How to Win Fleet Battles in Space Games, With Numbers From Our Own Bot Trial

If you keep losing big fleet clashes in space games, the usual cause is not a weak ship. It is that the fight was decided in the first minute by things you did not do: you spread your damage across many targets, you let the fast enemy ships choose their fights, and your supporting ships never got an order. We learned how much orders matter in Cosmic Starship, our own browser game, when a bot flew all 57 of its capital ships through the same big battle with no orders at all and not one of them won. This guide uses that trial and the game’s own data for the specific numbers, and general fleet-combat principles for the rest, and it keeps the two clearly apart.

What our bot trial showed

For the bot trial we wrote about in our dev notes, every one of the 57 capital ships 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 random seeds. The bot closes to about three quarters of its gun range, fires, circles a little and casts abilities on a timer. It never gives its squads an order and never aims an ultimate with any thought. No ship won. The median ship lasted 141 seconds.

That is a useful baseline for a player: a flagship that only flies and shoots is not a fleet. Half of your force in our game is the six squads, and the bot ignored them. If you are losing, ask first whether you have been using yours.

The trial also says something about roles. Carriers lasted 192 seconds on average, supports 170, artillery 162, vanguards 151, and strikers only 125. Two ships, the Morrow Hive and the Sun Pyramid, were still alive at the 240-second cap in both runs, and fifteen of the 57 reached 200 seconds or more. At the bottom, the Bowline lasted 82 seconds, the Night Pilgrim 92 and the Scarab 94.

Read that carefully, because it is easy to take the wrong lesson. The Morrow Hive has the lowest damage output of all 57 ships (23.5 per second, rank 57 of 57) and one of the shortest reaches (range 300, rank 55 of 57), yet it outlasted nearly everyone, because it carries a brood of fighters that soak fire and a carapace that regrows 7 hull a second. Surviving is not the same as winning, and the bot cannot tell you which ship wins in good hands. What it does show is where a ship’s strength sits: a carrier lives by what it launches, a striker by where it is not standing.

The mistake behind most strikers’ deaths

Strikers averaged the shortest lives, and the bot’s flaw explains why. The game describes strikers as built to dart in, hit and leave, with agile handling that stops on a coin and dashes far. The bot parked them at gun range and let them trade blows, so they died like vanguards. The practical rule for a human is to pick a target, close in, use the power, and boost out before the return fire lands. A boost gives about 0.34 seconds of invulnerability, and if you boost through a “HEAVY STRIKE” (the flickering white rim and the alarm that mark an attack that would take at least 8 percent of your hull) you take no damage, gain 12 energy and get half your boost cooldown back.

Running the numbers across all 57 ships also gives a rough pattern: faster ships lasted shorter in the bot trial (a correlation of about minus 0.5), and total hull plus shield lasted longer (about plus 0.4), while a longer gun range made almost no difference to survival. With 57 ships and two seeds, treat that as a hint and not a law. It does fit the idea that standing in the open with thin armor is the thing to avoid, whatever your range.

Orders: the cheapest improvement you have

Press Tab for command mode. Time slows, you pick squads with the number keys 1 to 6 (0 selects all), pick a duty and click a target. The duties are Escort (E), Attack (T), Defend (F), Hold (H), Intercept (I) and Objective (O). The game’s squad guide suggests a default we agree with: two interceptor squads on Intercept to deal with fighters, lancers on Attack against the biggest enemy capital, and bulwarks on Escort. If you only learn one sequence, learn Tab, 0, T, click, which sends everything at the most dangerous target. G does something similar without slowing time: every squad attacks the target under your cursor.

The reasoning is general, not unique to us. Splitting fire across ten targets leaves ten damaged ships that all still shoot back. Killing one at a time removes its damage from the fight. Our squad command guide covers the squad types and duties in more detail.

Formations matter more than most players expect. The game has nine, each with a skill on the N key. Wedge is for breaking a line, Shield ring for surviving a swarm, Screen for putting a wall between you and heavy incoming fire, Pincer for flanking slow capitals, Hornet swarm for fighter-heavy waves. Our rule of thumb: if you are being swarmed, switch to Shield ring and use its skill (Aegis, 40 seconds between uses, restores a third of your shields and a quarter of every escort’s hull). If you are pushing a capital, Wedge or the Battering ram. The squads also rank up with kills, and a squad that dies loses nothing, so the game itself is telling you not to hoard them.

General principles that hold in most fleet games

  • Focus fire. Choose a target in order of threat, not distance. In our game, you can press X on a large capital to aim at its engines (it slows down), weapons (it hits less) or reactor (it takes more damage).
  • Control range. Artillery should fight from where the enemy cannot reach it, which is why the game’s heavy-handling ships need you to choose a firing position before the fight arrives.
  • Screen what is fragile. A support ship or a carrier should have something between it and the enemy. Supports in our game are advised to stay near their squads.
  • Save cooldowns for turning points. An ultimate used at the start of a fight, when nothing important is in range, is wasted. The generic capital power costs 70 energy (each ship’s own cooldown differs; the Starship’s Full Stack Strike takes 22 seconds), and Fleet repair costs 40 energy and restores 22 percent of hull to every allied ship, so use that when several ships are damaged, not one.
  • Do not chase. Fast fighters will drag you out of formation. Intercept and Hold duties exist so your squads handle them while you finish the big target.

Reading the first 60 seconds

You can often tell whether you will win within a minute. Look at four things, in this order. First, what is coming: a field of fighters needs interceptors and area damage, a few large capitals need lancers and focus. Second, who is shooting at you. On Veteran and above, the game says hunters go for your flagship, and on Nightmare two in five of every wing hunt you, so if your shields fall faster than your target’s, your orders are wrong, not your aim. Third, the reinforcement clock: reinforcements arrive every 70 seconds on Veteran, 55 on Nightmare and 40 on Cataclysm, so on the hardest setting a second group arrives before the first minute is over. If you are not winning by then, you are fighting two fleets. Fourth, your energy: if it is empty at 30 seconds you are casting too early.

A practical opening that follows from all this: issue your squad orders in the first ten seconds, hold your ultimate, take the first heavy-strike dodge you are offered, and decide at 30 seconds whether the plan is working. Changing a plan costs less at 30 seconds than at 90.

Choosing the right difficulty to learn on

Hard difficulties raise the enemy level floor: 11 on Commander, 15 on Veteran, 19 on Nightmare and 24 on Cataclysm. Our bot trial ran on Commander at level 20, and no ship won, so if you are failing at Commander with an unmodified starter ship, that is not unusual, and it is a good moment to drop one step, learn the orders, and come back. The difficulty and Threat guide explains what each step adds. If you have not picked a ship, our best ships for beginners page is the place to start, and we think carriers are the kindest to learn on, since their fighters take fire for you.

Questions people ask next

Do these tips work in other space games? The principles do (focus fire, screening, cooldown timing, range), but the keys and numbers in this article are about our game. For other fleet games, see games like Homeworld.

Is the bot trial proof that a ship is balanced? No, and we say so in the note. It is a crude ruler for finding outliers, based on two seeds per ship, and a difference of ten or twenty seconds between neighbors is noise. If a ship feels very different in your hands, trust your hands.

Want to test a plan without risk? The Who Would Win simulator lets two fleets fight each other.

Key sources: our own bot trial note and the game’s data and help text (stats, ranks, squad and formation rules, difficulty table) as of the current release.