Direct answer
A custom model is a model created by the user at Predora.si to adjust the reading of matches per market.
It does not start from scratch: it starts from a baseline reading and uses factor boosts to give more or less weight to each signal.
Quick summary
| Definition | Model created by the user to personalize the reading of matches. |
|---|---|
| Baseline | Starts from a Predora.si baseline reading, not from a blank sheet. |
| How it adjusts | Uses factor boosts in each market. |
| Markets | Results, Over 2.5 goals, BTTS, and Over 9.5 corners. |
| Common mistake | Changing model weights without comparing historical performance. |
How a custom model works
A custom model works like your own lens within Predora.si. The baseline reading is adjusted according to the weight defined for each factor.
Customization is done per market, because a signal that is good for Match Result may not be as useful for goals, BTTS, or corners.
- None: the factor does not take priority in that market.
- Light: the factor has a smaller influence.
- Medium: the factor starts to weigh more.
- Strong: the factor receives one of the highest weights in that market.
Formula or logic
A custom model does not have a single formula that the user calculates by hand. It changes the reading that generates the estimated probability.
Custom model = baseline reading + factor boost adjustments in each market Fair odds = 1 / estimated probability EV = (estimated probability x odds) - 1Example in football
Imagine a match in the Over 2.5 goals market. The baseline reading estimates 54% and the best odds are 2.00, generating an EV of +8%.
If a custom goals model gives more weight to attacking volume, goals scored, goals conceded, and recent pace, its probability may rise to 57%. At the same odds, EV becomes +14%. Compare the model’s history and match factors to assess this new estimate.
How it appears on Predora.si
Custom models appear in the My Custom Models area. After saving, a model can be selected as the Active Model.
- Create a model with a name that is easy to recognize.
- Configure boosts per market.
- Hit Rate History, Return History, and Base describe the same slice through different dimensions.
- Compare with house models before trusting the reading too much.
House model vs custom model
No type is automatically better. What matters is the logic, the baseline, the market analyzed, and performance over time.
| Type | Best for those who want |
|---|---|
| Predora.si Models | To start with a ready-made reading. |
| Custom model | To test their own reading and refine analysis per market. |
Common mistakes
Setting everything to Strong without criteria
Too strong a boost can reinforce noise, duplicate similar signals, and leave the reading confusing.
Thinking custom is always better
A custom model only makes sense when it has clear logic and performance tracked with a sufficient sample.
Practical checklist
- Define the model's objective.
- Avoid setting every factor to Strong.
- Customize per market.
- Compare with house models.
- Compare Hit Rate, Return, and Base in the same slice.
FAQ about Custom model
Is a custom model created from scratch?
No. It starts from a Predora.si baseline reading and applies adjustments according to the chosen factor boosts.
Is a custom model better than house models?
Not necessarily. It may become more aligned with your style, but it needs logic and performance to make sense.
Does a custom model change bookmaker odds?
No. The odds remain the market odds. What changes is the probability estimated by your model.
Can I customize every market?
You can adjust blocks such as Match Result, Over 2.5 goals, BTTS, and Over 9.5 corners.
How do I compare a custom model with the default model?
Compare the same markets, periods, and sample sizes to assess how each model performed. Explore whether changed weights produce a useful difference.
Customization with method
A custom model works like a controlled experiment when it has a clear hypothesis, sufficient sample, and comparable performance.