How many TCs on Arabia?

What players do

Over a full game, the typical Arabia player ends up with three TCs. The higher the level, the more TCs: above 2,100, nearly one player in two has four or more. The second TC goes down around 24 to 25 minutes, the third around 27 to 30, right after reaching the Castle Age (21 to 22 minutes on average).

How many TCs it takes to win

Players who stay on a single TC win 45% of their games; those who go up to three or four win 53 to 54%. Beyond that, the curve drops again: five TCs or more is 49%. Every game counts, whether it lasts ten minutes or an hour.

The best number shifts a little with level. Below 900, the peak is three TCs. From 1,500 up, it is four. Above 2,100, three and four are level, at 58%.

Do TCs win games, or do winners build TCs?

That is the real question: a player who is ahead can afford more TCs. To separate the two, we compared players who were in the same situation at 15 minutes, well before the second TC: same Elo gap, same villager gap, same army produced, same Feudal time. Every game of at least 15 minutes counts, short and long alike.

From the same situation, going from one to three TCs adds 15 points. Booming is not just a sign that you are ahead: it wins games. The fourth TC adds almost nothing.

No rock-paper-scissors

You might think a one-TC player with a big army punishes the boomer. We classified each player by their whole game: one TC with a lot of army (aggressive), one TC without (passive), two, three, four TCs or more. Then we looked at every match-up.

It is a hierarchy. The aggressive one-TC player crushes the passive one (72% wins) and wins 49% of games overall, as much as a two-TC player. But against two TCs it wins only 43%, and against three, 36%. Whoever builds more TCs wins more often, whatever the style across the map. Between three and four TCs, it is even.

One TC: you have to win fast

The player who stays on one TC wins half the games that end before 20 minutes. The longer the game, the lower the chances, down to 36% beyond 40 minutes. The player with three TCs or more does the opposite and wins more and more as the game goes on.

When to build your third TC

Among players who build a third TC, those who do it early win more, from the same position at 20 minutes. They also reached the Castle Age earlier: the third TC goes with a fast Castle.

The second TC is built once in the Castle Age. Only one civilization truly breaks the rule: the Cumans, whose bonus lets them build a Town Center in the Feudal Age. Almost all of them use it: 74% build their second TC before the Castle Age, around 11 minutes. It gives them no measurable edge: they win 46% of their games, as many as the Cumans who wait for the Castle Age.

Where to build your TCs

For every TC built after the starting one, we looked at the starting map within six tiles, the range at which it works as a drop site: gold, stone, forest. And its position: behind the starting TC, level with it, or towards the opponent.

The best spot combines a mine and a forest: gold and wood, or stone and wood, give +2.7 to +3 points. A TC in the open, far from everything, is the worst: −3.2 points for the second, −4.3 for the third. A forest alone or a mine alone does no better than average.

Pushing your TCs towards the opponent goes with winning: +3.3 points for the second, +3.8 for the third, while half the players build them behind. The best case observed is a second TC on stone, against a forest, forward: +8.6 points. A player who pushes TCs forward often already controls the map, even from the same position at 20 minutes: this figure also measures that confidence.

Civilizations

For each number of TCs, here are the civilizations that do best and worst, in win rate above what Elo predicted. Only civs with at least 150 players in the cell count. With these numbers, the margin of error is 4 to 8 points: only read the big gaps.

Two findings stand out. With a single TC, Burgundians, Bohemians and Lithuanians win far less than Elo predicted, by −15 to −18 points. And the Varangians sit at the bottom with two, three and four TCs.

What really matters at 25 minutes

Finally, a small model puts every factor side by side, in games that reach 25 minutes. It picks the winner in 64% of games, against 57% with Elo alone. The heaviest factor is the Castle Age time, then the villager gap, the army produced and Elo. The number of TCs weighs little once villagers are known: one more TC with the same villagers is worth about two points. A TC pays off because it makes villagers afterwards.

And what about the drop TC?

It is the famous “Persian Douche”: you delete your Town Center and drop a new one right next to the enemy base, before the Castle Age. Persians and Sicilians are the ones who play it, and 6% of their players try it. The new TC goes down around 9 minutes, closer to the opponent than to home in 98% of cases, and typically about an eighth of the distance from the enemy TC.

The drop TC pays off for both civs: +6 win points for Persians and +21 for Sicilians compared with those who play otherwise. The gap to Elo says the same: +8.8 and +19.9 points, against +2.6 and −0.6.

Take it with a pinch of salt: 122 players in total, 100 of them below 1,200 Elo. At that level, the drop TC wins 63% of its games against 53% for the others. Higher up, there are too few games to conclude.