The 8th Day Engine

The model makes the prediction.
Every result stays on the record.

Data in. Prediction out. Every match runs through the same modelling pipeline before a single pick is published — and once the game ends, the verdict is graded and kept, win or lose.

Same process every match · 2,400+ graded
The pipeline

How a CleverScores prediction is made

Six stages, applied the same way to every fixture — whether it's a World Cup final or a mid-table league game. The engine builds the case, tests it, then locks it before kick-off.

01

Build the matchup

The engine assembles both sides from team strength, recent form, home and away records, scoring patterns, rest and availability, and the other factors that shape how a game is likely to go.

02

Estimate the likely result

Statistical models estimate how many goals or points each side is likely to produce — the baseline expectation for the match, not a single guessed scoreline.

03

Simulate the match

The engine plays the matchup out many times over to measure win, draw and loss probabilities. A prediction describes the range of what could happen, rather than betting on one imagined outcome.

04

Check the wider context

Injuries, squad strength, venue, scheduling and tournament conditions can adjust the baseline — but only where the evidence supports it, never on a hunch.

05

Publish the verdict

The final prediction, its probabilities and the likely score are locked before the match begins. Once it's published, it can't be quietly edited after the fact.

06

Grade the result

After the game, the prediction is settled and added to the public accuracy record. Losing predictions are graded and kept, exactly like winning ones.

The difference

How that compares

Most prediction sites show you the pick. Fewer show you the process, and fewer still keep the ones that lost.

Most prediction sites
CleverScores
Picks with no method behind them
The same process, explained openly
Winning examples cherry-picked
Every settled prediction recorded
Opinion-led
Model-led
Odds presented as predictions
Model and market kept distinct
Results can quietly disappear
A permanent public ledger
Non-negotiables

The rules every prediction follows

These don't change from match to match, and they don't bend for the popular team.

Model first

The published pick comes from the prediction engine — not from an article writer, and not from a chatbot.

No result hiding

Correct and incorrect predictions both stay in the public record. Nothing is deleted to protect the hit rate.

Probabilities, not certainty

A 65% prediction can still lose. The percentage describes likelihood — it was never a guarantee.

Market independence

Bookmaker odds may be used for comparison and to spot value, but they don't choose the model's winner.

Same process every match

The engine applies the same core method regardless of how big or small the teams are.

Locked before kick-off

The verdict is published before the game starts, so it can never be tuned to fit a result after the fact.

Questions

Straight answers

Can the model be wrong?

Yes. A prediction expresses probability, not certainty. Even a correctly calculated favourite can lose. The standard isn't perfection — it's whether the model performs consistently over a meaningful number of recorded predictions.

Do you just copy the bookmaker odds?

No. Bookmaker odds may be used for comparison and to identify value, but they don't choose the model's winner. The model and the market are kept separate on purpose.

Are losing predictions removed?

No. Every settled prediction is graded and kept — correct and incorrect alike. That's the whole point of a public record.

Does a person or a chatbot write the pick?

The verdict comes from the prediction engine. Any surrounding text only explains the reasoning — it can't overrule the model's call.

Is every sport handled the same way?

The same core method runs across football, MMA/UFC, NBA and MLB: build the matchup, estimate the likely result, simulate it, check the context, then lock and grade. The inputs differ by sport; the process doesn't.

The model makes the prediction.
The data explains why.

See the pipeline in action against real fixtures — and check the picks against the results.

See today's predictions →