What a crude bot says about 57 ships

Before a release, we sometimes want a quick, rough answer to a plain question: is any capital ship wildly out of line with the others? Nobody can fly fifty-seven ships across dozens of battles by hand, so we asked a bot. This note is about what the bot said, and mostly about how little it can say, because the second part is what we learned.

The set-up

The bot is the same one we use for everything. It flies toward the nearest enemy until it is at about three quarters of its gun range, fires at it, circles a little and casts the ship’s four abilities on a timer when it has the energy. It never gives a squad an order and never aims an ultimate with any thought. Every one of the 57 capitals flew the same fight: a large fleet clash against the Iron Mandate at Commander difficulty, level 20, for up to 240 seconds, twice with different seeds. We recorded how long it survived and how much hull it had left.

This is deliberately a poor pilot. It is a ruler, and what we want from a ruler is that it is the same every time, not that it is clever.

What came out

No ship won the clash. That tells you the first thing: a capital flown with no orders and no aim cannot beat a Commander-level fleet on its own. The game is not built that way, because half of the fleet you command is the six squads that the bot ignored.

The median ship survived 141 seconds. The shortest life was the Bowline’s, at 82 seconds, followed by the Night Pilgrim (92) and the Scarab (94). Two ships were still alive at the 240-second cap in both runs: the Morrow Hive and the Sun Pyramid. Fifteen of the 57 lasted at least 200 seconds.

Grouped by role, the pattern is the one you would guess, and that is reassuring:

  • Carriers, nine of them, lasted on average 192 seconds, because the fighters they launch soak fire the bot never thinks to manage.
  • Supports (170 seconds) and artillery (162 seconds) came next, and vanguards averaged 151.
  • Strikers, thirteen of them, lasted only 125 seconds on average. A striker is built to dart in, hit and leave. The bot parked it at gun range and let it trade blows.

What we make of it

It would be easy to read the Bowline’s 82 seconds as a balance problem. We do not. The Bowline is a line battleship in the Lantern Line, whose trait rewards allied ships lined up abreast with it, and the bot never forms a line. The ship was asked to do the one thing it is not built for. The same goes for strikers: fly them like a vanguard and they die like one.

That is the useful result. The test could not rank two ships that are close to each other, and with two seeds it should not be asked to. What it can do is catch an outlier that is far outside its role’s range, and none of the 57 was. A ship that lasted far longer than its role, or far shorter for no reason of design, would have gone on our list to look at by hand.

What it cannot tell us

  • It cannot tell whether a ship is fun. A tough ship that is dull to fly would score well.
  • It cannot tell how good a ship is in skilled hands. A support’s ability to rescue the fleet at the right moment is invisible to a bot that casts on a timer.
  • With two seeds per ship, a difference of ten or twenty seconds between neighbours is noise.

We publish the numbers on the warship pages as a small line under each ship, labelled as a crude test, so that you can see what we saw. If you fly a ship and the line looks nothing like your experience, trust your experience. And tell us, because that is exactly the case where a better test is needed.