April 17 to May 7, 2026
Presented by Horizon Hobby & the Gies College of Business. Build a predictive model, present to industry judges, and win cash prizes.
LOSI 22S Sprint 2WD, the product line at the heart of this year's forecasting challenge
Our Partner
Headquartered right here in Champaign, Horizon Hobby is the world's largest manufacturer and distributor of radio-controlled cars, planes, boats, and helicopters. For four decades they've been bringing hobbyists the most innovative products in the industry.
They're also partnering with Gies on a real-world operations challenge worth $7M+ annually.
The challenge
Horizon sells world-class RC products. When customers need a replacement part, Horizon wants it to be available, but predicting which parts will be needed and when is hard.
Your job: build a predictive model on the LOSI 22S Sprint Car data, then apply it to the new 22S Dirt Oval launch.
On campus
Professor Seshadri’s class with the LOSI 22S Sprint, the product line at the heart of this year’s forecasting challenge. Real product, real demand history, real launch decision.
Diane and Steve Miller Classroom, Gies College of Business.
Form a team of two currently enrolled Gies students.
Sign the NDA, download the sprint car dataset and prediction template.
Build your model, submit predictions, code, methodology, and slides by May 1.
Top three teams present to Horizon Hobby on May 7.
Prizes
The three finalists who present to Horizon Hobby are selected by a panel weighing all of the following. Topping the leaderboard alone will not guarantee a finalist slot.
Your wMAPE rank on the public leaderboard — weighted MAPE on the three held-out Old SKU parts, 50/50 yearly + monthly. Lower is better; 0% = perfect forecast. Because actuals for the held-out parts are hidden, your leaderboard score is effectively a clean validation signal you can use to iterate.
Clarity of approach, data-use choices, assumptions, validation, and diagnostics — as evidenced by your PDF methodology and code submission.
How your forecasts would inform inventory and launch planning for the Dirt Late Model parts. New-SKU predictions are not scored on the leaderboard, but Horizon Hobby reviews them directly — this is the business deliverable.
Invest in all four submission components (prediction file, code, methodology PDF, presentation). A strong write-up with modest accuracy can beat a top-accuracy submission with weak analysis.
Key milestones from announcement to final presentations.
Website live. Team registration opens.
Horizon Hobby briefing. Data package available after NDA.
Follow-up clarifications from Horizon Hobby.
All four components due by 11:59 PM Central.
Top three teams announced.
Finalists present to Horizon Hobby. Winners announced.
You must be a currently enrolled Gies College of Business student. Teams of exactly two. Registration closes May 1.
Register your team