SURVIVOR (USA) WINNER PREDICTION CHALLENGE

4 winners × $100 each

ANYONE, ANYWHERE CAN COMPETE. Winners determined by prediction accuracy and confidence. The top 4 participants with the best logloss scores will each receive $100.

Build a machine learning model to predict the winner of Survivor Season 49 when there are 4 participants remaining in the final tribal council. Your model should output probability predictions for each of the final 4 contestants based on historical data and game dynamics.

Survivor is a reality competition TV show where contestants are stranded in a remote location and must survive while competing in challenges and voting each other out. Players must "outwit, outplay, and outlast" their competitors through strategic alliances, physical and mental challenges, and social gameplay. The game progresses through multiple eliminations until only a few contestants remain for the final tribal council, where a jury of previously eliminated players votes to determine the sole survivor and winner of the million-dollar prize.

New to Survivor? Watch this video for a comprehensive explanation of how the game works (The only difference is in the current version of the game, there are 4 people at the end instead of 3):

How Survivor Works - Explained

New to building Survivor prediction models? Check out our example repository that demonstrates how to:

• Load and explore Survivor historical data
• Build a machine learning model for winner prediction
• Format your predictions for submission
• Understand the competition requirements

The example includes R code, data analysis, and a complete walkthrough of the prediction process.

View Example Code Repository

• 🗳️ One Model Per Castaway

  Each participant may submit one machine learning model. All predictions must be submitted before the final episode airs.

• 🏝️ Submission Requirements

  Submit your complete code/model and a CSV file with 4 rows: contestant names and winning probabilities (must sum to 1.0).

• 🔍 Data Sources

  Use data from the survivoR R package (github.com/doehm/survivoR) or its Python wrapper (pinwheels.org/project/survivorpy) package, or any sources these packages reference. All data sources must be documented.

• 📋 Submission Format

  Submit:

  • Your complete code/notebook. Winning predictions will be verified using your submitted code, your predictions should be EXACTLY reproducible using your code.
  • CSV file with contestant names and probabilities. There should be 4 rows with 2 columns in your .csv submission: Contestant Name, Probability.
  • Brief model explanation.

Submission will be open until the day prior to the seasson finale (Not announced yet, but most likely mid December).

ONLY SUBMISSIONS THAT PREDICT THE WINNING PROBABILITY OF THE FINAL 4 CONTESTANTS WILL COUNT. MEANING YOU PROBABLY WANT TO WAIT UNTIL AFTER THE EPISODE BEFORE THE FINALE TO SUBMIT

Vote for Winner (Soon)