What a 41% chance of winning actually means

Start with the event being predicted

A percentage needs a label. A chance of Liverpool winning a match is different from a chance of a player scoring, or a chance of starting given that the player appears. Before reading the number, check the fixture, competition, event and any condition attached to it.

The example below is hypothetical. It is not a REDS forecast, an archived match or a claim about a particular opponent. We use it because the arithmetic can be checked without asking the reader to trust an unavailable match record.

Why three displayed percentages can add up to 99

Imagine the unrounded chances for a match are:

• Liverpool win: 41.4%.
• Draw: 30.3%.
• Opponent win: 28.3%.

These add up to 100%. Displayed as whole percentages, they become 41%, 30% and 28%, which add up to 99%. The missing percentage point comes from rounding each number separately. Nothing has been assigned to a fourth result.

This explains how a 99% total can arise. It does not verify any historical forecast with those displayed numbers. Without the underlying values, we cannot say that an old discrepancy was definitely caused by rounding. We also should not change a displayed probability merely to make a total look tidy.

The largest slice can still be less than half

Liverpool winning is the largest of the three individual possibilities in this example. Yet the draw and opponent-win chances together are 58.6%. An honest reading is: Liverpool have the highest individual result probability, while a draw or defeat is more likely in combination.

That distinction matters when a headline says REDS “favours Liverpool”. It should not turn a 41.4% estimate into “Liverpool will win”, “a safe pick” or “better than an even chance”. A percentage expresses uncertainty; it does not remove it.

A scoreline answers a different question

An estimate for a particular scoreline concerns one result among many possible scores. It cannot be substituted for the overall chance of winning. A scoreline table and a win/draw/loss estimate also need not have been produced by the same calculation.

Read each output on its own terms. If the page records that different predictions point in different directions, keep that disagreement visible. An article should not manufacture agreement by editing the numbers or presenting one output as proof of another.

What the final whistle can tell us

A match result establishes which event happened. It does not, on its own, establish whether a probability was well judged. An outcome given a substantial chance can fail to happen; an outcome given a small chance can happen.

To judge the forecasts, examine predictions saved before their matches, then compare them with the later results. Look for a clearly stated evaluation period, the number and type of matches included, and a comparison with a simple alternative. For a probability forecast, the useful question includes whether events assigned similar chances occur at roughly the expected frequency across repeated predictions. That is an evaluation question, not a replacement for producing a forecast.

A good historical result does not guarantee the next one. A claim of improvement needs the relevant evaluation record; this article does not claim a particular accuracy rate, a profitable strategy or a minimum sample size that makes REDS reliable.

Before sharing the number

Keep the event label, fixture and forecast time with a screenshot. Preserve conditions such as “if the player appears” and any “too early to call” or reference-only label. Check whether the result was locked before kick-off, calculated later, or remains a live estimate.

For the current answer, open REDS. For the reason pre-match records matter, read the companion article on locking predictions. None of the illustrative numbers here should be quoted as today’s forecast.