Streamer-inspired matchup study: which CoM fantastic creature reliably beats the Colossus, and what the simulator runs revealed about the broader fantastic-creature meta.
Links: Master of Magic — Economic Analysis, Tier System and MIRR, BattleValue
Project: masterofmagic — scraper at scraper/scrape_com.py, simulator at simulator/battle.py, analysis at analyzer/com_vs_colossus.py
A streamer’s stack of invisible Great Drakes was getting shredded by an enemy Colossus in the endgame. Why the plan failed — and what should they have used instead?
The Colossus has Illusion Immunity, which in CoM also functions as see-invisible. The Drakes weren’t actually hidden, so the Colossus opened with its boulders (4 ammo × 25 strength) and then ate the survivors with melee 25 + First Strike. The plan was structural: the Drakes’ edge was supposed to be “Colossus can’t shoot what it can’t see.” Without that, it was naked stat-on-stat — the Capability Without Leverage thesis at its most vivid: the leverage (invisibility), not the capability (the Drakes), was the asset, and a single counter removed it.
The deeper “why” is meta: the streamer was applying original-MoM strategy to Caster of Magic.
The Drake-with-Invisibility play is a MoM-era solution: one buff makes one elite untouchable, you rampage. CoM removes that pillar (Illusion Immunity is now common on tier-2+ defenders) and replaces it with a war of attrition the player has to actually grind through.
This also explains the BV/Slot vs BV/Cost tension in the parent README: “9 maxed slingers might beat a Colossus per-cost, but 9 Colossi will whip 9 slingers.” The 9-unit stack cap is the binding constraint, not mana.
Scraped 44 fantastic creatures across all 5 realms (data/com/fantastic-units.json). Filtered to the 20 candidates ≥ 200 mana and ran each as the attacker against a naked Colossus through the Markov simulator. Faster unit initiates round 1; alternates from there.
Two simulator bugs surfaced during this analysis and were fixed:
BattleEngine.fight — initiator’s full attack always resolves first; defender’s FS is ignored. This matters because every flying or teleporting candidate can sidestep the Colossus’s First Strike just by being the melee initiator.doom_damage (bypasses defense + resist) and life_steal_distribution (per-figure 1d10 vs resist - penalty, with Magic/Death Immunity granting full immunity). Healing-from-Life-Steal back to the attacker is not yet modeled — defender’s HP loss is.A third “bug” turned out to be correct behavior: Missile Immunity does NOT block boulders per the wiki. Storm Giant’s MI doesn’t help vs the Colossus.
| Unit | Mana | Win % | Per 100mp | Why |
|---|---|---|---|---|
| Hydra | 480 | 96.1% | 20.0 | 9-figure × 10HP stack absorbs boulders, Fire Breath 6 chunks Colossus on contact. Regen 7 not modeled — true rate is higher. |
| Great Wyrm | 500 | 92.0% | 18.4 | Merging-teleport skips the boulder approach; Armor Piercing halves Colossus def 12 → 6. Poison 25 is wasted (Colossus has Poison Immunity). |
| Great Drake | 500 | 87.9% | 17.6 | Naked stats favor it cleanly. The streamer’s pick was right; the plan (relying on Invisibility) was wrong. |
| Demon Lord | 666 | 68.7% | 10.3 | Caster 40 self-buffs not modeled — likely undervalued. |
| Death Knights | 420 | 36.6% | 8.7 | Armor Piercing helps; Life Steal -5 vs resist 15 yields zero damage (max d10 = 10 vs mod resist 10). |
Everything below 30%: Arch Angel (16.4%), Sky Drake (9.5%), Behemoth (2.8%), Gorgons (2.7%), Djinn (0.7%), Stone Giant (0.7%), Wraiths (0.2%), and full-zero results for Doom Bat, Storm Giant, Efreet, Chaos Spawn, Angel, Night Stalker, Great Lizard, Unicorns.
A side experiment ran the Hydra vs every other ≥ 200mp candidate. Hydra goes 18-2. The two losses are structural and revealing:
Drakes → beat → Hydra (high-strength breath vs multi-figure stack;
each head rolls defense separately, all die)
Hydra → beats → Colossus (mass + breath vs single-figure boulder)
Drakes → beat → Colossus (naked stats; only loses if plan depends
on a counter the Colossus has)
So Great Drake is the strongest 1v1 unit on raw stats — beats both Hydra and Colossus. But Hydra is the safer reusable answer because it post-battle full-heals if it wins, and it’s not vulnerable to “the AI also brought Drakes.”
For the streamer’s specific Colossus problem: Great Drakes without Invisibility, or Hydras as the cheaper, more reusable option. Hydras at 480 mp also dodge the trap of being hard-countered by enemy Drakes that could later show up.
The unit-vs-unit lens is incomplete for CoM endgame. The probable real answer for high-resist single-figure threats like the Colossus is spell chains, not unit matchups: debuff-resistance → damage-or-kill spell, or terrain-targeting effects that bypass resistance entirely.
Crack’s Call is the canonical CoM example — targets the ground rather than the unit, ~25% chance to instantly kill any non-flying target regardless of resistance. That mechanic structurally hard-counters the Colossus (high res, ground-only) in a way no unit can. A future analysis should model spell economy: cost-per-expected-kill across the spell list, with debuff chains as multi-step pipelines.