Price Estimation Algorithm

We are dedicated to providing our community with the tools and resources they need to navigate the Open Metaverse seamlessly. One of a core goal of THE HUB DAO is to establish a decentralized Metaverse state by developing various projects and experiences within the most influential virtual worlds. To achieve this vision, we faced the challenge of acquiring fairly valued LAND parcels, which led us to develop a highly accurate, AI driven price estimation model for Metaverse Assets.

Methodology: Harnessing Data and Machine Learning

Our price estimation model is built upon a robust methodology that combines historical sales data and machine learning algorithms to estimate the price of LAND parcels. To ensure accuracy, we retrieve historic LAND sales data from reputable sources such as Opensea's and Etherscan's APIs. This data is then used to calculate the daily median of historic sales, which serves as a benchmark for the model's accuracy.
The machine learning algorithm is trained using the historical price movements of the asset in question. By analyzing and learning from past data, the algorithm predicts the daily median of LAND sales on t+1, which serves as the reference point for valuing individual parcels.

Neighboring Weighting Algorithm: Enhancing Accuracy and Precision

In the virtual real estate market, just like the physical real estate market, the location of a parcel plays a significant role in its value. To account for this factor, we have implemented a neighboring weighting algorithm that significantly enhances the accuracy of our model. This algorithm analyzes the historic prices of neighboring LANDs surrounding the parcel being evaluated and assigns them weights based on their sales history. The more a neighboring LAND has been sold, the higher its weight and influence on the valuation of the target LAND.
To optimize the weighting process, we trained neural networks using different combinations of neighbors. Through multiple rounds of training, we identified the optimal combination of 11 neighboring parcels that minimizes the median average percentage error (MAPE). This approach ensures that our valuation model captures the local market dynamics and provides accurate and precise valuations for individual LAND parcels.
The model is currently set up to evaluate LANDs in The Sandbox, Decentraland and Somnium Space and we are constantly looking to expand to other metaverses.
For more information please refer to the model's KPIs
MGH Valuation Algorithm KPIs.pdf