Local prototype guide

The Workshop and Play tabs now use the same catalog and engine as CLI games and batches.

Open the kit builder. Open Play and the melee trials.

Review the builder

1. Open Workshop. Choose a named core box.
2. Click an empty component outline or its labeled button. Choose parts in any order.
3. Browse manufacturers within each component. Review the old and new stats, then select Fit component. Arms and held equipment are independent choices.
4. Choose a pilot. Select equipment or an empty hand for each hand. A two-handed item occupies both hands and replaces conflicting equipment.
5. Complete all components. Save an independent library build or a roster copy. Drafts stay separate. The build and roster must meet all hard limits.
6. Preview pilot reassignment, including any teammate swap and faction/bond changes. The existing pilot name is the provisional public designation; a Workshop nickname stays private.

Drafts and rosters persist in this browser. Existing saved builds import their old arm loadouts through an explicit catalog mapping. Each old combined arm’s Weight was split between a structural arm and its suggested held item. Its body Size, Armor, and Durability remain unchanged. Default kit weights and the ranged worked example remain unchanged.

Core boxes retain the existing named kit identities. Foreman, Reaver, and Lantern currently share the Medium core stats. The box shows its core family; this pass does not invent distinct performance for those cores. Structural arm IDs retain their old names for saved-build compatibility. Display names describe the separate arm.

Review melee

Play includes three hotseat trials: twin melee weapons against a two-handed weapon, twin weapons against a shield, and a two-handed weapon with its carrying arm disabled. Each trial starts adjacent. Select Melee, then the enemy on the board.

Both pools appear before decisions. Select a successful weapon die and a body part. Declare the strike. The defender can select enough successful dice to parry, or take the hit. The same sequence pauses for human decisions when a bot initiates an exchange. The log records pools, declarations, responses, and damage.

Every exchange costs its initiator a Full action. The defender retains its activation. Both receive fresh pools for each new action. If one pool is exhausted, the other mech continues attacking. Damaged arms affect usable dice and equipment. A two-handed weapon remains usable with one working arm.

Workshop, Play, Battle and Batch remain mounted after first use, so switching tabs preserves the current session. Changes to a roster apply when starting a new game.

The owner confirmed that successful rolls should determine both attack and parry availability. Failed rolls fund neither. This shared resource is a settled design decision; the exact roll formula and tuning values remain provisional.

Explicit prototype defaults

These values are implementation choices for review. They are not owner-approved balance conclusions.

Existing damage-condition modifier dice are rolled once per weapon pool, before its d20s. Their penalty applies to each die in that pool. Positional hit modifiers do not apply.

The selected hit is not rolled again after declaration. Penetration, damage, critical and heat dice are rolled when a strike is accepted. Mid-exchange damage does not reroll existing successes. A disabled source loses its remaining dice; a usable two-handed weapon retains its dice. Defense-only leftovers do not prolong an exchange once both mechs lack attacks.

The defaults are stored in data/melee.json and in the equipment and arm entries of data/parts.json. The Tuner edits the separate arm and equipment catalogs. Weapons remain absent from the hit table. Hard build limits apply even when a scenario disables team composition checks.

What this pass does not establish

The bot uses a deterministic baseline: target Core and parry when enough dice remain, spending defense-only dice first. This policy tests the shared engine and human decision windows. It is not evidence that Core attacks, light specialists, or the three equipment styles are balanced. The old power score does not model interactive exchanges and is labeled as a legacy estimate.

The local match flow uses four locked roster entries per side and three deployed mechs. Both BO1 and bounded BO3 are playable. Scenario trials can use one to three mechs per side.

Battlefield interface

Starting a game opens the focused battlefield. The packet navigation gives way to the map, current decision, score and unit strip. Exit battlefield returns to the packet. The board keeps its size and position while controls change.

The HUD identifies the player who must decide now, including the responding player in melee. It shows the active mech, remaining movement, Quick OR Full budget, Heat, current score, zones held now and surviving Hull. Click a unit card to inspect it. Inspection does not activate the unit.

Begin Player A’s turn / Begin Player B’s turn is the prominent button above the board. It remains visible while inspecting units or reading Analysis. Melee and reactions use Respond as Player A/B in that same location; taking control does not spend an action. If several mechs are ready, choose one of the activation cards or a highlighted mech on the board. Issue orders returns from inspection to the active decision.

Orders lists each command once under Movement, Ranged, Melee, Defense or Systems. All orders is the default; the category buttons filter the list. Walk and Boost stay together, Throw and Break grapple sit with Grapple, and preparation, Overwatch and Guard are under Defense. Costs and unavailable reasons are visible without a hover; the question-mark button explains the rule. A selected order stays named in the HUD and command panel when changing filters or inspecting a unit, with Cancel order and Show selected order controls.

Weapon and ammunition controls are available in the ranged/defense context before selecting an order, so a player can switch away from an unusable gun. If damage removes the selected weapon, the selector and preview switch together to an available slot. A confirmation identifies the actor, target and cost. Melee identifies the first striker before committing and uses both pools rather than the ranged weapon selector. End activation stays at the bottom of the side panel while its contents scroll.

Melee shows the acting player's successful opportunities first. Source groups distinguish weapons, arms and defense-only dice. Choose a target part after the upfront roll. The confirmation stays visible while optional opponent pools and roll details scroll. A parry explains the required cost and shield adjustment.

Committed movement animates along the accepted path. Pending reactions stop at their legal decision point. Skip animation settles the display immediately; it does not change the commands or dice. Commands wait while displayed units catch up to the accepted state. Initial loads, reset, replay seeks and reduced motion show the correct position immediately. Fit units frames the engagement; Reset board shows the whole map. Follow actor and Part labels are optional and off by default. Deployment keeps the whole map visible.

The latest-effect strip shows hits, actual durability lost and Heat without requiring the log. Explain result opens Analysis. Its latest attack breakdown separates the upfront melee opportunity or ranged hit check from location, penetration versus Armor, damage, overflow, critical effects and disabled parts. Exact event data, filters and annotations remain available for rules analysis.

Game tools holds save/export and scenario tools plus display and playtest preferences. It is a modal dialog: background commands cannot receive keyboard input. For a managed local match, Game tools returns to the match's save/setup controls. Replay opens separately and leaves the live decision intact. Full unit sheets and the complete event log remain available below the board or in the expandable battlefield reference.

Moving across elevation

New games use paid vertical movement (H3). Each ordinary-leg step pays the destination terrain entry cost plus one movement point for every elevation level changed, uphill or downhill. Ordinary legs may cross any adjacent height the mech can afford. Working Jump legs ignore the elevation surcharge in both directions and pay only destination terrain entry; destroyed Jump legs use ordinary costs. No vertical step adds Heat.

Drops do not deal fall damage or by themselves cause Prone or end movement, including forced drops. Throw, collisions and other rules retain their own written effects. The route preview shows total movement cost and the elevation sequence; waypoints can select a particular affordable route. Existing saves without the paid-vertical capability restore their embedded legacy climb, Heat and fall behavior rather than silently changing an in-progress game.

The strategy interface review records the research, review passes and verification evidence.

Play a local match

1. In Play, choose Local match. Add a reserve if the Workshop roster has only three entries.
2. Choose BO1 or BO3 and a seed. Lock both public rosters.
3. Pass the device for each private team selection. Reveal both teams, roll d6 for ends, and alternate confirmed placements.
4. Alternate activations. Scoring boundaries refresh Reactions without changing whose activation comes next.
5. For Overwatch, select a ranged weapon and aim a cone at one elevation. A provisional 60° cone is in use while its exact geometry awaits owner follow-up. Respond to a committed path by choosing a reactor, source and firing point, or pass the remaining movement. Earlier path points become unavailable once passed.
6. BO3 offers one private reserve swap or stand pat, plus legal ammunition changes. The next map is visible. Parts and pilots remain locked. Draws consume a game slot; stop at two wins or three results.

Saves, replay and observations

Permanent offensive Enraged

New games enable permanent offensive Enraged (E9). An enemy-caused teammate death qualifies each survivor from its next own activation through the rest of the game. Pending means the next activation's benefit is not usable yet. Each own activation refreshes +2 ordinary movement allowance and one shared offensive reroll, shown as ready or spent. This does not increase Speed, Boost or Charge; ordinary restrictions apply. Repeated losses never stack or refill the current reroll. Both teams can qualify, even when tied or ahead in surviving units. Existing saved games retain their recorded duration.

The prototype automatically spends the reroll on the first failed ranged hit or failed offensive opportunity in a melee exchange the mech initiates. It keeps the replacement result and does not reroll the modifier dice. Reactions, defender pools, and parry-only dice are excluded. Inspect the attack breakdown or melee roll ledger to see the original die → replacement. No reroll choice is required. The active bonus and unused reroll expire when that activation ends.

Enemy Heat and lethal enemy pushes count; self-inflicted Heat and environmental deaths do not. New games use permanent offensive Enraged. Existing base-v4 saves resume with their temporary activation duration preserved, and existing permanent saves retain their duration. Earlier incompatible session versions still require the matching engine; import recovery keeps a rejected file downloadable. The retained comparison reports document the evidence and its limits.

Accepted commands and pending choices autosave on this device. Resume entries reopen the appropriate actor or privacy curtain. Named checkpoints are independent pinned copies. Export JSON to retain the exact setup, catalog, tuning, commands, events and notes. Export Markdown for a readable digest. Incompatible files remain unchanged; do not load them under different rules.

Replay supports event, activation and round steps, speed, a timeline, a pinned inspector and bookmarks. Mark a moment during play and add a note immediately or later. Optional categories and postgame feedback do not block play. Reports describe one observation and do not claim balance. Optional decision-time observation excludes hidden tabs, handoffs and replay; pause it while AFK.

Twenty ordinary completed saves are retained per save shelf. Unfinished and pinned saves are kept. Local files contain both players’ private setup choices; they are not a secrecy boundary.

Sandbox and help

Use the melee trials or focused Overwatch, Heat, rooftop and pilot-synergy presets. Create Scenario accepts legal library builds, placement, part damage, Heat and advanced statuses. Dice scripts override named stages. An exhausted or incompatible stage stops with an error; unconfigured stages retain normal seeded rolls.

Reset restores the exact initial setup. Branching retains the original history. Mirror rotates the initial terrain and positions 180° and exchanges sides, with the same seed and stage-script order. Existing maps have the even row count this transform needs. A mirrored trial is a new experiment, not a balance verdict.

Help and preferences contains a skippable quick guide, deeper Heat/synergy/sideboard guidance, keyboard controls, enlarged text, reduced motion and optional timing. Desktop/laptop and tablet landscape are the current targets.

Site and parry interface

The packet now uses shared navigation, colors, and typography across the home page, Workshop, Play, World registry, rules, and development pages. The illustrated rulebook is the main rules page. Existing rule links still work. The Evidence page groups dated reports and links to Battle, Batch, and Tuner. pnpm dev serves the complete site, with the bench at /bench/.

The melee panel labels the defending pool and each selectable die. It states when too few successful dice remain to parry. Pointer and keyboard checks cover one-die, two-die, and shield parries. Workshop and Play navigation preserves a pending exchange and its selected dice.

Verification

The ranged worked example, seeded game reproducibility, and controller/CLI parity remain checked. New checks cover incomplete drafts, independent equipment, hard limits, two-handed impairment, targeted strikes, shared-pool defense, fresh exchanges, and human responses during bot activations. The current test harness runs Chromium at desktop and tablet landscape sizes. Exact run results and remaining review findings belong in the session handoff.

Current demo tuning — 2026-09-09

Start a new match to test the burst-fire catalog; existing saves retain their embedded parts and weapon values. Ranged profiles now use multiple smaller hits and +2 printed Accuracy. Aimed Shot is a Full action that calls the same body part for every projectile. The hit/Pen/damage rolls remain independent, while modifier dice and weapon Heat/ammunition use are shared per Fire. Melee is unchanged. Conduit/Braced/Field arms now exchange capacity and protection rather than presenting same-weight strictly better choices.

Read the burst/lethality report and body tradeoff audit. Compare fresh-action reliability with repeated attack kill times; neither alone measures whole-game power. Aimed Core and concentrated fire are the main current playtest hotspots. Exports now include first death, cause when known, survivor-count changes and later activations so actual human games can inform lethality. The old Tuner P/Pwt estimate is explicitly uncalibrated.