The tools for each game feed into an AI-powered engine that determines which games and strategies to allocate resources to in order to maximize ROI. The engine then acts as a resource allocation strategy, which deploys capital into specific asset generating initiatives. Each cycle is then evaluated, scored and re-inputted for the next iteration of the engine.
The model will include multiple variables involving (1) historical inputs from our operations, (2) metrics from game analytics (users, marketplaces, tokens, etc.), and (3) other macroeconomic factors prevalent in the P2E ecosystem. As the model continues to evolve, each variable would transition from being human-influenced into being fully automated, with proactive adjustments made whenever deemed necessary in feature selection and parameter optimization. Instead of being married to a single paradigm or machine learning approach, decentralized governance allows for any relevant externalities to be included in the formulation of the algorithm, which in turn will continuously be improved upon through proposals and efforts of the DAO. In its final state, the engine will be fully automated with a plethora of heuristic rules and models that are optimized using gradient descent-based algorithms, where multiple versions can automatically dictate fund allocations that will optimize yield. As more iterations are realized, the model continues to learn with the eventual goal of full automation in dictating the operational management of these NFT assets.
Figure 2. Stochastic Gradient Descent was a long standing base algorithm for AI/machine learning, with recent advances in the form of tweaks (i.e. optimizers like Adam, or strategies such as mini-batch sizes) allowing for faster convergence or reduction in error rates.