Main strength
Compare numerical forecasts across a fixture list.
Main limitation
Model calibration and long-run profitability are not independently verified.
What we checked: a concrete example
Lokomotiv Moscow U19 vs Spartak Moscow U19
Checked 7 November 2026 · Public pages and retrieved source snapshots · Check protocol
| Find a match | Observed The football list separates 1X2, goals and BTTS sections; fixtures can be located by competition. Inspect primary page |
|---|---|
| Inspect the analysis | Observed The sampled archive row retains a home-win selection and a 3–1 score projection. Inspect primary page |
| Find earlier material | Observed The Yesterday link opens a previous-day forecast listing. Inspect primary page |
| Inspect results reporting | Observed, limited The same row reports a 3–0 full-time score. It is not an itemised betting-profit ledger. Inspect primary page |
Our assessment: Useful for comparing several numerical outputs, provided the market being evaluated stays consistent.
What is available without payment?
The public football pages include outcome percentages, a predicted score and match data, with separate navigation for goals and both-teams-to-score forecasts. The distinction between these outputs matters: a predicted home win and a predicted exact score are different propositions. A forecast can get one right and the other wrong. [1]
How the predictions are produced
Forebet describes a statistical approach using historical information and probability models, including Poisson-based methods. That explains the broad family of analysis, but it does not make the complete model reproducible. We have not examined its code, training data or how its estimates are calibrated. [2] A reader should ask whether a stated probability is consistently meaningful across many comparable matches, rather than interpreting the largest number in a row as a promise.
What can you check in the results?
The previous-day page displays forecasts and completed match scores in the same listing. This makes it possible to inspect individual successes and failures. [3] It does not establish that an entire historical series is immutable or that the odds shown were available when each forecast first appeared. We have not calculated a strike rate, return on investment or profit from these listings.
What the checked example reveals
The sampled home-win forecast agrees with the displayed outcome, while its exact-score projection does not. These are separate success conditions. A results review that counts a correct winner as a correct score would misrepresent the product. The archived fields make this distinction inspectable; this example is not a performance sample or proof that forecasts were unchanged after publication. See the dated four-task check and primary pages.
Usability and interpretation
The numerical columns make comparisons quick once the market labels are understood. The main reading burden is keeping outcome percentages, score projections and match scores separate. The feature check below shows why that distinction matters.
Who should consider the alternatives?
If you want a written case for a particular selection, compare the VictoryPicks review. If you want historical team data to inspect yourself, see FootyStats. Neither format is inherently more accurate. Our editorial methodology separates access to results from proof of predictive performance.
Editorial rating breakdown
Assessment dated 7 November 2026 · Based on the dated primary-source sample below.
| Verifiable results (30%) | 3/5 Previous-day forecasts and final scores can be examined together. Full-period completeness, publication timestamps and settlement reconciliation have not been audited. [3] |
|---|---|
| Analysis and methodology (25%) | 3/5 The provider explains its mathematical approach, including Poisson modelling; probabilities and score outputs are inspectable. The complete model and calibration have not been reproduced. [1] [2] |
| Free content (15%) | 4/5 Public outcome probabilities, score forecasts and supporting tables offer substantial free fixture research in the inspected sample. [1] |
| Coverage (10%) | 3/5 The inspected navigation confirms multiple competitions and prediction markets. Continuous season-long coverage has not been sampled. [1] |
| User experience (10%) | 2.5/5 · Estimated A comparable small-screen and keyboard-navigation assessment is still needed. No measured loading-speed or local-network test was performed. Neutral midpoint assumption; not a verified assessment of this feature. |
| Local relevance (10%) | 2.5/5 · Estimated A country-scoped assessment of domestic coverage, useful local context and access is incomplete. Country names or worldwide availability alone do not establish this criterion. Neutral midpoint assumption; not a verified assessment of this feature. |
Sensitivity to missing evidence: holding the assessed scores fixed, the total could range from 5.1 to 7.1/10 if the unverified criteria were all 0 or all 5. This is an assumption range, not a statistical confidence interval.
These are provisional editorial judgments about service features, not measured prediction accuracy or a betting-return forecast. How estimates and ranking work.
Sources and review date
Background primary sources were initially checked on 5 November 2026. The four-task check above records the newer 7 November 2026 observations and retrieval limits. Added sources explicitly identify their later check date. Dynamic content may change; this is not a paid-account or prediction-performance audit.
- Predictions and market navigationInitial check 5 November 2026
- What is Forebet: model explanationInitial check 5 November 2026
- Previous-day predictions and scoresInitial check 5 November 2026