplying the result by 5.25 to get 30M veCRV.
SinceBecause we know that the total veCRV acquired is 47M, and the sum of the 3 last features is 43M, we can deduce that there are 4M Organic veCRV voting for this pool on Curve governance layer (highly likely to be Frax’s own veCRV).
Thus, the sample of our pool revenue contribution features is :
The next information that we will need is the yield captured by each layer of governance, we can see in the image below the 3 tranches. From left to right we have the $ captured by veCRV, by vlCVX, and by vlCVX delegated (earns FXS emission on top of CVX).
We are now able to calculate the Shapley Value of the bribe feature within our model thanks to the following table :
Author’s calculations
By reiterating this method on each feature and for every Frax base and metapool we should be able to set up an optimized incentive framework and maximize the efficiency for every participant in the ecosystem built on top of Curve Finance.
Eventually, because we have noticed that the cost of votes tends to grow when the value of a bribe increases, Paladin Quest offers the perfect tool to mitigate the risk of overpricing votes and prevent the rate increase, because it is possible to cap the ratio of part of the incentive program, and offer the second part the opportunity to capture uncapped votes.
Moreover, it is possible to set up and manage a blacklist of addresses directly from the front end of the dapp, to prevent double dipping and optimize even more the incentive distribution.