Project · forecasting / competition

CrunchDAO Structural Break

A deterministic submission for the CrunchDAO Structural Break competition — a regime-change detection challenge on financial time series. The entry runs on CrunchDAO infrastructure, produces forecasts in the required format, and is fully auditable before submission.

Boundary: This is a competition submission, not a trading system. It risks no capital, holds no positions, and does not connect to any brokerage. It computes forecasts from provided data and submits them to CrunchDAO for scoring.

Competition context

CrunchDAO Structural Break: 100k USDC prize pool. Competition open until 2026-10-01. Code runs on CrunchDAO infrastructure; the submission must be a deterministic Python entry point.

Public competition page verified: HTTP 200 at hub.crunchdao.com/competitions. Official competition definitions cloned locally.

Method

The baseline implementation (submission.py) is deterministic and passes 6/6 local validation tests. It ingests the provided time-series data, applies a structural-break detection algorithm (changepoint detection with regime classification), and outputs forecasts in the competition's required schema. No external API calls, no random seeds, no network access at runtime.

Local validation (2026-09-11): 6/6 deterministic tests pass. Submission format verified against competition spec. Next iteration: complete CrunchDAO account registration/authentication, then submit before 2026-10-01 deadline.

Sources and verifiability

Competition platform: hub.crunchdao.com/competitions (public)

Local repo: The verified submission is in the project repository — commit 2d4620d.

All inputs are provided by the competition; the algorithm is pure Python with no hidden state. Any recorded forecast can be re-computed from the competition data.

Verification

Run python3 -m pytest or make test in the project directory — 6/6 deterministic tests must pass. Inspect submission.py for the complete algorithm. The evidence boundary is the competition rules and provided data; no private data or external APIs are used.