Definition
The weather forecast says "70 per cent chance of rain". It stays dry all day. Was the forecast wrong?
Probability is a measure of how likely something is to happen, from impossible to certain. It can be written as a fraction, a decimal or a percentage. Probability describes what tends to happen over many occasions, not what will certainly happen on any single one.
In simple words
If you toss a coin, heads and tails are equally likely: an even chance, or 50 per cent. That does not mean ten tosses will give exactly five heads. But toss it a thousand times and you will get close to five hundred. Probability is about the long run, not the next toss.
Why it matters
Every AI prediction is a statement of probability: "92 per cent likely to be spam", "80 per cent chance of rain". Understanding chance helps you read these correctly, make better decisions under uncertainty and avoid common traps.
Key terms
- Probability
- How likely an event is, from impossible to certain.
- Outcome
- One possible result, such as heads when tossing a coin.
- Event
- An outcome or group of outcomes we are interested in, such as rolling an even number.
- Equally likely
- Outcomes that have the same chance, such as each face of a fair die.
- Experimental probability
- A chance estimated by repeating an experiment many times and counting results.
- Gambler's fallacy
- The false belief that past random results change the chance of the next one.
Explained in detail
The probability scale
Probability runs from 0, meaning impossible, to 1, meaning certain. In percentages that is from 0 to 100 per cent. An event with probability one half, or 50 per cent, is as likely as not. Rolling a 7 on an ordinary die is impossible. The sun rising tomorrow is as close to certain as anything. A fair coin landing heads sits exactly in the middle.
Working out simple chances
When all outcomes are equally likely, count the outcomes that give the event and compare with all the possible outcomes. A die has six faces. Three of them, 2, 4 and 6, are even, so the chance of rolling an even number is three out of six, which is one half. A bag holds 3 red and 7 blue marbles: the chance of drawing red is 3 out of 10, or 30 per cent.
What a forecast percentage means
"70 per cent chance of rain" means that on days with conditions like today's, it rains on about seven out of ten. On the other three, it stays dry. A single dry day does not prove the forecast wrong. The forecast is judged over many days: if it says 70 per cent and it rains on about 70 per cent of those days, it is doing its job well. AI predictions work the same way. A spam filter that says "90 per cent likely spam" will still be wrong about one in ten such emails, if it is well calibrated.
Common traps
| Trap | Example | The truth |
|---|---|---|
| Gambler's fallacy | "Five heads in a row, so tails is due" | The coin has no memory; it is still 50 per cent |
| Small numbers | "I tossed it 4 times and got 3 heads, so heads is 75 per cent" | Few trials vary a lot; you need many |
| Treating likely as certain | "90 per cent safe, so nothing can go wrong" | One time in ten, it will |
| Ignoring rare but serious risks | "Only a 1 per cent chance of a fire" | Over many years, rare events do happen |
Learning chance by experiment
Some chances cannot be worked out by counting outcomes, such as the chance that a thrown bottle cap lands upside down. You can still estimate it by experiment: throw it many times and count. The more times you repeat it, the closer your estimate settles towards the true chance. This idea, that averages settle down over many trials, is called the law of large numbers, and it is the basis of insurance, opinion polls and the way machine learning estimates probabilities from data.
How it works
- List all the possible outcomes.
- Check whether they are equally likely.
- Count the outcomes that make up your event.
- Compare with the total number of outcomes, and express it as a fraction or percentage.
- If you cannot count, experiment many times and use the share of times it happened.
Try it yourself
Fifty coin tosses
A class tests whether a coin really lands heads half the time.
Coin and bottle-cap experiment
You need a coin, a bottle cap and paper.
- Toss the coin 50 times and tally heads and tails.
- Work out the share of heads after 10, 20 and 50 tosses.
- Now throw the bottle cap 50 times and tally "open side up" and "open side down".
- Estimate the chance for the bottle cap from your results.
Think about it
- Did the coin's share get closer to one half as you tossed more?
- Why could you not work out the bottle cap's chance by counting faces?
Common mistakes
- Believing a run of results changes the chance of the next one.
- Judging a probability forecast by a single day.
- Treating "very likely" as "certain".
Check yourself
Summary
- Probability measures how likely an event is, from impossible to certain.
- It describes the long run; single results can surprise.
- Avoid the gambler's fallacy, and read AI percentages as chances, not certainties.
Finished reading?
Sources: Khan Academy: Statistics and probability · India Meteorological Department