Define the event before counting
For a fair six-sided die, one specified face has probability 1/6 and an even face has probability 3/6. The event even combines three distinct outcomes. In roulette, red similarly combines several pockets, but zero is outside the red and black groups. Counting only visible categories can therefore create an incorrect denominator.
In a standard single-zero wheel with equally likely pockets, red is 18/37, about 48.65%. That is a model based on the wheel's 37 pockets and conventional colour assignment. A different wheel or special game requires its own outcome space.
Theoretical and observed proportions
A modelled probability is not necessarily equal to the proportion seen in a short record. Ten die rolls can include no sixes without disproving the 1/6 model. A large appropriate sample gives more useful evidence, but even then assumptions about independence and the data-generating process matter.
Conditional probability and combined events
The probability of one event given another can differ from its unconditional probability. Two selections from the same match may share information and cannot automatically be multiplied as independent. Use the actual conditional model when combining events. Mutually exclusive exhaustive outcomes must have probabilities totalling one.
Avoid certainty from a percentage
An 80% event can fail, and a 1% event can occur. A betting price's implied percentage incorporates a quote, not an independently measured true chance. Probability describes uncertainty rather than removing it. Read its source, date, scope and assumptions before using it to interpret a payout, forecast or claim that something is due.
What is the difference between at least one win and all wins?
For two independent events each with 50% win probability, all wins has probability 25%. At least one win has probability one minus the chance both lose: 1 − 0.5 × 0.5 = 75%. These describe different events. A standard all-to-win accumulator cannot use the 75% figure as its success probability. Define the qualifying event before multiplying or adding percentages. Overlapping events also require care because simple addition can count a shared outcome twice.