Incentive Analysis: Arbitrage & Market Maker Keepers in MakerDAO
by Eden Dhaliwal & Geoff Lefevre
We’re excited about what’s happening in the crypto space right now. Yes, the crypto winter still has a long thaw ahead, but you can really see the community is working hard, learning from past lessons and applying more rigour to developing viable cryptonetworks. There is a next generation of decentralized n-sided networks appearing with greater sophistication due to the integration of more participants, better validated mechanisms and a greater understanding of game theoretic blindspots. With more roles and greater complexity, comes with it the challenge of sufficiently incentivizing actors in the network while optimizing towards economic alignment. And while we all know rewards should be higher in the early going (in order to effectively motivate new actors and overcome switching costs), knowing what it takes to attract new actors in quantitative terms helps in developing a computational model for a network.
To examine what it takes to encourage new agents on a network, we decided to create a simple business case around running a single Arbitrage and Market Maker Keepers agent on MakerDao. We took the view of an independent actor rather than a commercial, scale-minded operator. We estimated fixed and variable costs to be the following:
Fixed Costs:
Hardware = $1000
Development Time = $5000 (25hrs @$200/hr)
Electricity = $300 (Single desktop: 1yr 24/7)
Variable Costs:
Exchange Fees = .02%
Gas Price = $.0003
Methodology
For this work, we used @RISK, a simulation tool for excel. This model was developed as a baseline benchmark and used to produce the case for these Keepers. Lastly, the cost of gas, market deltas, and spreads were modelled statically.
First, we took random samples on Market Orders from Oasis. Using the same time window, we then used Dai/Eth prices to determine the mean and standard deviation of key metrics used in the model to simulate asset prices and market buy/sell orders at each time step. Further research was conducted to identify wallet addresses managed by Keepers, upon which we took more random samples and examined the data to estimate the average number of transactions a Keeper executes in a day.
To model both Market Maker and Arbitrage Keeper opportunities, we used Markov Decision Process (MDP) where at each time step, the Keeper decided to execute a trade based on set requirements and constraints.
The mean output values from the simulations were used to produce a 12 month Net Present Value model for each Keeper. All prices are in Dai.
Arbitrage Keeper Analysis
Here’s a snapshot of how we set up our model:
Which provided us the following outputs:
And the resulting 12 month NPV for Arbitrage Keeper:
Market Maker Keeper
Here’s a snapshot of how we set up our model:
Which provided us the following outputs:
And the resulting 12 month NPV for Market Making Keeper:
Future Extensions
Refinement of this model can be grouped into four categories for better results:
Data Collection
Further sampling is required to clarify the daily transaction throughput of each agent. In addition, more sampling is required to model the market deltas for Market Maker Keeper inventory of ETH (PETH) and Dai.
Optimisation of Keeper strategies
The spreads for each Keeper can be further optimised to increase revenue
Multi-Agent Dynamics
You can create a multi-agent MDP model. In order to execute a trade, each agent needs to set the optimal gas price. This creates a multi-agent auction where agents bid on their gas prices to execute their trades quicker.
Gas Arbitrage
There is evidence that Keepers are loading the chain with empty transactions when the cost of gas is low and cancelling those transactions when the cost of gas goes up, and are therefore able to collect the difference in gas prices between each time step to fund gas costs associated with their profitable transactions.
Findings
Our study revealed the following figures:
Keeper Operations
Average # Transactions / Day = 87.52
Average % Profitable Transactions (Market Maker only) = 0.21
Average # Profitable Transactions / Day (Market Maker only) = 18.38
Arbitrage Keeper Economics (all prices in Dai)
Average Variable Cost / Transaction = 0.52
Average Revenue / Transaction = 1.29
Average Profit / Transaction = 0.77
Market Maker Keeper Economics (all prices in Dai)
Average Variable Cost / Transaction = 0.02
Average Revenue / Transaction = 19.41
Average Profit / Transaction = 19.38
Conclusion
As mentioned, this simulation illustrates a simple business case for running a single Arbitrage and Market Maker Keeper agent on the Maker network. From our simulations, running an Arbitrage Keeper agent implies an average 12-month NPV of $22,778 and an annual IRR of 29%. For the Market Maker Keeper, an agent implies an average 12-month NPV of $125,351 and an annual IRR of 168% — quite compelling. Furthermore, each agent’s return diminishes as it attempts to offer more liquidity (Dai), which means you cannot scale with a single agent and therefore incentivizes an operator to create many agents in order to scale returns.
Again, this is only a simulation and just a static snapshot, but it does provide important insights in terms of what kind of switching/opportunity costs need to be overcome if you are looking to acquire agents that can perform arbitrage and market making activities in your network. Performing this type (comparative) analysis will increasingly become an important part of Token Ecosystem Creation in a world where new network opportunities are proliferating.
