How our AI models evolve
September 15, 2026
Artificial intelligence is constantly evolving.
At SignalTurf, our analysis models are regularly improved through new learning methods, our research work, and the ongoing growth of our historical data.
Every new generation has to prove itself
Creating a new model doesn’t mean it will be used in the app.
Before being deployed, every new generation is compared to the model currently used by SignalTurf using a rigorous evaluation methodology.
A new generation is only released if it demonstrates better performance than the one it’s meant to replace.
Otherwise, the previous model stays in production.
This requirement ensures that every update brings a real benefit to users.
An update that benefits every user
As soon as a new generation is validated, it directly replaces the previous model for all SignalTurf users.
There is no different version depending on subscription type: everyone benefits from the same analyses, based on the same model.
This approach has evolved since SignalTurf launched — find the details and reasons behind this choice in our release notes.
Performance shown in the app
The statistics available in SignalTurf always correspond to the model currently in production, used by all users.
This ensures that every user views the same performance indicators, based on a sufficiently representative history.
A fully automatic update
You don’t need to take any action.
As soon as a new generation is validated, SignalTurf automatically uses it to produce your analyses.
No configuration, no additional download, and no settings changes are required.
Go further
This page only concerns the process for evaluating and deploying our models.
To understand how they work, check out the AI models chapter.
The Types of bets chapter then presents the main bet types offered in the app and how predictions can be used to interpret them.