One result versus the whole picture
A single outcome tells you little. What matters is the distribution: a list of every possible result with its probability. Once you have that, you can say which results are common, which are unlikely, and how far a typical session is likely to stray from the average.
A distribution has a shape, and the shape depends on how the game produces its results. Two shapes cover many of the games players meet.
Summed dice: a bell shape
Roll two six-sided dice and add them. There are 36 equally likely combinations, but the totals are not equally likely, because some totals can be made in more ways.
| Total | Ways | Probability |
|---|---|---|
| 2 or 12 | 1 each | 1/36, about 2.8% each |
| 3 or 11 | 2 each | 2/36, about 5.6% each |
| 4 or 10 | 3 each | 3/36, about 8.3% each |
| 5 or 9 | 4 each | 4/36, about 11.1% each |
| 6 or 8 | 5 each | 5/36, about 13.9% each |
| 7 | 6 | 6/36, about 16.7% |
The ways add up to 1+2+3+4+5+6+5+4+3+2+1 = 36, which is a quick check that the table is complete.
The result is a peak at 7 that tapers evenly to both sides. Add more dice and the shape becomes smoother and more bell-like, with the middle totals crowding together and the extremes growing rarer. This is why a character who rolls several dice to determine a score tends to land near the average, and why extreme scores are the exception.
The table also answers “typical” in plain numbers. The totals from 5 to 9 together have 4+5+6+5+4 = 24 ways, so they appear 24/36, or two-thirds of the time. A total of 2 or 12 appears together just 2/36, about 5.6 percent of the time.
Single rolls and drops: flat and skewed shapes
Not every outcome is a sum. One die gives a flat distribution, with every face equally likely. Waiting for a success is different again.
Suppose an item drops with a 10 percent chance per attempt. The chance of getting nothing in 10 attempts is 0.9^10, which is about 34.9 percent. So even after ten attempts at the 10 percent chance, the item is missing more than a third of the time. Meanwhile, some players will get it on the first try, which happens 10 percent of the time.
The distribution of “attempts until success” is skewed: the single most likely number of attempts is one, yet the average is 10, because a long tail of unlucky players pulls the average up. This is why the average and the typical experience differ in many video game luck systems. Reading only the mean misses the tail, and the tail is where players feel frustrated.
Typical, rare and the tail
A few ways of describing where a result falls are worth knowing.
- Typical results sit near the centre. In a bell shape, that is the mean, median and mode together. In a skewed shape they separate.
- Unlikely results sit in the tails. They are not impossible, just infrequent.
- Rare is relative to how many chances there are. A 1 in 100 event is rare in one session but expected somewhere among a few hundred sessions.
That last point matters. If a million players each have a one-in-a-thousand chance of something amazing, around a thousand of them will see it. Hearing about an extraordinary result usually says more about how many people tried than about the odds being broken. Seen this way, long shots do happen, and they can be understood without calling them miracles or evidence of rigging.
Reading a distribution before choosing
Knowing the shape helps you pick a plan. If outcomes cluster tightly, as with summed dice, then planning around the average is sensible. If outcomes have a long tail, then planning around the average can leave you short, so you may want a buffer or a fallback. In RPG progression, for example, the time to find a rare item may stretch well beyond the average, and it helps to have other ways of advancing in the meantime.
It also helps in bet sizing and risk choices. Where a decision has a bad but unlikely outcome in the tail, the question is whether you could absorb that result if it came. That is a question about the tail, not the average.
Limits
Distributions describe a game as it is designed; they do not tell you what any one session will do. Real games also sometimes adjust their own odds, for example with pity timers or matchmaking, which changes the shape. When a distribution matters to you, check whether the rules or published rates are available, and treat any claim about exact odds with caution if its source cannot be verified. Even with the shape known, a particular result remains uncertain.