The first method an audit should always employ is that of learning from similar cases. When coin offerings were initially invented, there was a high chance that a new project was creating a proposition or mechanics never encountered before. Now, it is likely that a new project will find at least some points of similarity with existing projects.

For example, there is a lot that can be learned from the successes and failures of different types of stablecoins. From the controversy that has surrounded USDT [4], to the success of Terra/Luna or the bank run and the eventual collapse of Iron Titanium [5], it is likely that a project can learn a lot through similarities with existing projects.

The analysis becomes even more useful if data is available which can be used to make an argument. An example of this approach is seen in an audit of the tokenomics of Frax by Albaron Ventures, where they compared the peg stability of Frax and other stablecoins, as shown in Figure 1Figure 1.

Comparison of peg stability of different stablecoins. Source: [6]

One of the methods suggested by the author in the past is agent-based modelling [7]. Agent-based modelling is used to study complex systems and to solve problems that are difficult or impossible to solve analytically.

The agent-based modelling process starts with the identification of the system’s components and their relationships. The next step is to identify the rules governing how these components interact with each other, which can be done by observing the system in operation or by using expert knowledge. The final step is to run simulations with different sets of initial conditions and parameters and compare them against one another in order to find an optimal solution.

The flexibility of agent-based modelling has given rise to different flavours of this approach. For example, some simulations might incorporate intelligent agents, through the use of reinforcement learning, a famous example of this being the multi-armed bandit problem.

In the domain of networks, diffusion models are a popular choice, and they have been used in many different areas from biology [8] to social networks [9].

Also, agent-based modelling is often implemented via Monte Carlo methods, whose objective is to use repeated random sampling in order to identify equilibria and possible evolutionary paths in a model [10], [11].

Agent-based modelling is an excellent method for analysing token economies, given their complexity and the interconnected parts.

Game theory is a branch of mathematics that studies the mathematical models of conflict and cooperation between rational decision-makers, first devised by Von Neumann [12]. Game theory is mainly used in economics, political science, and psychology in order to understand how humans interact with each other when there is limited, or no, trust.

Game theory has an integral significance to blockchain, given that the root of the blockchain lies in how, through the use of algorithms, trust can be ensured in a network without trust. It therefore follows that game theory can be a useful tool for a tokenomics audit.

One of the best examples of this is probably OlympusDAO. While OlympusDAO did not go through a tokenomics audit (it has gone through a smart contracts audit only) [13], the project is well known for its use of game-theoretic analyses to prove the sustainability of its protocol [14]. A famous image depicting the game theory behind this project is shown in Figure 2Figure 2.

Olympus DAO game theory. Source [15]

Another type of analysis, closely related to game theory, is what we will term “structural” analysis but which can also be called “balance-of-forces” analysis. This is a higher level of abstraction, where the analysis lists all possible dynamics and their impact upon a token economy, but without explicitly creating an incentives matrix.

The dynamics and the tools employed create a narrative, which aims to convince an informed reader that they should work as expected when deployed in real life. The balance of the different forces that are applied to a token and an ecosystem should ideally be driven by a clear objective, such as token appreciation, stability, or sustainability.

While this is not explicitly stated, the majority of new blockchain protocols follow this logic, starting with Satoshi Nakamoto’s original bitcoin paper [2]. A more recent example of a successful project outlying this is Terra/Luna’s documentation [16].

However, the aim of a structural analysis for a tokenomics audit should be to formalise the different dynamics at play in a more structured way, and troubleshoot for issues that might arise, thereby highlighting weaknesses.

Another mechanism which is often employed, albeit informally, is the study of marginal cases. These are purely hypothetical scenarios that could break a system, though they never happen in practice. This is similar to the stress test practice which financial institutions undergo. It has become popular for analysts to publish these types of analyses on Twitter, as a series of posts. An example of such an analysis is shown in Figure 3Figure 3.

Example of a popular thread discussing a potential scenario was Terra/Luna crashes. Source: [17]

Some analyses prefer to resort to probability theory and stochastic models, sometimes also combining aspects of some of the previous tools mentioned, like game theory and agent-based modelling. A good example of this approach is the analysis of Bitcoin’s defences against attack vectors [18] and mining pools [19].

This case study presents the way in which the BankX audit was conducted. While the full audit can be found in the references list [20], this case study will present an overview of the thinking processes governing this audit.

The audit was broken down into a series of steps.

BankX is a stablecoin that has been inspired by elements of two other blockchain projects, Hex [19] and Frax [16], while also adding its own mechanisms. Both Hex and Frax seem to have flourished since their inception (both of them were launched in 2020). At the time of writing, Hex is at position 201 of Coinmarketcap [21] and Frax is at position 204.

The goals of the analysis were tied to the goals of the system in a hierarchical manner, with the most important goals at the beginning. The goals are copied and pasted below from the original audit [22].

Once this foundation is established, the audit then proceeds to describe the various mechanisms and how they interact. The audit also contains a full table of the dynamics at play and describes whether they work in an inflationary or deflationary way.

The audit then proceeds to an analysis of the peg stability of different stablecoins, demonstrating through empirical evidence that Frax, at least at the time of the audit, could perform just as well, or even better, than “traditional” stablecoins, such as USDT and USDC. The similarity of BankX’s and Frax’s mechanics is, therefore, considered a positive aspect, which provides evidence that the stablecoin can work as expected and maintain its peg.

The audit then goes on to use game theory in order to analyse BankX’s bonding curve mechanism, before moving on to the marginal cases analyses. One of the marginal cases includes data from the Terra/Luna crash that took place in 2021 (shown in Figure 4Figure 4.

The Terra/Luna crash. Source: [16]

The author declares no conflict of interest.

Not applicable.

Not applicable.

This study was not funded.

Not applicable.