[terraform estimator]

Price estimator for Terraforms by Mathcastles parcels — terraformestimator.xyz.

Architecture

• Frontend — Next.js 15, deployed to Vercel (frontend/).
• Backend — Node/Express API, deployed to Render (backend/).
• Pricing — a hedonic model fitted to ~20,000 settled sales (backend/src/hedonicModel.js, coefficients in backend/src/pricing-v2-coeffs.json), scaled by the live collection floor. Each parcel gets a range: what it would clear at taking a bid, and at a listing that gets bought. Rare specials (Godmode, Plague, seeds, Lith0) fall back to the hand-tuned v1 model, which is also served on the unlinked /legacy page.
• Refit — nightly at 04:20 on the maintainer's machine (ops/daily-refit.sh): pull new sales, refit, re-calibrate against the last 30 days of sales, verify, test, commit, push. The sales database never leaves that machine; the pipeline scripts are in sales database/.
• On-chain — traits and artwork come straight from the Terraforms contract over Ethereum RPC.

Local development

cd backend && npm install
# backend/.env: RPC_URL, ALCHEMY_API_KEY (live floor), OPENSEA_API_KEY (listings and sales feeds)
npm run dev          # :3001

cd frontend && npm install
# frontend/.env.local: NEXT_PUBLIC_API_URL=http://localhost:3001
npm run dev          # :3000

https://ethereum.publicnode.com works as an RPC. eth.llamarpc.com returns null and cloudflare-eth.com fails network detection — don't use either.

Tests: npm test in backend/ (sales comparison rules and model invariants) and in frontend/.

Deployment

Pushing to main redeploys both: Render builds the backend, Vercel the frontend. The pre-push hook appends a live floor sample and commits it, then asks for the push to be run again.

Maintaining the model

See CLAUDE.md for the moving parts and sales database/fit_hedonic.py for how the fit is chosen (its header records the backtest behind the current half-life and regularisation).

License

The code is free software under the GNU Affero General Public License v3.0 or later (LICENSE): you may use, study, change and share it, and if you run a modified version as a service you must offer its source to its users.

The sales data published for download (frontend/public/data/, the open-data section of /stats) is dedicated to the public domain under CC0 1.0 (full text): use it for anything, no permission or credit needed.