Perhaps the most talked about aspect of Civilization V is how the game completely transformed the balance between tall (a few large cities) and wide (many smaller cities) strategies. With the announcement of Civilization VI, I have no doubt that this discussion will once again take centre stage amongst fans of the series when the new game releases.



For the first time in the civilization series, Civilization V added penalties to the founding of new cities; a global happiness limit, an increase in technology cost, and brutally expensive national wonders for large empires. The result of these changes was that empires with a smaller number of cities were equally as viable, or even superior to large sprawling empires. This made multiplayer very interesting, as players had to make important decisions between expanding and conquering, or sitting back and trying to win with more peaceful tactics.

So I decided to see if there was any relationship between the number of cities founded and the likelihood of winning. To do this, I turned to my dataset of game data from the multiplayer streamer FilthyRobot (see the about page for more info if you are new to this site).

This therefore means that any conclusions we try and draw from the data really only apply to FilthyRobot as a player. It’s possible that things might also apply to other multiplayer players, but I’d caution against assuming that we can learn anything about the single player experience. With that said, let’s dive in to the data.

How many cities is typical?

In order to think about this, we need to make an important distinction. A city founded or conquered in the first 100 turns will have plenty of time to contribute valuable gold/culture/science to the game. However, cities conquered during the late game won’t really tell us much about the effect of wide or tall builds on winning. I have therefore decided to use the number of cities at turn 100 as a proxy for the number of cities in the overall strategy. This cutoff is somewhat arbitrary, but it represents the time (on quick speed) after which new cities don’t tend to contribute much to the game. Consistently, it’s rare to see players founding new cities after this point anyway.

The number of cities at turn 100 therefore includes cities founded using settlers, as well as cities conquered with early war (more on this distinction later). Below is a plot showing the number of games (frequency) with different numbers of cities at turn 100. It is clear that “tall” builds (2-5 cities) are by far the most common strategy used by FilthyRobot, whereas wide builds (6-10 cities) are more rare. Just based on the sheer number of games with 4-5 city openings, this might suggest that 4 or 5 cities is the sweet spot.

When we split this data into games played with the tradition or liberty policy trees, the pattern is consistent with how these social policies are typically used. The tradition tree was more commonly opened for “tall” games, whereas the majority of “wide” games (7+ cities) used the liberty tree.

Does founding or conquering more cities improve the chance of winning?

This leads to the question of whether the most commonly used strategies (4-5 city tradition, or 5-7 city liberty) are associated with an increased likelihood of winning. Below is a graph showing the win % achieved with different numbers of cities at turn 100. Faded bars represent data with <10 games

The above graph shows that FilthyRobot’s win rate is surprisingly consistent across most builds, and is similar between tall and wide set ups. Despite the fact that he used 3 city openings much less often, this does not seem to affect the likelihood of winning. There is a reduced win rate associated with 2 and 10 cities, which might represent the lower and upper limits. 2 cities is perhaps too small to compete, whereas 10 cities might be a case of over-expansion.

The similarity in win rate across different numbers of cities is perhaps not surprising, as a skilled player will know when founding a new city is a tactically sound move, and therefore wont employ a wide build unless the conditions will support it. However, the above graph doesn’t distinguish between cities founded with settlers and cities conquered by war.

On the face of it, it might seem like cities conquered by war are “more expensive” in the early game, as they require the investment to build an army in order to take them. However, they have the benefit that the hard work of building infrastructure can be done by another player and taken advantage of by the aggressor, as well as the obvious advantage that an opponent loses cities in the process.



I was therefore curious as to whether the investment into early war harmed or improved FilthyRobot’s win rate overall. To examine this, I looked at the win rate associated with different numbers of cities captured before turn 100. Faded bars represent data with <10 games





To my slight surprise, the data seem to suggest that strategies involving the capture of cities in early war performed better overall than peaceful openings. I had previously assumed that building important infrastructure might be the more efficient strategy. While there isn’t enough data to properly correlate the number of captured cities with win rate, games with at least one city captured by early war were associated with a 13% higher win rate than peaceful openings.

This difference is similar in magnitude to beneficial start conditions that increase the likelihood of winning, such as coastal access and mining luxuries. The data therefore seem to suggest quite strongly that for FilthyRobot (who is a skilled warmongerer), the early investment into a military rush is likely to pay off in most games. Peaceful play, on the other hand seems to have served him less well.