ADIA Lab Structural Break Challenge: Real-Time Edition
ADIA Lab, in partnership with Crunch Lab / CrunchDAO.
📌About the Event
ADIA Lab Structural Break Challenge: Real-Time Edition is a global machine learning and data science competition by ADIA Lab in partnership with Crunch Lab. Participants build models that detect structural breaks in time series data in real time, meaning they must identify when the behaviour of incoming data changes instead of analysing the full series later. The challenge is useful for fields like finance, healthcare, industrial systems, and climate science, where early detection of sudden changes can be critical.
ℹ️ Event Details
Mode of conduct: Online competition hosted on the CrunchDAO / Crunch platform.
Competition type: Machine Learning / Data Science / Time Series Challenge.
Task: Participants receive a historical reference segment and then an online segment revealed one observation at a time. After each new observation, they submit a score from 0 to 1 showing confidence that a structural break has already occurred.
📅 Important Dates
- Start date: May 6, 2026, 4:00 PM UTC.
- Start date in India: May 6, 2026, 9:30 PM IST.
- End date: September 17, 2026, 4:00 PM UTC.
- End date in India: September 17, 2026, 9:30 PM IST.
- Quota refresh: Every Wednesday at 4:00 PM UTC.
- Winners announcement: Winners will be announced at the ADIA Lab 2026 Symposium, scheduled for October 26–28, 2026
🎯 Eligibility & Rules
- Who can participate: The challenge is meant for data scientists, machine learning engineers, statisticians, and researchers working on time series analysis, sequential modelling, and real-time systems.
- Student eligibility: No student-only or college-only restriction is mentioned on the official challenge page; it appears open to relevant global participants on Crunch.
- Registration requirement: Participants need to create a CrunchDAO account to access the competition dataset and submit models.
- Core rule: At every new observation, the model must output a confidence score between 0 and 1 that a break has already occurred.
- Break condition: A structural break may occur at any point in the online segment, or it may not occur at all.
- Evaluation metric: Submissions are scored using Time-Stratified AUC (TS-AUC), where per-step AUC is calculated across live series and then weighted into one score.
- Compute limit: The challenge mentions a 15-hour weekly compute budget across 10,000 series and up to 1,000 steps each.
- Allowed approaches: Participants may use methods like streaming statistical tests, online change-point detection, Bayesian tracking, supervised deep learning, or time series foundation models.
🏆 Prize Details and Value
- Total prize pool: USD 100,000; Crunch listings also show it as 100,000 USDC.
- Main incentive: Opportunity to compete in a global, high-impact ML challenge connected to real-world problems in finance, healthcare, industrial monitoring, and climatology.
- Recognition: Winners will be announced at the ADIA Lab 2026 Symposium, giving top participants international visibility.
- Portfolio value: Participants can showcase work in time series modelling, real-time ML, and structural break detection.