G
Awaiting your property
Enter the listing details on the left. Don Guillermo will run the numbers and issue a BUY, HOLD, or PASS verdict.
01 Price prediction
02 Rental analysis
03 Written verdict
A data-driven second opinion on any Madrid property. Trained on thousands of real listings, cross-checked against current rental data, and delivered as a one-page investment brief in under a minute.
Enter the listing details on the left. Don Guillermo will run the numbers and issue a BUY, HOLD, or PASS verdict.
Five inputs: address, district, asking price, size, and rooms. That's all the signal the ML model needs.
~10 secondsA BigML ensemble (10 decision trees, trained on 4,198 Madrid listings) outputs what the property should cost based on its features alone.
BigML · R² = 0.90A curated Google Sheet with Q1-2026 rental yields, sale prices, and growth rates for 11 Madrid districts is queried for the matching row.
Idealista 2025/26Given both the prediction and the district data, Claude Sonnet 4.6 writes a five-section investment brief and issues a BUY / HOLD / PASS verdict.
Claude · ~20 secondsA Random Decision Forest trained on 4,198 Madrid listings from Idealista (April 2019 snapshot). Features include size, number of rooms and bathrooms, and district. The ensemble averages across 10 decision trees, which reduces variance and handles non-linear relationships between location and price.
We trained several model types (linear regression, single tree, boosted trees) and chose the ensemble because it delivered the highest R² on the held-out test set.
A curated Google Sheet covering 11 Madrid districts across three market tiers: Premium (Salamanca, Centro), Mid-market (Tetuán, San Blas), and Value (Usera, Villaverde, Vicálvaro, plus the outlying Pinto zones).
For each district we store the 2026 average rent per m², 2026 average sale price per m², estimated gross rental yield, year-over-year rent growth, and a qualitative context note used by Claude when drafting the brief.
"I have been watching this city buy and sell itself for forty years. The numbers always know first."
Don Guillermo is not a single tool — he is a small committee of specialists wearing one coat. A statistical model that has read thousands of listings and knows what they cost. A reference book of current district data, kept up to date. An analyst who can read both and write the verdict in plain language.
The architecture is a classic Retrieval-Augmented Generation pattern: the ML prediction provides quantitative intuition, the Knowledge Base provides current market context, and a large language model synthesizes the two into a decision. None of the components can do the job alone. Together, they produce a brief that is grounded in real data and written like a human would write it.