A fee adjustment tool using a PID Controller for liquidity spreads and a mean reversion controller for a margin.
Currently, it is still unclear for which areas of fee setting the tool can be used. The tool adjusts fees according to predefined parameters. These parameters primarily determine the speed of fee adjustments. The parameters must be specified by the user, which means that, at present, an understanding of the economic and mathematical model behind the tool is required to configure them.
At this stage, it can be assumed that the tool is best suited for nodes whose fees are generally in a good state but need regular adjustments to ensure optimal liquidity flow and counteract the depletion of channels. A good analogy is the autopilot of an airplane, which assists pilots during normal conditions but is not activated during takeoff or in severe turbulence. Similarly, this tool is not expected to deliver immediate positive results if simply activated on nodes with fees that are currently not market- appropriate. Additionally, it will be necessary to regularly validate the parameters used by the tool, as the algorithm cannot respond to all changes in the network. For new channels, it will also be necessary to manually set fees externally, as the tool does not provide a good initial estimate but instead applies a configured value meant only to prevent depletion.
The tool is also capable of setting inbound discounts. However, this does not eliminate the need for rebalancing in cases of significant demand sinks. The tool can support finding a suitable fee level that simultaneously represents an opportunity for rebalancing.
requirements.txt lists all dependencies
for this Python version. Using Python 3.12.8, you can install the dependencies
with:
pip install -r requirements.txt .
This can be done in a virtual environment or with a docker setup. If you are using a Python version different from 3.12.8, you can try:
pip install -r base.in .
Adjust the feelancer.toml for your needs. Especially config your database in
sqlalchemy.url section. For productive use postgres is recommended.
An example of the configuration file can be found here.
feelancer --config [CONFIG_FILE]
It's hard to say. First, learn how the model works. You can find an xlsx-example
in the doc directory. In my humble opinion, a good first step is to set
- Writing more tests and more documentation. Testing has been more explorative until now.
- The dependencies between the model parameters must be researched, and it must be investigated whether all parameters are required. Perhaps a generally different modeling approach is also preferable.
- Maybe building more analytical tasks.
For development you need to install
pip install -r dev-requirements.txt -e .