A repeatable method
Probability problems in games look different on the surface but can usually be handled with the same four steps:
- Define the outcomes and what counts as success.
- Count or model the cases.
- Combine the probabilities with the right rule.
- Check the answer.
The steps sound obvious, yet most errors come from skipping the first or the last.
Step 1: Define the event precisely
State exactly what you are measuring. “The chance of a good roll” is not an event. “The chance that the total of two six-sided dice is at least 10” is.
Settle these details before calculating anything:
- What is the sample space, meaning every possible result?
- Are the outcomes equally likely? Faces of a fair die are; sums of two dice are not.
- Are the trials independent, or does one result change the next?
- Is the question about one trial, or about at least one success in several?
Much of the confusion about game odds comes from an event that was never pinned down, such as whether a stated rate applies per attempt or per session.
Step 2: Count the cases
When outcomes are equally likely, the probability is the number of favourable cases divided by the total.
Two six-sided dice have 6 × 6 = 36 equally likely ordered pairs. Count the pairs with a total of at least 10:
- Total 10: (4,6), (5,5), (6,4) is 3 pairs.
- Total 11: (5,6), (6,5) is 2 pairs.
- Total 12: (6,6) is 1 pair.
That is 6 favourable pairs out of 36, so the probability is 6/36 = 1/6, about 16.7 percent.
The common mistake here is to treat the eleven possible totals (2 to 12) as equally likely. They are not, and counting ordered pairs avoids the error. A total of 7 can occur in 6 ways, so its probability is 6/36 = 1/6, while a total of 12 has only 1 way, 1/36.
Step 3: Combine with the right rule
Once the building blocks are known, the combining rules are short.
- Independent “and”: multiply. A 50 percent hit chance followed by a 20 percent crit chance gives 10 percent for a crit.
- Exclusive “or”: add. Totals of exactly 11 or exactly 12 on two dice: 2/36 + 1/36 = 3/36.
- At least one: subtract the chance of none from 1.
Example: at least one success
A 25 percent effect has three tries, much like repeated attempts at a crafting step with modest upgrade success rates. The chance of no success is 0.75 cubed, which is 0.421875. The chance of at least one is 1 minus that, 0.578125, or about 57.8 percent. Adding 25 percent three times would give 75 percent, which overcounts the cases where the effect triggers more than once.
Example: a roll-twice-keep-higher rule
Some mechanics let you roll a twenty-sided die twice and keep the higher result. Suppose success needs a 15 or more.
On one roll, 6 of the 20 faces succeed (15 to 20), so the probability is 6/20 = 30 percent.
With two rolls keeping the higher, the result fails only if both rolls are 14 or lower. Each roll does so with probability 14/20 = 0.7, so both do so with probability 0.49. The chance of success is 1 minus 0.49, which is 51 percent.
That is a large jump from 30 percent, and it shows how a modest-looking rule can change odds substantially. The average of the kept roll also rises. Questions like this are the starting point for judging the expected value of a mechanic, not only its success rate.
Step 4: Check the answer
A probability calculation that is never checked is a guess in disguise. Several checks are cheap.
Range check. Probabilities must be between 0 and 1. A result above 1, such as adding three 40 percent chances to get 120 percent, is a clear sign that overlapping cases were double-counted.
Complement check. Compute the probability of the opposite event and confirm the two add to 1. In the two-dice example, the chance of a total below 10 is 30/36, and 6/36 + 30/36 equals 1.
Extreme cases. If a chance is raised to 100 percent, does the formula give certainty? If it is lowered to 0 percent, does it give impossibility? A formula that fails these cases is wrong.
Rough simulation. Roll physical dice or write a few lines of code to run the event thousands of times and compare the observed share. A simulation will not give the exact value, since it carries its own noise, but it should land close to a correct answer. With 10,000 runs of a 1/6 event, the typical error in the observed share is about 0.4 percentage points.
Sanity check against intuition. If the answer surprises you, check whether the intuition or the arithmetic is more likely to be off. Often it is the intuition, but not always.
Common mistakes
- Adding when you should multiply, or the reverse.
- Treating dependent events as independent. Drawing cards without replacement changes the odds on each draw, which is why mulligan decisions rely on counting what remains in the deck. Drawing two aces from a standard 52-card deck is (4/52) × (3/51), not (4/52) squared.
- Mixing up “exactly” and “at least”. Exactly one success in three tries is a different number from at least one.
- Ignoring order. Whether (3,4) and (4,3) are the same outcome changes the count. For two dice, they are distinct and equally likely.
- Forgetting hidden mechanics. If the game adjusts odds after failures, independent-trial formulas do not apply as they stand, and the published drop tables may describe only part of the real system.
When to stop calculating
Exact counting works for small, tidy mechanics. For complicated ones, with many interacting effects, conditional bonuses or changing states, writing a simulation is usually more reliable than deriving a formula. The same four steps still apply: define the event, model the cases, combine them, and check the output against something you can verify by hand.
The result is a number that describes a frequency, not a prediction of the next roll. Used that way, careful analysis shows what to expect over many trials and how much the next few results can reasonably vary.