1Football Software: Daily Correct Score Predictions From AI Software
Read Benfica–Gil Vicente's stored scoreline and goals tables, with checked calculations and clear limits on timing and performance evidence.
A ranked correct-score table shows how a model distributes probability across possible results. This guide uses the Benfica–Gil Vicente screenshot stored with this article to explain the numbers, without treating one selected example as a performance test.
The stored output
| Rank | Score | Displayed probability |
|---|---|---|
| 1 | 2-1 | 8.33% |
| 2 | 3-1 | 8.21% |
| 3 | 2-0 | 7.63% |
The top two scores differ by 0.12 percentage points, but they remain separate predictions. A second-ranked score occurring does not turn the first-ranked selection into a success.
Together these scores represent 24.17% of the displayed probability distribution, leaving 75.83% for other outcomes if the entries are part of the same distribution.
Reading the goals block
| Total | Over | Under |
|---|---|---|
| 0.5 | 97.15% | 2.85% |
| 1.5 | 90.08% | 9.92% |
| 2.5 | 75.76% | 24.24% |
| 3.5 | 56.45% | 43.55% |
Over 3.5 means four or more goals, not a particular exact score. With a consistent distribution, the difference between Over 2.5 and Over 3.5 assigns 19.31% to exactly three goals.
Several scorelines with two or three goals can individually rank highly even when four-or-more goals has a larger combined probability. That is not automatically a contradiction: a range contains many separate scores.
What has and has not been checked
The figures above have been checked against the stored image. It shows neither a capture time nor a product version. The earlier article described it as pre-match output and highlighted a 3-1 result without a precise fixture date or linked official match record. Those claims are not established by the screenshot alone, so this revision uses it as an output-reading example.
The GOAL HOME and GOAL AWAY panels have additional labels and percentages. Their exact meaning needs product-specific documentation; this guide does not treat them as a verified confidence scale.
Using an output record
Record the fixture, product and version, generation time and your evaluation rule. Keep the original output and add the result later. Include every eligible match, rather than only screenshots that match the outcome.
See the correct-score product listing for its documented features. The free website record and methodology concern a separate model and do not prove paid-product accuracy.
Correction, 19 September 2026: Rechecked the screenshot figures, clarified ranked-score interpretation, and removed unsupported timing, result-context and product-performance claims.
Predictions are estimates, not guarantees. 18+. Support.