Competitive balance research
Decision frame
Rule B gives each mech a class cap and each deployed team a 145 Weight cap. A roster has four mechs, but only three deploy, so the fourth acts as a revealed-map sideboard (I1–I2). Every surviving mech acts once per round, with two Quick actions or one Full action (E1–E6). Kills score 3; zones score each round (I4). Map symmetry repairs are in progress and the new study will freeze their final objective layouts. Heat, reactions, part loss, cover, and the location die add tactical costs that Weight does not describe.
An earlier, pre-burst diagnostic identified failure modes. In that historical sample on Foundry, 38.2% of ranged Fires deal no body damage, 66.0% of Cannon Fires deal none, and Bastion first deals damage in round 4.49 when it deals damage at all. Causeway has a confirmed side-associated mirror bias, and a Foundry bot can oscillate without attacking. These are historical local measurements, not the new study baseline, from docs/tuning/2026-09-09-balance-audit.md; they must remain separate from the observed rules below.
Observed mechanisms in other games
| Game | Observed rule or live-balancing practice | Design inference for Mecha Skirmish |
|---|---|---|
| Kill Team | A battle has four turning points. Players start with 2 command points (CP) and gain 1 each in the first Strategy phase. Later, the initiative player gains 1 and the other player gains 2. Ploys normally cost 1 CP; each ploy is limited to once per turning point except Command Re-roll. Strategy ploys are committed before alternating activations, while firefight ploys respond during them. If one team exhausts its operatives first, Counteract grants bounded 1 AP actions, with movement capped at 2 inches.[^1] | CP can create forecastable round plans and local reactions without changing chassis stats. The extra CP for the player without initiative compensates tempo, not score. Mecha Skirmish cannot copy it literally: under E2, equal three-mech teams can give the same player the first activation in successive rounds. |
| Kill Team generally fixes team size and selection rather than pricing every operative; specialist and weapon repeats are restricted. Games Workshop says it uses tournament data as an input, then applies concise changes to particular rules, cards, and abilities. Its April 2026 update weakened elite teams that exerted too much threat and control, while also improving opposing objective options and loosening restrictions on weaker teams.[^2][^3] | When a dominant package owes its value to activation control, objective denial, or a repeatable combo, a narrow legality or ability change is easier to reason about than a global Weight or damage change. Matchup counterplay and scoring access are valid balance levers alongside nerfs. | |
| Age of Sigmar | Each player receives 4 CP at the start of a battle round, chooses among 10 commands, and loses unused CP at round end. The player with fewer victory points gains 1 extra CP; having fewer auxiliary units can also grant 1.[^4] | Expiring round budgets prevent hoarding and make the amount of tactical interruption predictable. A score-based underdog point is a clean comeback experiment, but it should be tested against Enraged because two overlapping catch-up systems can overcorrect. |
| Infinity N5 | A standard list uses 300 Army Points and 6 Support Weapons Cost (SWC), permits at most 15 troopers, and limits each Combat Group to 10. Availability caps repeats. Regular deployed, non-null troopers each contribute an Order to their group; those pooled Orders may all be concentrated into one trooper. The second player can spend a Command Token to remove two first-turn Regular Orders when the opponent has more than ten total Orders.[^5][^6][^7] | This is layered budgeting: total power, premium equipment, copies, and action economy have separate controls. It also shows the danger of coupling unit survival to action income: early losses reduce both board presence and future Orders. Mecha Skirmish’s one-activation-per-survivor rule already has that snowball, so CP generation should not scale upward with surviving mechs. |
| Alpha Strike | The current quick-start framework gives each side 250 Point Value (PV). Unit cards price approximate battlefield strength, including the unit at a default Skill rating. It offers objective families with different role incentives: carrying objectives home, reaching the far edge, holding ground, destroying targets, or destroying units.[^8] | Weight should be checked against whole-kit battlefield contribution, not only isolated damage. Objective families can change which mobility, durability, range, and size profiles are valuable; a roster is strategically diverse only if its good choices change with the mission and map. |
| Shatterpoint | A strike team contains two squads. Each squad must contain one Primary, one Secondary, and one Supporting unit; the Primary supplies a Squad Point cap for the other two, and duplicate names are forbidden. Activations come from a shuffled unit Order Deck plus a wildcard, with one revealed card reservable for 1 Force. Victory is a best-of-three tug of war over changing active objectives; the loser of a struggle chooses between maps when the next card offers a choice.[^9] | Composition templates bound the value of flexible stars. Limited activation correction preserves uncertainty while giving agency. Changing active objectives and a bounded loser choice can keep one fortress position from deciding the whole game. |
These rules solve different game lengths and model counts. Their shared pattern is to cap separate sources of advantage and balance missions, terrain, and action economy alongside damage.
Testable hypotheses for this game
1. Add a small command economy before adding more passive comeback power
Run the current game as the control and compare these two rule prototypes:
- Tempo CP: start with 2 CP. Gain 1 at every round start; the player who does not take the first activation gains 1 more. Cap stored CP at 3. This is the closest Kill Team translation, adjusted for a seven-round game.
- Reset CP: receive 2 CP at each round start and discard the remainder at scoring. Starting in round 2, the player behind on score receives 1 more. A tied score gives neither player the extra point.
Begin with only two universal spends, each limited to once per mech activation: Command Re-roll costs 1 CP after a failed hit, offensive melee opportunity, or penetration die; keep the replacement. Emergency Step costs 1 CP after ordinary movement for up to 2 more movement points and 1 Heat; it does not increase Speed, Boost, or Charge. Do not allow Command Re-roll on Enraged’s automatic replacement, and do not let Emergency Step stack with Enraged movement. Those exclusions make the two resources observable instead of multiplicative.
The bot must choose whether to save or spend CP. Model at least four policies: immediate legal spend; expected-value threshold by target and part; reserve 1 CP for a high-value attack or objective swing; and score-aware risk seeking when behind. It must evaluate penetration rerolls against the actual struck part, value movement by reachable attacks and zone control, and know remaining rounds. Re-run policy grids for both players. A single greedy policy will make the CP rules look more or less powerful depending on its scripted preference.
Accept a CP variant for human testing if it lowers empty ranged Fires by a practically meaningful amount against the freshly measured baseline and brings Bastion’s first-damage timing forward by at least half a round, while changing median game length and wipe frequency by less than 10% relative to control. For comeback, report win probability after losing the first mech and after trailing at the end of round 2. The candidate should improve at least one trailing cohort without increasing the leading side’s conversion rate. Also report CP earned, spent, wasted, spend type, round, and outcome; rejection signals include one spend consuming more than 80% of CP or frequent cap loss.
2. Audit marginal team value, then add a premium budget only if Weight cannot price it
Weight prices physical parts, while pilots are free and team effects can multiply accuracy, action access, or reactions. For every legal part or pilot swap, measure marginal objective points, kills, body damage, attacks enabled, damage prevented, and activation denial across each map and side. Compare equal-Weight substitutions and packages such as Pilot + Head + weapon.
First tune Weight and class caps. If the same premium effects remain concentrated in most best-response teams across missions, prototype an Infinity-like secondary allowance: 3 Systems points per deployed team, charged only to effects that create free actions, extra attacks or reactions, activation manipulation, or team-wide modifiers. Do not charge ordinary weapons merely for being strong; Weight should continue to carry that job. Success means at least three distinct roster archetypes remain non-dominated across the map/mission matrix, and no single package is present in more than 70% of best-response rosters. These are prototype decision thresholds, not claims about all competitive games.
3. Make mission diversity part of the balance matrix
Keep Hold Zones as the baseline. Add one prototype each for carry/extract and timed relay, as H7 already anticipates. Use the same maps and rosters before changing terrain. Then test the four-to-three sideboard against revealed missions: track stand-pat versus swap frequency and whether the reserve changes role rather than merely replacing a weaker kit.
A useful mission pool should change roster ordering without creating a forced specialist. Require at least two different roster archetypes to lead somewhere in the mission-by-map matrix, while each archetype retains a playable counterplan outside its best mission. Measure objective points, contest turns, first contact, unreachable scoring states, and score margin. Borrow Shatterpoint’s loser choice only as a later BO3 variant: after losing game one, choose between two seeded next maps. It must not be mixed into the first mission experiment.
4. Treat side, terrain, and dominant packages as first-class balance dimensions
Do not average away side effects. An identical-roster mirror should produce no material side association before a roster verdict counts; report each side separately with uncertainty. Causeway currently fails that prerequisite. On every candidate ruleset, publish results by map, side, roster, first activator, first mech lost, and CP policy.
For live tuning, copy Kill Team’s narrow-change discipline. If one package dominates, identify whether its excess comes from damage, objective control, activation denial, safety, or low execution cost. Change the smallest causal rule or availability gate, then rerun the entire mission/map matrix. Global Armor, accuracy, or Weight changes should follow only when the failure is similarly global.
Sources
[^1]: Games Workshop. “Kill Team Lite Rules.” October 2024. [^2]: Games Workshop. “Kill Team Metawatch: The Shape of Things to Come.” January 19, 2023. [^3]: Games Workshop. “Kill Team Balance Update: Battleclade Gets Software Upgrades.” April 29, 2026. [^4]: Games Workshop. “React to Your Opponent With Powerful New Commands in NewAoS.” April 10, 2024. [^5]: Corvus Belli. “Army List.” Infinity N5 Wiki, updated for N5.2, October 2025. [^6]: Corvus Belli. “Orders and the Order Pool.” Infinity N5 Wiki. [^7]: Corvus Belli. “Command Tokens.” Infinity N5 Wiki. [^8]: Catalyst Game Labs. “Alpha Strike Quick-Start Rulebook, third printing.” 2026. [^9]: Atomic Mass Games. “Star Wars: Shatterpoint Core Rules.” January 9, 2026.