Sókrates Wiki

Icelandic Real Estate Residual Model (Spalíkan)

3 min read

Icelandic Real Estate Residual Model (Spalíkan)

Summary

The Spalíkan is a sophisticated five-layer predictive model designed to analyze the Icelandic residential real estate market using the complete transaction history since the year 2000. Developed as a high-visibility demonstration of the Sokrates autoresearch framework, the model decomposes property value into structural and controllable components to optimize net value for sellers.

Details

The model is trained on a comprehensive dataset comprising over 100,000 property listings and every kaupsamningur (purchase agreement) registered in Iceland since 2000. This dataset allows the model to achieve a resolution in pricing trends and value drivers that exceeds human brokerage capabilities.

Architecture and Methodology

The Spalíkan utilizes a five-layer decomposition strategy. Rather than training on raw price data, each successive layer is trained on the residual value left by the previous layer. This methodology isolates specific signals from structural noise:

  • Layer 1-2 (Structural): Focuses on fundamentals such as location, size, age, and basic property characteristics.
  • Layer 3-5 (Controllable/Presentation): Analyzes variables that a seller or agent can influence, including listing descriptions, photography quality, timing of the listing, and the specific performance residuals of real estate agents.

A key differentiator of the Spalíkan is its optimization target. While traditional automated valuation models (AVMs) optimize for the final sale price, the Spalíkan optimizes for net value to the seller, integrating carry costs and the seller’s specific cash flow position into its timing recommendations.

Strategic Implementation

The model is intended to serve as a “PR stunt” and proof-of-concept for the Sokrates “basis compounding” thesis. By demonstrating that the autoresearch framework can identify non-obvious alpha in a well-understood market like real estate, the project aims to build credibility within the Reykjvík FinTech cluster. The model is slated for public hosting at sokrates.is/spalikan.

Commercial Trajectory and e-fasteignir

There is a strategic focus on e-fasteignir (operated by Blue E Solutions ehf), a digital self-service property platform in Iceland. While e-fasteignir currently utilizes a valuation model (Kaupsagan) for comparable sales and price index corrections, the Spalíkan offers the “optimization” layers they currently lack. The proposed business path involves using the model’s public visibility to initiate an equity-based partnership, where the Spalíkan becomes a premium feature on the e-fasteignir platform in exchange for a stake in the business.

Technical Evolution

The development of the model is intertwined with the evolution of the Sokrates ORM and query builder for Neo4j. This includes:

  • Gap 4 Implementation: Wiring writes through the query builder to replace technical debt associated with cast(LiteralString, query) patterns.
  • Graph Topology: Implementing depth limits (default 0 for writes, 1 for reads) and cardinality caps per hop to prevent query explosions in the knowledge graph.
  • Hypernode Patterns: Utilizing hypernodes for N-ary relationships (3+ entities) while maintaining simple edges for binary relationships.
  • Eidos
  • Autoresearch Framework
  • e-fasteignir
  • FinTech Cluster
  • Hyle Type System

Graph View

Backlinks