Agent-Based Modelling (ABM) is a technique used a lot these days in Cryptoland. Historically, modelling financial ecosystems was (and still is) essentially performed using glorified spreadsheets. Around the time of the GFC, ABM methods gained popularity as the increasing complexity of financial markets necessitated better approaches to understanding complex economic systems. Using ABM today is considered standard practice in finance and economics.
ABM works by simulating a system by programming the rules of the system, rather than attempting to understand, and program, a complex system’s behaviour ahead of time. Essentially the approach is to establish the starting conditions of a game, then to run the simulation and see what happens. There are three basic ingredients:
- The rules of the game to play; this is the system under examination.
- A set of (generally) profit-seeking players, known as ‘agents’. These are the players of the game. They are programmed with rules that simulate the real-world decisions of people and companies (and, in DeFi, programs) so the way they behave when interacting with the game may be observed.
- The environment in which both the game and players exist. The environment includes all of the information that’s external to the system, like what’s happening in external markets, currency prices and the like. Typically, the environment is just a whole bunch of data that both the game and the agents can access to make decisions.
I see. So… what’s been happening with Bumper’s modelling work?
We’ve been working on modelling Bumper for some time. Everyone has shown some heroic patience. So, we recently asked our modelling team to pull together an interim report for us to try and share some of our early results.
It’s probably not a surprise to learn that we’ve had some setbacks along the way, including the need to change partners for this work midway through. Sans specifics, we are now fortunate to be working with the Swiss Centre for Cryptoeconomics to take things to a conclusion for this phase. We’re looking forward to a long and fruitful relationship with Dr Matthias Hafner and his team.
Just tell me the latest.
Right then, to the good bits.
We have an interim report.
If you don’t want to read that (although, we very much encourage you to), here are some of the key takeaways:
1. We have a working simulation of Bumper.
Woot. All coded up in Python and deployed to a beefy server to munch through lots of data.
2. It works, but we need to tune it.
We are getting a lot of encouraging results, but also some odd and undesirable ones too.
For example, on the very first full run of the simulation, we compared the cost of premiums of our model with Hegic, a decentralised options protocol, and found that Bumper’s protection mechanism is about half the cost of the premiums on Hegic. Yup, you heard that correctly… half the cost!
Having said that though, we need to shape the settings so the results are more consistent under a range of volatility scenarios. So that some outcomes are improved, we anticipate this will require us to start to make some trade-off decisions around what set of scenarios to actually cater for, which settings are the defaults, and what scenarios are supported in the initial releases versus later releases.
Community feedback on this aspect will be important so that we’re tuning things to the community’s liking.
3. What specifically is left to be done?
Testing, and, lots of parameter-setting.
Our objective is to ensure the protocol operates really well under a range of scenarios. This doesn’t just mean different price volatility scenarios, but also monitoring things like the ratios between all four pools, rates of change of liquidity, and differences in expected changes within the pools to what is measured. Dozens of individual parameters need to be configured. As of now, we’re in the process of ensuring we are confident that all the knobs and dials are working correctly and providing appropriate results. We’ve been surprised a couple of times already by what the simulation has produced, and that’s a very good thing.
4. What results can we expect?
Scenario analyses, with the costs of premiums for Takers and the expected yields for Makers.
One of the important considerations at this point is to gain community feedback on the results. Bumper is not magic, and there is a fundamental trade-off between risk and cost. In practical terms for users, this means we collectively must agree how this trade-off should be struck. For a Protection Taker, the question is either:
- For what risks do I want to be covered?
- What kind of price drop over what time period?
- How much am I willing to pay for price protection?
For a Liquidity Maker, the questions become:
- What price risk scenario am I willing to supply liquidity for?
- What is the probability of an untenable price movement?
- 1% chance in 1 year?
- 1% chance in 10 years?
- 1% chance in 100 years?
Some would argue the data that would ordinarily support these kinds of decisions is unreliable, as blockchains have not been around long enough to anticipate everything that might go wrong. But this apparent risk has not stymied the growth of DeFi. Plenty of risks also exist in traditional venues to which DeFi is immune. Unforeseen events affect asset prices all the time, both on- and off-chain. So, the immaturity of data may indeed be a concern, but it’s one that every blockchain project shares. Welcome to the Frontier!
5. What happens then?
Once we have socialised different scenarios and agreed the protocol parameters, we can load them into our configuration files ready for launch. If you’re inclined to read the source code, you’ll notice lots of lines at the beginning that look something like this:
And if raw source code doesn’t do it for you, perhaps Kelly LeBrock will…
Ahem.
But then, once we’re live, this effort doesn’t stop there. Far from it; this will be just the beginning of a literally endless journey of continuously updating, augmenting and experimenting with the system. There’s a laundry list of bits and pieces we want to add in the first half of 2022 (check out a snapshot over at bumper.fi/roadmap). Stay tuned…
About Bumper
Bumper protects the value of your crypto using a radically innovative DeFi protocol. Set the price you want to protect and if the market crashes, your asset will never fall below that price. Importantly, if the market pumps, your asset rises too.
Stay Connected to the Bumper Project:
Join our Telegram — https://t.me/bumperfinance
Follow us on Twitter — https://www.twitter.com/bumperfinance
Join our Discord — https://discord.gg/YyzRws4Ujd
Visit our Website — https://bumper.fi
