Forebet, PredictZ and WinDrawWin: a transparent comparison

How popular prediction sites work and how to compare them fairly

Forebet, PredictZ and WinDrawWin all publish football predictions, but they do not present them in the same way. Forebet foregrounds algorithmic probabilities. PredictZ combines score and result tips with form, tables and accumulator content. WinDrawWin places predictions beside a large collection of team and league statistics.

This is not an accuracy ranking. We did not find a common, independently audited dataset that would support a fair “most accurate” claim across all three services. Instead, this article compares what a reader can see and verify on their public pages, checked on 7 November 2026.

What each prediction site publicly shows

The table below records features visible on official public pages. “Not located” does not mean a feature cannot exist elsewhere on the site. It means we did not locate it in the official pages cited in this review. Interfaces and policies can change after the check date.

The table describes disclosure, not forecast quality. A model can be detailed and still perform poorly. A simpler service can make useful calls. The point is that readers need enough information to test whichever claim is being made.

How Forebet works, according to Forebet

Forebet describes itself as a statistics-based prediction system. Its public match table can display home, draw and away percentages, a predicted result, a correct score, expected average goals, weather and prices. Its FAQ says the predictions come entirely from statistical and mathematical analysis, with no additional human adjustment to the prediction numbers.^[1]

Forebet also names parts of its mathematical framework. Its “What is Forebet” page refers to Poisson, “Beyond Poisson”, current form, home advantage and head-to-head records.^[2] Its features page lists several markets, live predictions, value selections, team comparisons and historical statistics.^[3]

That is more methodological context than a bare pick, but it is not a full reproducible model specification. A reader still cannot rebuild the current output from those descriptions alone. The public pages reviewed did not provide the complete feature weights, training and validation windows, version history, calibration chart or a downloadable pre-match forecast ledger.

What Forebet’s percentages do and do not mean

If a table shows 52% home, 27% draw and 21% away, it is presenting a distribution rather than a certainty. The 52% outcome remains expected to fail often. A correct-score prediction is usually the most likely single score cell, not a claim that this score has a high absolute probability.

Forebet distinguishes upcoming and finished predictions by colour and provides date navigation. Its FAQ also says live predictions are recalculated during play. That makes the distinction between pre-match and live output important: a changing live probability must not be used later as if it were the frozen pre-match forecast.

How PredictZ presents predictions

PredictZ’s public home page describes free football predictions, analysis, form, statistics, results and league tables. Daily pages show a predicted score and a home/draw/away indicator, while separate sections cover correct scores, both teams to score, totals and accumulators.^[4]

Its public company page identifies PredictZ as a trading name of Web Analysis Solutions Limited. The site also states that its picks do not guarantee wins or profits.^[5]

For a reader, the strength of this presentation is convenience: a compact tip sits beside form and competition information. The limitation is that a score or H/D/A letter does not show how probability is divided among the alternatives. In the official pages reviewed, we did not locate a model description comparable with Forebet’s high-level algorithm explanation, a fixed model-version history, or a public calibration report.

PredictZ also publishes accumulator and bookmaker-linked content. That does not make a prediction right or wrong. It means commercial links and prediction evidence should be assessed separately. A price shown on a page can change, so any return claim needs a source and publication time.

How WinDrawWin presents predictions

WinDrawWin says it covers more than 140 football leagues and provides free predictions with statistics and analysis. Its prediction directory links to today’s and tomorrow’s predictions, predictions already played today and yesterday, an archive and league pages.^[6]

Its statistics section covers BTTS, over/under goals, corners, half-time markets, draws, scoring and defensive records. The site says these statistics are updated after each team’s latest match and are used to formulate its football tips.^[7]

This makes WinDrawWin useful for inspecting the descriptive evidence around a selection. But descriptive statistics are not automatically a probability model. The official pages reviewed did not expose a complete prediction formula, feature weights, probability calibration or a versioned pre-match ledger with a publication timestamp for every forecast.

WinDrawWin and PredictZ identify the same registered company, Web Analysis Solutions Limited, on their about pages.^{[5] [8]} They remain separate public products and should be evaluated from their own outputs rather than treated as independent evidence confirming each other.

What popular prediction sites often do not show

The most useful missing information is not a secret “winning formula”. It is the information required to audit a performance claim.

An immutable pre-match timestamp

A fixture kick-off time is not the same as a forecast publication time. A strong record shows when the forecast was first published, which version was active and whether it changed. Ideally, the old version remains visible.

Without that trail, a reader cannot distinguish a genuinely pre-match prediction from a page that may have been updated later. This is not an accusation that any named site edits results; it is a limitation on what an outside reader can prove from the reviewed pages.

A complete denominator

“17 correct predictions” is incomplete without the total number of eligible predictions. A 70% figure also needs:

• the market being scored;
• the date range and leagues;
• rules for void, postponed and abandoned matches;
• whether every eligible prediction is included;
• whether several markets from one match are counted separately; and
• the number of forecasts behind the percentage.

Calibration, not only winners

A probability forecast should be checked by probability band. If events labelled 60% occur about 60% of the time across a suitable sample, that band is calibrated. A pick-only record cannot answer this question because it discards the probability assigned to the alternatives.

Hit rate can also be misleading across different markets. Home-win tips, double chance, over 1.5 goals and exact scores have very different base rates. Comparing them under one “accuracy” number is not fair.

A model change log

Models and editorial rules evolve. A public change log should say when inputs, weights, data providers or settlement rules changed. Otherwise, a lifetime headline can combine several systems that were never evaluated as one fixed method.

Error analysis

Publishing losses is necessary but not sufficient. A useful review asks where errors cluster: promoted teams, sparse leagues, late lineup changes, draws or particular probability bands. This turns a result archive into evidence about limitations.

Figure 1. A fair comparison uses the same consecutive sample and rules for every service. It does not start by selecting memorable winners.

What you can verify yourself

You do not need access to any site’s source code. A small prospective audit can test the quality of its public record.

1. Define the sample before the matches

Choose one competition, one market and a fixed sequence such as the next 50 eligible matches. Do not switch leagues or markets after seeing early results.

2. Save the public output

Record the URL, forecast, displayed probabilities, predicted score, page time, your capture time and scheduled kick-off. A screenshot helps document presentation, but a structured sheet is easier to score.

3. Use one settlement rule

Decide whether the market covers 90 minutes plus stoppage time, and define voids or postponements. Apply the same rule to Forebet, PredictZ, WinDrawWin and any other site.

4. Keep every eligible forecast

Do not remove low-confidence predictions, unpopular leagues or losses. If a site does not publish a forecast for a match, mark it as no prediction rather than treating it as a success or failure.

5. Compare like with like

For 1X2 selections, report hit rate with sample size. Where probabilities are available, also calculate calibration and a proper score. Do not compare a correct-score tip directly with a double-chance selection.

6. Separate selection quality from price

A prediction record and a return record answer different questions. Return requires the exact available price, its source and timestamp, stake rule, all losses and consistent treatment of voids. A later or best-case price cannot be inserted after the result.

How ProofPitch applies the same criteria to itself

The standards above also apply to ProofPitch. A comparison page is credible only if the publisher accepts the same test.

Read how a mathematical football model works for the modelling pipeline and how to audit a prediction site for the general checklist.

ProofPitch’s calculated card price is a documented calculation using the market reference and a 5% margin. It is not a bookmaker quote. A live market price must have its own source and timestamp. Likewise, the live public record and retrospective model check are different datasets and must not be merged into one performance headline.

Frequently asked questions

Is Forebet accurate?

That question needs a market, date range, sample size and scoring rule. Forebet publishes probabilities and finished-result indicators, but this review does not claim an independently verified accuracy rate. Test a consecutive pre-recorded sample rather than relying on selected wins.

Is PredictZ better than Forebet?

There is no fair basis here for that claim. Forebet reveals more high-level information about an algorithm and displays 1X2 probabilities; PredictZ foregrounds compact score and result tips with form and statistics. “Better” depends on a predefined, common evaluation.

Are PredictZ and WinDrawWin the same company?

Their official about pages identify both as trading names of Web Analysis Solutions Limited. They are still separate websites with different presentation and content, so one should not be treated as independent validation of the other.

Does WinDrawWin show past predictions?

Its prediction directory links to recently played predictions and its footer includes an archive link. For a rigorous audit, still record forecasts prospectively because archive access alone does not establish an immutable publication time or version history.

Which site gives probabilities rather than only tips?

Of the three reviewed, Forebet visibly publishes home/draw/away percentages in its main match tables. PredictZ and WinDrawWin pages reviewed focus more on selections, score forecasts and supporting statistics.

Can a correct-score prediction be called accurate if the match winner was right?

No. A correct-score forecast and a 1X2 forecast are different claims. If the predicted score was 2–1 and the match ended 1–0, the home-result direction was right but the exact-score prediction was wrong.

Should I trust a site with many winning tips today?

One day is too small and easy to select after the event. Define the sample in advance, retain all forecasts and measure the same market over many matches.

Can these sites make a profit certain?

No forecast can guarantee a result or profit. PredictZ explicitly states that its picks do not guarantee winning bets or profits, and Forebet describes its predictions as informational rather than betting services. Treat every prediction as uncertain.

Compare evidence, not headlines

Forebet, PredictZ and WinDrawWin organise useful football information in different ways. The fair question is not which homepage looked right yesterday. It is whether you can reconstruct a consecutive pre-match sample, understand what was predicted, preserve its time and version, and evaluate every outcome under one rule.

18+ only. Use only where legal. Gambling involves a risk of loss. Never treat a prediction site as a source of guaranteed income, never chase losses and do not use money needed for essential expenses. In South Africa, the National Responsible Gambling Programme lists a toll-free counselling and support line, 0800 006 008.^[9]

Notes and sources

1. Forebet. Frequently asked questions. Method, live changes, result colours and value sections; checked 7 November 2026.
2. Forebet. What is Forebet. Public description of mathematical frameworks and inputs; checked 7 November 2026.
3. Forebet. Main features. Public outputs and site features; checked 7 November 2026.
4. PredictZ. Free football predictions and tips. Public prediction, score, form and statistics presentation; checked 7 November 2026.
5. PredictZ. About PredictZ. Company identity and responsible-gambling statement; checked 7 November 2026.
6. WinDrawWin. Football predictions. League coverage and links to recent past predictions; checked 7 November 2026.
7. WinDrawWin. Football statistics and form. Statistics categories, update statement and use in tips; checked 7 November 2026.
8. WinDrawWin. About WinDrawWin. Company identity; checked 7 November 2026.
9. National Gambling Board of South Africa. National Responsible Gambling Programme. Toll-free support line; checked 7 November 2026.