Started
2025
The project began as academic work around football data, prediction quality, and what actually creates betting value.
Roadmap
How RobotBets Is Evolving
RobotBets started in 2025 as academic research. Since then, we have moved from basic football ML experiments to adaptive models and a broader vision for smarter sports intelligence.
Breakthrough
December 2025
That was the moment we shipped the first model that truly worked and kept improving with fresh data.
Next
Beyond Football
We want deeper model families, stronger research pipelines, and carefully expanded coverage into new sports.
Timeline So Far
The product did not appear overnight. It was built step by step, with each stage teaching us what actually matters in prediction quality and long-term profitability.
Academic Research Foundation
RobotBets began as academic research focused on football data, model validation, and understanding where betting value really comes from.
01
02
First Basic ML Models
We built the first generation with more classical machine learning methods and started testing whether structured football data could consistently surface useful edges.
Model Reset During a Faster AI Shift
As the world of AI moved quickly, we chose not to stay stuck with the first stack. We reworked the architecture and started aiming for models that adapt better over time.
03
04
First Self-Improving Model in Production
This was the first model that truly worked for us in a meaningful way: it performed well, learned from new match data, and improved through continuous retraining.
What We Are Building Next
The roadmap is not just about shipping more pages. It is about building a stronger research engine, better models, and wider sports coverage without lowering quality.
Research First
Stronger Model Families
We want to go beyond the current stack and add richer model families, better ensembles, and stronger calibration so probabilities become more useful in real decisions.
Next Wave
Research-Led Upgrades
Approaches like PNAS-style models and other research-driven methods are part of the next layer we want to test and bring into production.
Direction
Better Product Depth
The goal is not only more predictions, but also better explanations, cleaner confidence handling, and a system that becomes more trustworthy as it grows.
Longer-Term Vision
Football remains the core, but the long-term direction is bigger than one sport.
Next Wave
PNAS and More Advanced Architectures
We want to implement models inspired by approaches like PNAS and combine them with the systems already running on the platform.
Expansion
More Sports, Starting with Basketball
Once the research and model quality are good enough, we want to expand beyond football into sports like basketball and later others that fit the same standard.
Direction
From Tip Platform to Sports Intelligence System
The broader aim is to turn RobotBets into a smarter sports intelligence platform, where research, models, and product design reinforce each other.
We Are Still Early
The roadmap is ambitious on purpose. We would rather grow carefully, keep improving the models, and expand only when the quality is there.