The BetSignals star rating grades one thing: value. It tells you whether the price on a selection is better than it should be, not how likely the result is. This guide explains exactly how the rating is worked out, step by step.
If you have read this guide before, note that the rating has changed. It used to measure model confidence. It now measures value, which is a far better guide to whether a selection is worth backing. The win percentage still tells you how likely a result is; the stars now tell you how much value it carries.
---
What the star rating measures
Two questions decide the rating:
1. Do the two independent models agree on the outcome?
2. If they agree, are the market odds longer than the model's fair odds, and by how much?
The first question is a gate. The second is where the stars come from.
---
Step 1: do the models agree?
Every match is checked by two models built independently, including the attacking and defending Elo model. Each has a most likely outcome. If they point at different outcomes, you get a cross (✕) and no selection, whatever the percentages say. A cross means the models disagree. It is not a weak signal; it is no signal.
If both point at the same outcome, that becomes the pick and we move on.
---
Step 2: is there value?
Take the model's probability for the agreed pick and turn it into a fair price: fair odds = 1 divided by the probability. A 50% pick has fair odds of 2.00, a 40% pick has 2.50, and so on.
Now compare it with the market:
- If the bookmaker's odds are longer than the fair odds, the selection carries value.
- If they are shorter or equal, there is no value. The pick scores a single star and stops there.
A single star, then, is an agreed selection that the market has priced fairly or too short. It is the most common result, and there is nothing wrong with it. It simply is not a value bet.
---
Step 3: grading the value
When there is value, the rating grades how strong it is. Home and away run on different scales, for a reason explained below.
| Rating | What it means |
| --- | --- |
| ★★★★★ | Away value at odds up to 2.50 |
| ★★★★ | Away value at odds 2.50 to 3.50 |
| ★★★ | Away value at odds 3.50 to 5.00 |
| ★★ | Away value above 5.00 (higher variance), or any home value |
| ★ | An agreed pick with no value |
| ✕ | The models disagree, or there is not enough data |
---
Why home and away are graded differently
Home results are priced very efficiently by the market. Genuine value on a home side is rare, and when it does appear it is weak, so a home pick never scores more than two stars, however likely the win. Away sides are where real value tends to sit, so they carry the full ladder, with the shortest-priced value rated highest because shorter prices are the easiest to get right.
This is deliberate. The two sides behave differently in the market, so grading them on the same scale would be misleading.
---
Worked examples
- An away side is 45% to win, so fair odds are 2.22, and the market is offering 2.60. That is clear value at a mid price, so it rates four stars.
- A home favourite is 70% to win, so fair odds are 1.43, and the market is offering 1.42. Very likely, but no value, so it rates one star despite the high percentage.
- The primary model calls a home win while the Elo model calls a draw. They disagree, so it is a cross, with no selection at all.
---
How to use it
More stars means a stronger price advantage on the model's agreed pick. Treat the rating as a value filter, not a likelihood filter: a high star is a selection the market may be underpricing.
For likelihood, which is what you want when building an accumulator, read the win percentage instead. The two numbers do different jobs, and there is a full walk-through in How Percentages and Star Ratings Work.
The data surfaces where the price looks wrong. The decision is always yours.