Forecasting Cardano Native Tokens

We propose the development of an automated valuation method for native tokens based on tools of machine intelligence for complex systems.

Applications of reinforcement learning to improve actions (e.g. trading) in a complex environment (e.g. market for native tokens).

Since March 1, 2021, the Cardano blockchain has provided support for native custom tokens, which benefit from the same infrastructure and security features as the ADA cryptocurrency. In addition, by sidestepping the requirement of smart contract coding (in contrast with non-native tokens that run on Ethereum) and by being naturally coupled to work on the Cardano blockchain, Cardano native tokens are more efficient, error-prone and offer reductions in transaction costs.

However, the valuation of Cardano native tokens remains a subtle problem, given the fact that they are complex assets which may be backed up on or stand in representation of other underlying assets, services or goods as defined by the token issuer. A further complication in the valuation or forecasting of native tokens ensues from their limited circulation relative to the ADA cryptocurrency. 

It is a well known fact that financial markets are complex systems with interesting macroscopic behavior, such non-trivial time correlations, scale invariance and heavy-tailed distributions. Given the myriad non-trivial dependencies and correlations on underlying factors inherent to the market of native tokens, their valuation is not a straightforward matter and may be similarly amenable to understanding from the point of view of complex systems and time series analysis.

Given the spectacular successes of machine learning methods over the past decade for pattern detection and prediction in all kinds of practical and academic applications, we propose the development of machine intelligence systems, based on reinforcement learning methods for complex systems, with the purpose of the valuation and forecasting of Cardano native tokens. This machine learning application promises to reflect the real value of these assets based on the processing of information that is publicly available and relevant to the valuation of said assets, thereby providing a powerful tool for the community, whereby fair assessments of these tokens can be produced with a rigorous and systematic methodology.

The success of this project consists in the development of a practical, user-friendly machine intelligence system for Cardano native token valuation with an accurate performance as a value predictor of said assets.

Impact:

Nelson has met with Singularity.Net to discuss hosting the completed algorithms.

The reinforcement algorithms will provide the Catalyst community with tools to assist in the valuation of Cardano Native Assets. The F7 award will complete gathering of native asset data and correlated signals for prototyping of the machine learning algorithms. A F8 proposal will be submitted to evaluate the performance of the prototype. A F9 proposal will be submitted to demonstrate the tool with users.

Feasibility:

Team Capability: Nelson leads Photrek which has expertise in the development of machine intelligence algorithms for decision-making in complex environments. Zubillaga is an expert in modeling financial systems and is currently developing reinforcement algorithms. He will design the reinforcement algorithms for control of trading bots.

Roadmap:

A six-month plan with three 2-month phases.

• Phase 1: Research native tokens launches. Who is planning to launch what and when?
• Phase 2: Research price of existing native tokens and what are the primary influences of those prices.
• Phase 3:  Design the machine learning algorithm.

Total Budget: 9900 USD.

Each phase will be budgeted at 3300.

The project allocations will be: 50% Applied Research, 25% Data Analytics, 25% Reporting to Catalyst and Singularity Communities.

Auditability:

Success will be identification of a set of Native Tokens with significant potential future value and a set of features of the Cardano Ecosystem that are relevant for training forecasting algorithms.

Key Metrics:

• Number of native tokens analyzed.
• Number of features identified for forecasting.
• Number of engagements with community users.