Why average length is worth knowing
Game designers care about length because it shapes how a game feels. A race that usually ends in about eight turns plays very differently from one that drags on for forty. Players care too, since a “quick game” that sometimes runs three times longer than advertised is frustrating. The usual tool for the question is the expected number of turns, the long-run average over many plays.
You can often get that average from a single piece of information: the chance that the game ends on any given turn.
The waiting-time rule
Suppose that on every turn there is a fixed probability p that the game ends, and that turns are independent. The expected number of turns until it ends is 1/p.
Check it with a die. If you need to roll a six to finish, p = 1/6 and the expected number of turns is 6. If you need a five or a six, p = 1/3 and the expected number is 3. If you need any roll that is not a one, p = 5/6 and the expected number is 6/5, or 1.2 turns.
The rule is intuitive: a one-in-six event tends to show up once per six tries, so you wait about six tries. It is an average, not a promise. Any single game can end on the first turn or take far longer.
Adding stages together
Many games have several phases. A common pattern is “roll a six to leave the start, then reach the finish by rolling a five or six.” Averages of consecutive phases can simply be added, even though the phases are random.
- Stage one: p = 1/6, so 6 turns on average.
- Stage two: p = 1/3, so 3 turns on average.
- Total: 6 + 3 = 9 turns on average.
This works because expected values add, a property that holds even when stages depend on each other. It makes the stage-by-stage approach one of the most useful shortcuts for estimating game length. The same idea underlies the broader notion of expected value, applied to a count of turns instead of a payoff.
The average is not the typical game
Staying with the single-six example, the chance that the game is still going after n turns is (5/6)^n. After 6 turns, that is 15,625 / 46,656, which is about 0.335. So roughly one game in three runs longer than the “expected” six turns.
The median, the point where half of games have finished, is lower. The chance of finishing within 3 turns is about 42 percent, and within 4 turns about 52 percent, so the median is 4 turns. The average of 6 is pulled upward by the occasional long game. Whenever a game has a “wait until something happens” structure, expect this skew: many short games and a minority of long ones.
This is why a designer who only reports the average can hide a real problem. Spread matters as well, and the standard deviation of a waiting time with p = 1/6 is a bit under 5.5 turns, which is almost as large as the average itself. (The calculation is the square root of (1 - p) / p², which gives the square root of 30, about 5.48.)
When turns are not identical
The 1/p rule needs the same probability every turn. Real games often change as they go: a player closer to the finish has better chances, a deck gets thinner, a trap square only matters at certain positions. In those cases you track the situation, not just the turn count.
One exact approach is to define the expected remaining turns from each situation and relate them to each other, which is the logic behind Markov chains. Position-based movement, such as a token travelling around a board, is a natural fit. The topic of board game movement and landing odds on specific squares is a close relative, because average length depends on how often a token lands on shortcuts and delays.
Estimating duration by simulation
When the rules are too tangled for a clean formula, a Monte Carlo simulation answers the question directly. Play the game many times, record how many turns each play took, and average the results. As a check, test the method on the single-six example: the simulated average should come out near 6 if you run enough games.
Recording every game’s length, not just the mean, lets you see the full shape: the shortest, the longest and the median. How reliable the average itself is depends on how many simulation trials you run, since an average from a small sample will wobble from run to run.
Practical cautions
- Check independence. If one turn’s outcome affects the next, multiplying or dividing by a fixed p will be off.
- Decide what counts as a turn. Extra turns from doubles or bonus rolls change the counting, so define it before calculating.
- Report a range. “Usually between 4 and 12 turns, averaging 9” is more honest than a single figure.
- Do not read gambling meaning into it. In games with a house edge, a longer session means more rounds at a negative expectation for the player, so expected losses grow with duration. A short average game length is not a sign of safety, and chance-based games should be played only with money one can afford to lose.
Average game length is a modest but practical measurement. With a per-turn probability, a clear definition of a turn and an honest look at the spread, you can describe how long a game is likely to take without playing it hundreds of times.