Doctrine for planning a battle worth watching: assign roles to both sides, run the pre-battle checklist, and turn physics chaos into recorded results.
A planned matchup has four roles. The game only asks for creatures and counts — the roles below are what separate a designed experiment from two crowds colliding.
One creature at a fixed count that stays constant across the whole session. Every other change is measured against it. The Elephants at 200 in the official example are a perfect anchor: park them, and walk the Tigers instead.
200 Elephants, all session
The side you actually move. Give it the interesting question — can a swarm beat giants, how much quality beats quantity — and change its count in meaningful steps so each run's result maps to a number.
Tigers at 500, 600, 700…
Terrain is a documented variable (the description confirms different maps), so name the map in every card and hold it constant until you are explicitly testing it. The same counts on a chokepoint map can crowd differently than in the open.
Record: map name, per card
That is you, every battle. Winner, margin impression, and any physics drama worth remembering — piling, trampling, a late comeback. Battles run in a solo instance (the Roblox API reports a 1-player max), so nobody else is writing it down.
One card per run
Run this before every battle. It is short because the game's own design is short — sides, counts, map, go.
Not 'are Kaijus strong' but 'do 10 Kaijus beat 1000 Mechs, and if not, at what count do they start?' A question with a number in it gets an answer with a number in it.
Choose which side holds still and write the challenger's next three counts now. The ladder stops you improvising mid-session and keeps every run single-variable.
Different maps are official; their effects are not documented anywhere. Holding the map constant makes your counts comparable; changing it deliberately is its own later experiment.
Wrong counts are the most common silent failure — 500 vs 20 reads like 500 vs 200 at a glance. Read both counters back before releasing the armies.
The description promises total chaos: the pile-up that stalls a swarm or the trample that snowballs is the why behind the result. Note it in one line.
One run of a near-draw is a coin toss. Two or three runs give you a trend; the card records all of them so the variance stays visible.
Both use only documented counts (official description, checked 2026-10-03) so you can reproduce every number.
Squad card: anchor 10 Kaijus vs challenger 1,000 Mechs (ratio 100 : 1), your usual map, best of three with replays on any one-run disagreement.
First experiment: hold the Kaijus and step the Mechs down until the favorite loses — the ratio where it flips is the squad's first recorded finding. The official description's own example. A hundred Mechs per Kaiju is the kind of extreme count ratio the sandbox is built to test.
Squad card: anchor 200 Elephants vs challenger 500 Tigers (ratio 2.5 : 1), your usual map, best of three with replays on any one-run disagreement.
First experiment: hold the Elephants and step the Tigers down until the favorite loses — the ratio where it flips is the squad's first recorded finding. The official description's second example. A closer 2.5-to-1 ratio, showing the game supports anything from fair fights to absurd ones.
A side's army: one creature type and a count. The official examples keep each side to a single creature — 1000 Mechs, 10 Kaijus, 500 Tigers, 200 Elephants — which makes every battle a clean comparison of two compositions at chosen counts.
Decide first whether fair is the goal. Ratio 1 (equal counts) is the fairest test; the official examples instead stage 2.5-to-1 and 100-to-1 handicaps to probe creature worth. Pick the ratio that answers your question, fix the map, and change nothing else between replays.
Because the simulation has variance and your memory will smooth it into false confidence. A recorded card per battle — sides, counts, map, winner, replay notes — is the only way close matchups become knowledge instead of anecdotes.
For knowledge, yes — same-creature battles measure the physics, not the roster. The documented examples all cross categories (machine vs kaiju, animal vs animal of a different class), and that contrast is what produces a transferable answer about creature worth.
