NEDWaur In design
The project portal for the NED's LAN-party game: a forest brawler with a Warlock lava arena at its heart. Updated 2026-10-20 · 151 ideas logged.
The concept
NEDWaur is a chaotic top-down brawler for the NED's LAN crew. Everyone roams a big forest, fighting mobs and leveling up, and every so often the whole server gets yanked into a Warlock-style shrinking lava arena to settle things.
Forest phase. A large forested map with WASD top-down control. You fight mobs for XP and gold, level up, and decide whether to pick fights with other players or avoid them and farm.
Arena phase. Triggers (a timer, a usable item, more to come) teleport players onto a shrinking platform over lava, exactly like Warlock. Formats range from free-for-all to 1v1, team fights, and one-vs-everyone with god powers. Small fights stay short because everyone else is watching.
The vibe. Crisp controls and a high skill ceiling, with builds that stack in chaotic, unpredictable ways. The audience is the clan first, so imbalance and chaos are features for now. If it lands, maybe Steam later (and that is when copyright gets sorted).
Spelled NEDWaur: a play on the clan name NED's plus Warlock. This page is unlisted, not private. Nothing sensitive goes here.
Status at a glance
Phase
Prototype
v0.3: hex lava arena, Wizard + Xeno, bots (10/20)
Engine
Unity 6.6
URP + Mirror (host-authoritative)
Characters
14
10 from Mike, 4 from the character workshop
Players
~15
One LAN, no internet needed
Design principles
Everything here is proposed until Mike signs off. Mike's own principles come first, then the ones suggested by research.
From Mike
Crisp, clean gameplayProposed
Controls respond instantly and every hit, push, and death reads clearly on screen. If a moment feels mushy, fix the feel before adding content.
Why: Ryan, Rigby & Przybylski (2006) found that intuitive controls track with players' sense of competence, which predicts enjoyment. Game-feel work (Swink; Jonasson & Purho's 'Juice it or lose it') shows feedback is a big part of how good a game feels to play.
In NEDWaur: WASD movement with no input lag, hit-stop and a knockback arrow on every push, and lava damage that flashes the screen edge.
Rules anyone can understandProposed
A new player at the LAN should get the goal and their character's basic kit within one round. Depth comes from how the rules interact, not from how many rules there are.
Why: Spelunky is praised for applying the same small rule set to everything, so its emergent results stay predictable. Balatro's creator said he added the poker theme mainly to make the game easier to learn.
In NEDWaur: Forest: kill creeps, get gold, level up. Arena: don't touch lava, push others into it. Each character gets a one-line pitch on the select screen.
A high, rewarding skill ceiling (micro over autopilot)Proposed
Good play, meaning aim, dodging, positioning, timing, and push angles, should beat a better build piloted badly. Builds amplify skill; they don't replace it.
Why: SDT research ties enjoyment to felt competence. In Warlock-likes, knockback spells near lava already make positioning the core skill.
In NEDWaur: Skillshot pushes with travel time, dodge windows, and Mike's DDR brawler and mental-math wizard, which tie power directly to execution.
Chaos and randomness are featuresProposed
Every session should produce at least a few 'did that just happen?!' moments. Random events, wild abilities, and physics weirdness are welcome.
Why: Reviews of Gang Beasts and Spelunky 2 credit unscripted, systemic chaos as the main source of laughter and replay value in party and roguelike games.
In NEDWaur: Global wobbly aim, the gravity character's wormholes, pinball-ball abilities, and random arena triggers.
Builds compound in unpredictable ways (the Rounds rule)Proposed
Upgrades should multiply each other's effects, not just add to them, so two ordinary picks can combine into something absurd.
Why: The Steam page for Rounds advertises 11.2M+ powerup combinations from just 65+ cards. Vampire Survivors' evolutions and Isaac's item interactions are widely cited as the main reason players come back.
In NEDWaur: A 'bouncy' projectile card plus a 'pull on hit' card turns one spell into a ricocheting vacuum.
Harvest what the crew loves about DeadlockProposed
Find the specific Deadlock mechanics the clan enjoys and adapt them where they fit, rather than copying the whole game.
Why: The game is built for a specific group, and the crew already knows and likes these systems, so borrowing them lowers onboarding cost. (Mike's note: in Deadlock, farming creeps pays more than fighting players.)
In NEDWaur: Creep farming that out-earns player kills in the forest phase. The detailed Deadlock list belongs to the MOBA research dossier.
Spectated formats must be quickProposed
Any arena format that leaves people watching (1v1, 2v2, 1-vs-all) should finish fast so nobody sits idle for long.
Why: Party-game makers (Gaming Couch) argue rounds shouldn't last much over a minute. Warlock-likes shrink the arena every 10-30 seconds to force an ending.
In NEDWaur: A 1v1 duel shrinks faster than a free-for-all, with a hard sudden-death cap.
Necessary things should never feel like choresProposed
Creep farming, minion feeding, and puzzle-casting should be interesting decisions, not upkeep. If something is mandatory and boring, cut it or automate it.
Why: Sid Meier's 'a game is a series of interesting decisions'. Zagal et al. list grinding as a dark pattern that spends players' time against their interests.
In NEDWaur: The beastmaster's most basic upgrade both feeds and strengthens the beasts, so the tension only shows up on the juicy upgrades.
Fun first, balance laterProposed
During clan prototyping, an imbalanced but hilarious mechanic stays in and a balanced but boring one goes out. Real balancing waits for a possible Steam release.
Why: Players remember a fun game's broken builds as stories. Rounds and Vampire Survivors spread partly through people sharing absurdly overpowered builds.
In NEDWaur: The steam-train character can be wildly strong in the lava arena. Playtest first, then nerf.
Characters built for the crew's interestsProposed
Theme characters around what specific clan members love (motorcycles, Company of Heroes, Alien) so players see themselves in the roster.
Why: Relatedness is one of the three needs that independently predicted enjoyment of multiplayer games in Ryan et al. (2006). Spectator research also finds subject matter drives engagement.
In NEDWaur: The motorcycle momentum character, the CoH squad blob, and the Xenomorph.
Suggested by research
Randomize the offer, not the outcomeProposed
Put randomness where players can react to it, such as which cards appear, which events trigger, and map layout. Keep the core combat result (did my push land?) deterministic. Any percentage procs should use pseudo-random distribution so streaks stay rare.
Why: Burgun argues output randomness hides why you won or lost and slows learning, while input randomness creates decisions. Dota's PRD, inherited from Warcraft III, keeps chance effects from streaking.
In NEDWaur: Upgrade picks are a random 3-card draft, but spell damage and knockback are fixed. A '25% chance to freeze' uses PRD.
Upgrades change verbs, not just numbersProposed
At least half of all upgrades should change how something behaves (it bounces, splits, pulls, leaves lava) rather than giving +X%. Behavior changes are what compound into surprises.
Why: The Rounds cards people remember are behavior changes like Bouncy, Explosive Bullets, and Chase. Isaac's memorable items (Brimstone, Mom's Knife) change how you attack. A flat +10% never makes anyone laugh.
In NEDWaur: The 'Ricochet' card makes projectiles bounce off trees and arena edges; 'Magma Trail' makes your dash leave a short lava strip.
Get synergies from shared rules, not hand-authored pairsProposed
Give abilities and upgrades a small set of tags and hooks (projectile, on-hit, on-push, on-kill, pull, terrain) and have every upgrade work through them. Interactions then emerge on their own, without writing N-squared special cases.
Why: Spelunky gets emergent but predictable chaos from applying the same rules to everything. Isaac players report frustration when items that look like they should combine don't.
In NEDWaur: 'On-push: explode' works on the wizard's spell, the train's ram, and the gravity well's pull, because they all fire the same on-push event.
Chaos must be readableProposed
Show every player's picks on or near their character and in a spectator strip. Telegraph global events for 1-2 seconds. State stacking rules plainly and avoid hidden curves.
Why: Risk of Rain 2 players complain that items with similar wording stack in very different hidden ways. The Mario Kart critique is that catch-up nobody can see feels like cheating. Hokra's designer aimed for play that is clear to someone glancing across the room.
In NEDWaur: Card icons orbit each warlock in the arena. A klaxon and a screen tint warn of the 'wobbly aim' event before it hits.
Comebacks are visible and earnedProposed
Let the losing side catch up through choices and skill: losers pick an extra card and the leader carries a bounty. Don't use hidden rubber-banding or a blue shell that punishes the leader for leading.
Why: In Rounds the loser of each round picks a card, which reviewers say makes losing feel like an invitation, not a punishment. Riot's rule is that comebacks must be earned by beating someone stronger. The blue shell is the classic case of a comeback tool that wipes out good play.
In NEDWaur: The forest leader glows and carries a gold bounty. In arena best-of-N, the round loser drafts a card.
Chaos may steal control briefly, never for longProposed
Effects that take away control (inverted controls, wobbly aim, mind-control, pinball) last a few seconds, are telegraphed, and are funny to watch. Beyond that, chaos stops being fun and starts feeling like a punishment.
Why: SDT ties enjoyment to autonomy as well as competence. Burgun warns that randomness the player can't respond to causes the same problems as pure luck.
In NEDWaur: Inverted controls from a wormhole last 3-5 seconds with a visible countdown ring. Mind-swap ends early if the swapped body takes lava damage.
Spectators should have a stakeProposed
Anyone watching a duel should have something riding on it or some way to affect it, such as betting forest gold or voting on a hazard, so watching is part of the game.
Why: Spectator design work (Pirker, GDC 2019; UXPA) recommends making onlookers feel they win or lose along with the players. Hokra's designer found watching others play was as gripping as playing.
In NEDWaur: During a 1v1, everyone else bets gold on the winner. Eliminated players become ghosts who vote on the next shrink direction.
Make the big moments landProposed
Lava kills, multi-kills, comebacks, and absurd combos get hit-stop, slow motion, a sound sting, and a callout, so the room reacts together.
Why: 'Juice it or lose it' and Vlambeer's screenshake talk show that feedback effects, not systems, are much of what makes a game feel alive. For a LAN, shared reactions are the payoff.
In NEDWaur: When someone is pushed into lava there's a 0.3 s slow-mo, a splash, and a kill-feed line naming the push combo.
Easy in, deep outProposed
Every character has a baseline where pressing the obvious button does something useful, plus a mastery layer (timing, angles, cancels, risk-reward) that experts can push. Execution-heavy kits like the DDR brawler need a forgiving floor.
Why: Gang Beasts shows that a low skill requirement lets mixed groups play together, but it's criticized for wearing thin. Sirlin-style competitive design wants a high ceiling. A LAN with 15 people of different skill levels needs both.
In NEDWaur: The mental-math wizard gets easy problems for normal casts and optional hard ones for crits. Missing a DDR note weakens the combo instead of cancelling it.
Fun, not a casinoProposed
Variable rewards come from play inside a session (card drafts, creep drops, events), never from out-of-game currency, timers, or streak bonuses that punish you for stopping. A night ends cleanly at a natural stopping point.
Why: Zagal, Björk & Lewis define dark patterns as designs that work against players' interests, such as grinding, appointment play, and pay-to-skip. Vampire Survivors shows how strong gambling-derived hooks are, which is a reason to use them deliberately.
In NEDWaur: No login rewards and no persistent loot boxes. Unlocks, if they exist at all, are cosmetic or celebratory.
Games we're drawing from
| Game | Tier | What we take | What we skip | Why |
|---|---|---|---|---|
| Warlock (Warcraft III custom map) | Primary |
|
| The number-one inspiration and the arena phase's rulebook. This is the core arena phase of NEDWaur, and its descendants (Mages, Warlock Wars) confirm the pattern still works. [1] [2] |
| Deadlock (Valve) | Primary |
|
| The LAN crew's favorite game. Mike's LAN crew loves Deadlock. Players most consistently praise its movement, single-currency simplicity and active farming objects; those carry over to top-down FFA. Its lane and base structure does not. The clan already plays it and loves it. This agent did not research it; the MOBA dossier covers it. [1] [2] [3] [4] |
| Rounds (Landfall) | Primary |
|
| The model for compounding, unpredictable builds. Already covered by another agent; noted here only for how it connects to the economy recommendations. Mike named it directly. It is the clearest example of chaotic compounding builds that still feel fair, because the loser gets the choice. [1] [2] |
| Warcraft III | Reference |
|
| The forest phase's leveling model. The original Warlock map came from this ecosystem, and Mike's forest phase is explicitly Warcraft 3-style hero wandering. [1] [2] |
| StarCraft | Reference |
| — | The skill-ceiling model. |
| Company of Heroes | Reference |
| — | Some of the crew love it. |
| Dance Dance Revolution | Reference |
| — | Falling arrows as a combat input. |
| Alien / Alien: Earth | Reference |
| — | Mike was watching it. Copyright deferred. |
| Dota 2 (incl. Overthrow mode) | Suggested |
|
| Dota is the source of most creep and objective design, and Overthrow is Valve's own free-for-all MOBA mode, the closest precedent for 15 players in one forest. [1] [2] [3] |
| League of Legends | Suggested |
|
| Riot publishes the most explicit reasoning on comeback mechanics and readability. [1] [2] [3] |
| Path of Exile / Diablo | Suggested |
|
| They are the best-known answer to keeping repeated mob fights fresh, and they add the randomness Mike wants. [1] |
| Call of Duty: Warzone (Gulag) | Suggested |
|
| The closest mainstream precedent for watched 1v1s inside a larger match. [1] [2] |
| Fall Guys | Suggested |
|
| Mediatonic's level pillars are a ready-made rule for mixing Mike's skill-ceiling and chaos principles. [1] |
| Jackbox Party Packs | Suggested |
|
| The standard model for giving non-players a meaningful role in a party game. [1] |
| Vampire Survivors (poncle) | Suggested |
|
| It has the clearest 'discover a secret combo' payoff in the genre. Its designer came from the gambling industry, which is a useful reminder of where the healthy line sits. [1] |
| Risk of Rain 2 (Hopoo) | Suggested |
|
| It shows both the best case and the worst case for proc-based compounding. [1] |
| Brotato (Blobfish) | Suggested |
|
| Its rhythm of short fights and quick shops maps directly onto forest-to-arena cycles. [1] |
| Balatro (LocalThunk) | Suggested |
|
| It proves that 'add vs. multiply' layers create the 'it just keeps going' feeling, and that a familiar wrapper makes complex systems easy to learn. [1] [2] |
| Gang Beasts (Boneloaf) | Suggested |
|
| It is the model for laughter at a LAN, and it shows how physics chaos (Mike's water and gravity ideas) creates shared moments. [1] |
| The Binding of Isaac (McMillen) | Suggested |
|
| It is the reference point for chaotic, stacking attack modifiers. [1] [2] |
Mechanics
The current state of the game's rules. This section is the source of truth; the idea log below is the history.
| Area | Mechanic | Details | Status | Source |
|---|---|---|---|---|
| Core loop | Two phases: forest and lava arena | Players roam the forest (mobs, XP, levels, optional PvP). Triggers pull players into a Warlock-style shrinking lava arena, then everyone returns to the forest. | Proposed | Mike |
| Forest | Mobs, XP, and leveling | Warcraft III-style hero leveling from fighting mobs. Mob fighting should be necessary but interesting, never a chore. Detailed design pending the MOBA research. | Proposed | Mike |
| Arena triggers | Regular interval | Everyone is pulled into the arena on a timer. | Proposed | Mike |
| Arena triggers | Usable item | A player picks up and uses an item that summons players to the arena. | Proposed | Mike |
| Arena formats | FFA, 1v1, 2v2, 4v4, 1-vs-all | In 1-vs-all the lone player gets god powers. Fights involving only some players must resolve quickly so spectators don't get bored. | Proposed | Mike |
| Arena | Warlock rules inside the arena | Shrinking platform over lava, lava damage, knockback that grows as you take damage. Numbers from the Warlock dossier (the Dota 2 port by the original map team). | Proposed | Mike |
| Terrain | High ground, water, lava zones | Terrain should matter. Lava zones may appear in the forest too. | Proposed | Mike |
| Physics | Physics chaos: water and gravity | Flooding water that pushes players; gravity shifts and wells (see the Gravity Lord). | Proposed | Mike |
| Economy | Mobs vs PvP rewards | Undecided. Deadlock reportedly pays more for farming mobs than for kills. Awaiting research. | Open | Mike |
| Balance | Balance for both phases | Abilities can be overpowered in one phase and weak in the other. Tune through playtesting. | Proposed | Mike |
| Host settings | Everything tunable | Carried over from the Warlock rebuild: every gameplay number lives in data so the host can change it (shop time, rounds, lava damage, arena shrink, gold). | Decided | Mike |
Recommendations from research Proposed
| Area | Recommendation | What to do | Why |
|---|---|---|---|
| Match structure | Forest builds power, the arena keeps score | Make arena placements the only way to score match points; the forest is where you get strong enough to place well. Target a 30-35 minute match with an arena event about every 5 minutes and a double-points final free-for-all at the end. | One clear win condition keeps the rules easy to grasp, and it settles the farm-vs-fight question: both are only worth something because they set you up for the arena. MOBA match targets cluster around 30-35 minutes (Paragon targeted 35). [1] [2] |
| Economy | One currency that is both XP and gold | Use a single resource (Deadlock's 'souls' model) that both levels your hero and buys items. Every mob, kill and objective pays out in it. | Deadlock players single this out as what makes the game approachable: one number to watch instead of two. It also fits Mike's 'understandable rules' principle. [1] [2] |
| Economy | Farm pays more than fighting, about 65/35 | Tune so roughly 65% of a typical player's souls come from mobs and objectives and about 35% from PvP. Measure the real split in the first playtests and retune. | Dota, League and Deadlock all make creeps the bulk of income and kills the bonus. In a 15-player FFA, kill-heavy payouts would turn the game into hunting the weakest player. [1] [2] [3] |
| Economy | Kill bounties scale with how rich the victim is | Kill reward = a small base + a share of the victim's unbanked souls + a leader bonus for the top 3. Killing someone much poorer than you pays half base. | League rebuilt its bounty system around gold earned rather than kill streaks, and Dota's comeback gold scales with the victim's net worth. This makes hunting the leader attractive and bullying the weakest player pointless, which matters most in FFA. [1] [2] |
| Economy | Unbanked souls drop on death as stealable orbs | Fresh souls stay 'unbanked' for about 30 seconds, then bank automatically. Dying drops your unbanked souls as an orb anyone can grab, so third parties can swoop in. | Deadlock's soul orbs and denying create the risk-taking and vulture plays its players talk about. In a 15-player FFA this produces natural chaos and an incentive to fight a player who just cleared a big camp. [1] [2] |
| Economy | Earned catch-up, not handouts | Players below the lobby's median level get +25% XP from mobs; the top 3 get 15% less. Add one comeback objective per cycle that pays the trailing players more, scaled to how far behind they are. | Riot's rule is that comebacks must be earned or early leads stop mattering. Deadlock's 2026 Unstable Rift scales comeback bonuses linearly from 0% to 15% net-worth deficit, which is a clean, tunable model. [1] [2] [3] |
| Mobs | Four camp tiers, reward grows faster than risk | Small, medium, large and ancient camps, each with its own silhouette and color. Rewards should rise faster than danger (roughly 40 / 90 / 200 / 450 souls) so taking bigger risks is worth it. | Dota, Warcraft 3 and Deadlock all use tiered camps where tougher camps pay out more. It turns 'where do I farm' into a decision instead of a routine. [1] [2] [3] |
| Mobs | Random elite affixes on mobs | Some camps spawn an elite with 1 random modifier (2-3 later in the match), such as Vampiric, Splitting, Fire Trail or Soul Eater. Elites pay double and sometimes drop a one-time buff. | Diablo and Path of Exile keep farming fresh with randomized monster modifiers that escalate over time. This fits NEDWaur's chaos-and-randomness principle directly. [1] [2] |
| Mobs | Mobs that train arena skills | Give mobs telegraphed skillshots (fireballs, knockback charges, lava puddles) that use the same dodge-and-push rules as the arena. Clearing a camp cleanly, without taking hits, should pay a bonus. | This makes farming practice instead of a chore: skilled players clear faster and arrive at the arena warmed up. Telegraphed wind-ups are what make mob combat readable and fair (Valheim). [1] [2] |
| Mobs | Skill-timed farming objects | Scatter breakables and 'shrines' that pay more for timed play, modeled on Deadlock's Sinner's Sacrifice and Golden Statues: a well-timed heavy hit pays 3-4x a sloppy one. | Deadlock turns farming into a small timing game that rewards execution, which is the high-skill-ceiling principle applied to farming. [1] [2] |
| Mobs | Predictable respawns, with stacking allowed | Camps respawn on visible minute marks, and the camp only spawns if its spawn zone is empty. That lets skilled players stack camps (pull them out just before the respawn) for double payouts. | Dota's minute-mark spawning creates route planning and a stacking skill, and creeps blocking each other caps it naturally. Riot added jungle timers for clarity. [1] [2] |
| Mobs | Mobs stay relevant late | Camps upgrade a tier at set times (for example 10 and 20 minutes), and the ancient camps unlock mid-match, so farming still matters at minute 25. | Design writing on creep camps lists 'value fades mid-game' as the main way they become pointless; Deadlock upgrades its statue buffs at 10 minutes for this reason. [1] [2] |
| Mobs | One central boss with an arena-shaped reward | A Roshan / Mid-Boss style boss in the forest center with a randomized 7-10 minute respawn. It drops a single claimable token, for example one 'lava save' that pulls you back onto the platform once in the next arena. | Roshan works because one reward changes fights, the respawn window is uncertain, and it pulls everyone to one spot. Deadlock's Rejuvenator lets either team steal the drop, which fits FFA. [1] [2] |
| Mobs | Neutral mobs hit whoever provokes them | Camps aggro on whoever damages them, act as one unit, and leash and reset if pulled too far. Put the best camps at chokepoints so farming and third-party fights overlap. | Dota and Deadlock neutrals attack any provoker, which makes contested space; Warcraft 3 camps act as a group. Leashing prevents kiting exploits. [1] [2] |
| Arena triggers | Countdown-driven arena calls | The arena fires on a visible global timer (about every 5 minutes) with a 15-second warning, plus rarer player-triggered items (a Duel Challenge, a Summoning Horn). The format is announced at the warning so players can prepare. | Battle royale storms show that visible, predictable pressure creates fights while still leaving room for side goals. Fortnite tunes wait and close times separately for each phase. [1] [2] |
| Arena triggers | Small formats run on a hard 60-90 second clock | 1v1 and 2v2 arenas start smaller and shrink to nothing at 60 seconds (90 for 2v2), followed by sudden death. Winners gain a permanent stat stack like Legion Commander's Duel. | Mike's explicit worry is watchers getting bored. Duel's permanent damage for the winner is what makes a short 1v1 feel high-stakes, and Deadlock's Street Brawl proves 3-minute rounds work. [1] [2] |
| Arena triggers | The forest keeps running during small duels | During 1v1 and 2v2, everyone else keeps playing in the forest while the duel shows in a picture-in-picture window. Duelists get an arena payout to make up for the lost farm time. Only FFA and 1-vs-all pull everyone in. | Design writing on elimination agrees that sitting out is the main problem. Keeping 11-13 people playing solves it outright and leaves the LAN room yelling at the side screen. [1] [2] |
| Arena triggers | Spectators bet and heckle, but cannot decide the fight | Watchers can bet some of their unbanked souls on the duel winner (pooled odds) and throw a limited number of harmless 'rocks' or vote on one arena modifier. | Warzone's Gulag spectators throwing rocks and Jackbox's audience voting give watchers a role. The Gulag spectator-intrusion bugs show the limit: watchers must never be able to swing the result. [1] [2] [3] |
| Arena triggers | Pre-arena 1-of-3 draft with one reroll | Before each arena, each player picks 1 of 3 random arena-only upgrades, with one reroll. Trailing players see higher-rarity offers. | Deadlock's Street Brawl does exactly this, and it is the Rounds-style compounding chaos Mike wants. Weighting offers toward trailing players is an earned catch-up. [1] [2] |
| Match structure | A gold fountain pulls the 15 players together | A rotating 'hot zone' in the forest pays bonus souls and XP while you stand in it, and it moves after each arena. It pushes encounters on a big map without forcing them. | Dota's Overthrow FFA mode uses a central ring that spits gold and XP to pull everyone into fights. On a map big enough for 15 players, people would otherwise rarely meet. [1] [2] |
| Other | Universal movement kit for every hero | Every hero gets a dash and a parry or reflect on shared keys, in addition to their unique kit. A successful parry stuns and makes the target take more damage. | Deadlock players consistently name movement and universal mechanics (dash, slide, parry) as the main hook. They give everyone a skill ceiling without bloating any one kit. [1] [2] |
| Economy | Short shop with a fixed price ladder and imbues | Four item tiers at fixed doubling prices (e.g. 500/1000/2000/4000), components discount their upgrade, and 'imbue' items bind to one ability you choose. Keep the shop to about 30-40 items for the clan version. | Deadlock's 800/1600/3200/6400 ladder is easy to read, and imbues make builds compound in surprising ways. Players also complain about Deadlock's item overload, so keep the shop small. [1] [2] [3] |
| Readability | Clarity hierarchy: players > mob attacks > scenery | Strong silhouettes per hero, a unique color ring for each of the 15 players, and every effect has a visible cause. Player ability VFX always draw above mob VFX, and a crown marks the soul leader. | Riot's clarity pillars (recognizable silhouettes, a hierarchy of effects, every effect has a cause) matter even more with 15 players and chaos abilities like wobbly aim or gravity flips. [1] [2] |
| Readability | Show camp timers and bounties on the map | Show camp respawn timers, the boss window and each player's current bounty on the map or HUD. | Riot added jungle timers specifically for clarity, and visible bounties make the leader a shared target. [1] [2] |
Characters
Every concept, with a proposed kit. Tags show who came up with each piece. Below each kit, the ability pool collects candidate abilities with the slot left TBD; we trim to a final kit later. Nothing is final; this is the menu for the first playtest.
Steam Train · Motorcycle Rider · Beastmaster (Necromancer skin) · Tunneler (Mole Sapper) · Merchant · Math Wizard · Rhythm Brawler · Company Squad (U.S. Army) · Gravity Lord · Xenomorph · The Barkeep · The Harpooner · Pinball Wizard · Dam Beaver
Steam TrainMikeConcept
An armored locomotive with an artillery car that is slow to start, impossible to stop, and ruins anyone standing on the tracks.
| Slot | Ability | What it does | From |
|---|---|---|---|
| Basic | Auto-Gun | A turret on the tender auto-fires at the nearest enemy in range. Low damage, no aiming, so the driver can focus on steering. | Mike |
| Q | Artillery Shell | Lobbed shell with a slow, arcing travel time and a ground marker. Very high damage and knockback at the center, falling off toward the edge of a large AOE. | Mike |
| W | Lay Track | Drops a short temporary rail line ahead. While on it the train gains speed and turns perfectly along the curve; anyone standing on the rails when the train arrives eats a full-speed ram. | Agent: Characters |
| E | Steam Vent | Blasts steam out both sides: a short knockback and a screen fog on anyone hit. Doubles as an emergency brake that cuts speed by one lever notch instantly. | Agent: Characters |
| R (ultimate) | Full Steam Ahead! | Voice line plays, the lever slams to max, and for 6 seconds the train ignores acceleration lag, is immune to knockback, plows trees, and its ram knockback doubles. | Mike |
| Passive | Momentum and Steam Lever | Speed is set by a lever with notches (Reverse, Stop, Quarter, Half, Full) on W/S. Acceleration and braking lag behind the lever. Ram damage and knockback scale with current speed. Plows through trees and tree stands at speed. | Mike |
Ability pool: candidates, slot TBD
| Slot | Ability | What it does | From |
|---|---|---|---|
| TBD | Coal Shovel | Spend gold to shovel coal: max steam and ram knockback go up for 10 seconds. Turns money into a terrifying charge. | Claude |
| TBD | Train Whistle | A telegraphed blast of the whistle that scatters creeps and gives players a one-second warning to dive off the tracks. Pure intimidation, low damage. | Claude |
Open questions
- Which abilities from the pool make the final kit, and in which slots?
- Should lava damage the train or fuel it (crossing lava refills steam but costs HP)?
- Merge with the Motorcycle as one vehicle family, or keep both? (Recommendation: keep both, share the momentum code.)
- Can other players ride on the train car? A clan-party bonus if it is cheap to build.
- Can players ride on top of the train? (Funny, but maybe a nightmare.)
▸ Forest vs arena, fun, risks, tuning
Forest phase: Excellent. Trees become a speed check instead of a wall, so the train carves its own roads through the map and leaves visible scars other players can follow. It farms creep camps by driving through them and shelling the stragglers. Weak against anyone who gets beside or behind it, where the ram cannot reach.
Lava arena: Flagged: likely to break the arena, in both directions. On a shrinking platform a high-momentum vehicle tends to drive itself into the lava, and when it does connect, a full-speed ram on a small platform is an instant kill. Needs an arena variant (see fixes).
Skill ceiling: Reading where people will be in three seconds, setting the lever early so top speed arrives as the target commits, using Lay Track to bend a line around a dodge, and landing artillery where the victim is forced to flee from the ram.
Inspiration: Mike's design; Spirit Breaker charge (Dota 2), Mario Kart drift-boost thinking for the lever lag, SC2 siege tanks for the artillery.
Why it's fun
- The lever UI and the 'Full steam ahead!' line are a pure party moment that the whole room will hear.
- Committing to a line and being right feels enormous; the victim at least saw it coming from a mile away.
- Shelling a fight from long range with a slow arcing shell rewards prediction.
Risks
- Self-destruction in the arena feels like a controls failure rather than a mistake.
- Getting one-shot by a ram you were trapped against can feel unfair if the train came from off-screen.
- Hard to play well at all for anyone who is not comfortable with momentum, so it may sit unpicked.
Fixes
- Arena variant: a turntable (spin in place when stopped), lever capped at Half until the R is used, and a shorter acceleration lag.
- Heavy mass: the train takes 50% knockback, so it is an immovable object when parked and a missile when moving.
- Audio and visual tell: a loud whistle and a smoke column that is visible off-screen whenever speed is above Half.
- Ram damage capped at a percentage of max HP in the forest; knockback is the real payoff.
Tuning knobs
- Acceleration lag per lever notch
- Turn radius as a function of speed
- Ram damage cap and knockback multiplier
- Artillery travel time, AOE radius and center falloff
- Knockback taken (mass)
- Arena lever cap
Motorcycle RiderMikeConcept
A biker who lives at full throttle, drifting through the trees and running people down, but who cannot stop on a dime.
| Slot | Ability | What it does | From |
|---|---|---|---|
| Basic | Drive-By | Sidearm fired sideways from the bike toward the cursor. Usable at any speed, so the rider never has to stop to fight. | Agent: Characters |
| Q | Ramp Jump | Pops a wheelie and jumps a short distance. Airborne riders clear lava gaps, low walls, and ground-targeted spells. | Agent: Characters |
| W | Nitro | A short burst of speed. Hitting someone during Nitro deals ram damage and knockback. | Agent: Characters |
| E | Chain Lasso | Throws a chain that snags a target and drags them behind the bike for 1.5 seconds. Drag them into trees, creeps, or lava. | Agent: Characters |
| R (ultimate) | Wall of Death | The rider circles a point at speed, leaving a ring of fire for 5 seconds. Anyone inside is trapped; crossing the ring burns. | Agent: Characters |
| Passive | Momentum and Drift | Momentum and a wide turn radius (Mike). Holding Shift drifts: a tight turn that builds a drift charge, released as a mini-boost (Mario Kart). Ram damage scales with speed. When the bike is destroyed the rider bails out on foot with a small HP pool for a second life. | Mike (momentum, low turn speed); Agent: Characters (drift, bail-out) |
Ability pool: candidates, slot TBD
| Slot | Ability | What it does | From |
|---|---|---|---|
| TBD | Wheelie | Pop the front wheel up to block projectiles from the front, at the cost of turning even worse. | Claude |
| TBD | Burnout | Spin in place, leaving a smoke ring that hides you and anyone inside it for a few seconds. | Claude |
Open questions
- Which abilities from the pool make the final kit, and in which slots?
- Should Mike's 'overlap' idea mean a merge (rider is the train's outrider) or just shared code?
- Who in the clan is the motorcycle fan, and what do they want it to feel like?
▸ Forest vs arena, fun, risks, tuning
Forest phase: A raider. Fast travel between camps and the best map presence on the roster; it decides which fights happen. Trees punish bad lines, which is where the drift skill shows.
Lava arena: Workable, with care. Ramp Jump over lava gaps is the best escape in the game, and Chain Lasso is a lava-delivery tool. Risk is the same as the train: momentum on a small platform. Lighter weight and drift make it far more controllable.
Skill ceiling: Chaining drift boosts so the rider never drops below top speed, lasso angles that use trees and lava, and timing Ramp Jump to dodge a knockback spell mid-air.
Inspiration: Mike's design; Mario Kart drift boosting, Road Rash, Junker Queen / Roadhog chain pull (Overwatch).
Why it's fun
- Drift-boosting through a forest is satisfying on its own, even with no enemy around.
- Lasso-dragging someone into lava is a highlight-reel play.
- Bail-out gives a comedic second chance instead of a flat death.
Risks
- Overlap with the Steam Train (both momentum vehicles).
- Fast raiders that pick off farming players can make the forest feel unsafe for beginners.
- Low turn speed reads as clunky if the drift is not tuned tightly.
Fixes
- Differentiate by weight: the bike is light, agile, hit-and-run and takes extra knockback; the train is heavy, slow, takes less knockback.
- Engine noise audible from a distance so farmers get warning.
- Drift is the turn: tune it first, before anything else in the kit.
Tuning knobs
- Turn radius and drift turn rate
- Drift charge per second and boost size
- Bail-out HP
- Lasso drag duration
- Knockback taken
Beastmaster (Necromancer skin)MikeConcept
A handler with a pack of hungry beasts who are terrifying when fed and terrifying for you when they are not.
| Slot | Ability | What it does | From |
|---|---|---|---|
| Basic | Whip Crack | Short-range lash. Whatever the Beastmaster hits becomes the pack's focus target. | Agent: Characters |
| Q | Sic 'Em | The whole pack lunges at a target point or enemy with a burst of speed. | Agent: Characters |
| W | Toss Meat | Throw a chunk of meat from the wagon. Your beasts eat it for food, but any hungry beast (another Beastmaster's pack, or wild creeps) will chase it too. Land it on an enemy to make the beasts attack them. | Agent: Characters |
| E | Call Beast | Summon a new beast into an open pack slot. More beasts means more power and more hunger (Mike). | Mike |
| R (ultimate) | Release the Hounds | Deliberately let the pack go feral at a target area for 8 seconds: big damage bonus, they attack anything except you, and hunger resets afterwards. Converts the risk into a weapon. | Agent: Characters |
| Passive | Meat Wagon | One resource (meat, gained from creep corpses) is split between feeding and upgrades (Mike). The pack's hunger is one shared bar. Starve them and they go feral in stages: sluggish, then attacking the nearest anything, then turning on you and eating you (Mike). The basic upgrade improves both food capacity and beast strength; juicier upgrades force a choice (Mike). | Mike; Agent: Characters (shared hunger bar, feral stages, meat currency) |
Ability pool: candidates, slot TBD
| Slot | Ability | What it does | From |
|---|---|---|---|
| TBD | Starving Frenzy | Deliberately skip a feeding: beasts go berserk for 8 seconds with big damage, then roll a chance to turn on you. The feast-or-famine gamble as a button. | Claude |
| TBD | Pack Tactics | Passive idea: beasts hit harder when three or more bite the same target, rewarding focus over spreading out. | Claude |
Open questions
- Which abilities from the pool make the final kit, and in which slots?
- Is the currency meat, corpses, or XP? (Recommendation: meat from corpses, so it never competes with XP.)
- Is Necromancer a skin of the same kit or its own character?
- Can beasts be killed permanently, or do they respawn for meat?
- Do wild creeps also chase thrown meat, or only other Beastmasters' packs?
▸ Forest vs arena, fun, risks, tuning
Forest phase: Strong and central. The forest is a buffet: the pack kills creeps, the corpses become meat, the meat feeds the pack. The tension Mike wants lives here, deciding whether to spend this batch of meat on a stronger pack or a fuller wagon.
Lava arena: Weaker unless tuned. Beasts get knocked into lava, and a full pack on a small platform crowds everything. Feral beasts on a shrinking platform could be the funniest thing in the game, though.
Skill ceiling: Running the hunger bar close to empty for maximum upgrades, timing Release the Hounds so the feral phase becomes a weapon, and using Toss Meat to hijack enemy packs or pull creeps into an enemy.
Inspiration: Mike's design; Lone Druid / Beastmaster (Dota 2), Darkest Dungeon's stress tension, Don't Starve hunger.
Why it's fun
- Feast-or-famine risk management with a visible payoff (Mike's design is the fun).
- Toss Meat lets you weaponize other players' beasts and wild creeps, which creates chaos without extra rules.
- Being eaten by your own pack is a story people will tell.
Risks
- Feeding becomes a chore if it needs constant attention.
- Too many AI units make fights unreadable.
- A feral pack that kills a nearby innocent player feels random to the victim.
Fixes
- Meat auto-collects when the pack kills things and auto-feeds from the wagon; the only decision is spending, never clicking.
- One hunger bar for the whole pack, shown over the Beastmaster's head so enemies can see when to bait them.
- Feral stages are telegraphed: beasts glow red and growl for 5 seconds before turning.
- Arena: pack capped at 3 beasts; drowned beasts drop cooked meat.
Tuning knobs
- Hunger drain per beast
- Meat per corpse
- Upgrade costs and the 'both' starter upgrade
- Time between feral stages
- Pack cap (forest vs arena)
Tunneler (Mole Sapper)MikeConcept
A dwarven sapper who goes underground, reshapes the terrain, and pops up where nobody expects.
| Slot | Ability | What it does | From |
|---|---|---|---|
| Basic | Pickaxe | Melee swing. Extra damage to structures, walls, and ore veins. | Agent: Characters |
| Q | Burrow | Dive underground and travel quickly, leaving a visible moving mound (Mike's tunnel idea). Surfacing knocks up everyone nearby. AOE damage on the mound forces you out. | Mike (tunneling); Agent: Characters (mound counterplay) |
| W | Tunnel Network | Place a tunnel entrance. Entrances persist and link to each other; the Tunneler travels between them instantly, and others can use them too (slower). | Agent: Characters |
| E | Rampart | Raise a short dirt wall that blocks movement and projectiles. In the arena it can pin someone against the edge. | Agent: Characters |
| R (ultimate) | Undermine | Collapse an area of ground. In the forest it leaves a pit that slows and damages. In the arena it drops that chunk of the platform into lava early. | Agent: Characters |
| Passive | Ore Sense | Sees ore veins in the forest. Mining them gives bonus gold, giving the Tunneler its own economy route. | Agent: Characters |
Ability pool: candidates, slot TBD
| Slot | Ability | What it does | From |
|---|---|---|---|
| TBD | Cave-In | Collapse a tunnel under enemies to leave a pit. In the lava arena the pit briefly opens to lava, so it becomes a trap. | Claude |
Open questions
- Which abilities from the pool make the final kit, and in which slots?
- Should Mike's tunneling just live on the Xenomorph (vents) instead of a separate character?
- Can other players destroy tunnel entrances?
▸ Forest vs arena, fun, risks, tuning
Forest phase: Strong. Tunnel networks make it the map's logistician; burrow ambushes are great; ore veins give a farming route that does not compete with creep camps.
Lava arena: Strong, maybe too strong. Undermine accelerates the shrinking platform on demand, which is a thrilling but dominant power. Burrowing must not work over lava.
Skill ceiling: Placing tunnel networks that win rotations, walling enemies into lava with Rampart, and timing Undermine for when opponents are standing on it.
Inspiration: Mike's idea; Rek'Sai tunnels (League), Earthshaker Fissure (Dota 2), Nydus Worm (StarCraft).
Why it's fun
- Popping up under a fight is a classic surprise.
- Reshaping the arena is a unique power nobody else has.
- Tunnel networks are a toy the whole lobby can use.
Risks
- Overlaps with the Gravity Lord's teleport and wormholes, and with Xenomorph vent travel.
- Untargetable while burrowed is frustrating if it lasts long.
- Undermine could end arena rounds too fast.
Fixes
- Give it a distinct identity as the terrain changer (walls, pits, platform collapse) rather than just a teleporter.
- Burrow is never untargetable: the mound can be hit.
- Undermine in the arena has a 2-second crack-and-rumble warning and cannot be cast in the final shrink stage.
Tuning knobs
- Burrow speed and max duration
- Tunnel entrance count and duration
- Rampart length and HP
- Undermine radius and warning time
- Ore vein density
MerchantMikeConcept
A traveling trader who wins by making deals, not by fighting, and gets rich off everyone else's war.
| Slot | Ability | What it does | From |
|---|---|---|---|
| Basic | Coin Toss | Throws coins. Weak, but deals bonus damage proportional to gold held. | Agent: Characters |
| Q | Contract | Offer a non-aggression pact to a player: they pay you gold and you become invulnerable to each other for a duration (Mike). One-button accept/decline. Breaking a pact early (by attacking) marks the breaker as an Oathbreaker: revealed on the map and drops gold. | Mike (paid pact); Agent: Characters (Oathbreaker penalty) |
| W | Pop-Up Shop | Place a market stall anywhere. Any player can buy shop items from it without walking back to base, at a markup; the Merchant pockets the markup. | Agent: Characters |
| E | Hired Muscle | Rent a mercenary creep for 20 seconds. A short rental, so it does not compete with the Beastmaster's or the Squad's permanent armies. | Mike (buying minions); Agent: Characters (rental framing) |
| R (ultimate) | Gold Rush | Scatter a pile of gold coins across an area. Everyone scrambles; the Merchant collects coins at double value. Starts fights between other players. | Agent: Characters |
| Passive | Interest | Earns a small percentage of gold held every 30 seconds, capped. | Agent: Characters |
Ability pool: candidates, slot TBD
| Slot | Ability | What it does | From |
|---|---|---|---|
| TBD | Bounty | Mark a player: they're pinged on everyone's map every 3 seconds and wear a gold coin. Whoever kills them (lava counts) takes the bounty. Claude's take: the Merchant stakes their own gold and anyone can chip in; the Merchant earns a ~50% broker's fee only if someone else gets the kill; if the target survives ~40 s they keep the pot; no repeat bounty on the same player back to back. | Mike |
| TBD | Insurance Policy | Sell a player insurance: if they die within 60 seconds they keep their dropped souls and the Merchant pays out; if they survive, the Merchant keeps the premium. | Claude |
| TBD | Black Friday | Everything in the shop is 50% off for everyone for 10 seconds; the Merchant takes a cut of every sale. Causes a stampede. | Claude |
Open questions
- Which abilities from the pool make the final kit, and in which slots?
- Can two non-Merchant players make pacts with each other through the Merchant (as a broker fee)?
- Does pact gold count as the payer's loss in the end-of-game score?
- Bounty: how big is the stake, how long the timer, and does it carry into the arena if an arena call fires mid-bounty?
▸ Forest vs arena, fun, risks, tuning
Forest phase: The Merchant's best phase. Contracts let it farm in peace, Pop-Up Shops tax everyone, and Gold Rush starts fights it does not have to join.
Lava arena: Turns into politics. In FFA, the Merchant can sell protection or pay someone to target a rival. In 1v1 it is weak, so give it 'pay to win' arena spending: gold converts to a coin shield.
Skill ceiling: Reading who needs safety and pricing accordingly, positioning Pop-Up Shops on traffic routes, and timing Gold Rush to start a fight between two rivals.
Inspiration: Mike's idea; Alchemist and Bounty Hunter (Dota 2), Diplomacy and Sea of Thieves alliances, Gangplank's economy (League).
Why it's fun
- Social play the clan will love: bribes, betrayals, and Oathbreakers.
- A character whose win condition is economic is new to this genre.
- Gold Rush makes a whole lobby scramble.
Risks
- Pacts could let the strongest player buy safety from everyone.
- Weak in 1v1 arena formats where there is nobody to deal with.
- Pact popups interrupt the other player mid-fight.
Fixes
- Cap active pacts at 2, and pacts expire when the arena triggers (or carry over in FFA only).
- Arena coin shield: spend gold for a damage shield and knockback resistance.
- Pact offers appear as a small corner prompt that auto-declines after 5 seconds, never a modal.
Tuning knobs
- Pact price floor and duration
- Shop markup percentage
- Interest rate and cap
- Merc rental cost and length
- Oathbreaker penalty
Math WizardMikeConcept
A professor-mage whose spells only work if you can do the math, and who makes enemies take a pop quiz to dodge.
| Slot | Ability | What it does | From |
|---|---|---|---|
| Basic | Chalk Dart | A plain projectile with no quiz, so the Wizard is never helpless. | Agent: Characters |
| Q | Arcane Equation | Fireball-style knockback spell. Casting it pops a mental-math problem next to the character; correct answer casts, wrong answer fizzles (Mike). Holding Q longer requests a harder problem, which crits on success (Mike's 'hard one goes for critical'). | Mike |
| W | Pop Quiz | Targeted spell: the enemy must solve a problem within 3 seconds. Fail and they take full damage; succeed and they dodge or take half (Mike). | Mike |
| E | Proof Step | Blink. Requires tracing a tiny maze with the cursor (Mike's 'tiny puzzle' idea); a clean trace blinks further. | Mike (cursor puzzle); Agent: Characters (as a blink) |
| R (ultimate) | Final Exam | A large area. Every enemy inside gets the same problem at once; everyone who fails is stunned and knocked back. A great spectator moment. | Agent: Characters |
| Passive | Honor Roll | Each consecutive correct answer reduces cooldowns slightly; a wrong answer resets the streak. | Agent: Characters |
Ability pool: candidates, slot TBD
| Slot | Ability | What it does | From |
|---|---|---|---|
| TBD | Lightning Round | A global pop quiz: everyone gets the same problem, the fastest correct answer gets a buff and the slowest gets a tiny debuff. A whole-room moment. | Claude |
| TBD | Order of Operations | Queue three spells; solve one slightly longer problem and all three fire in order. | Claude |
Open questions
- Which abilities from the pool make the final kit, and in which slots?
- Should failed self-casts cost the cooldown or refund it?
- Puzzle types beyond arithmetic: which ones are readable in under a second?
- How hard do the problems get at high levels, and should each player set their own difficulty?
▸ Forest vs arena, fun, risks, tuning
Forest phase: Fine. Quizzes cost time, but forest fights are looser. Creeps do not need a quiz to be hit by the Chalk Dart.
Lava arena: Funny and viable. Doing math while dodging fireballs is the joke. Final Exam in a crowded FFA arena is spectacular.
Skill ceiling: Fast, accurate answering while still moving, choosing when to gamble on hard problems for crits, keeping the Honor Roll streak alive, and timing Pop Quiz on an enemy who is already busy dodging.
Inspiration: Mike's design; Invoker (Dota 2) for input-gated spells, Brain Age, WarioWare microgames.
Why it's fun
- Every cast is a mini-game; success feels earned (Mike).
- Pop Quiz is pure clan comedy, and the victim always has an out.
- Spectators can play along with the problem on screen.
Risks
- WASD answer selection (Mike's suggestion) stops the Wizard from moving while answering.
- Being quizzed repeatedly gets annoying fast.
- Math ability varies across the clan, which can feel like a skill test unrelated to the game.
Fixes
- Answers on 1-4 or by clicking the answer bubble, so WASD stays on movement.
- Pop Quiz has a per-victim cooldown (one quiz per 10 seconds per target).
- Problem difficulty capped so any answer is readable in about one second; mix in non-math puzzles (shapes, arrows, maze traces).
- Optional per-player difficulty setting in the lobby.
Tuning knobs
- Problem time limits (self vs victim)
- Difficulty tiers and crit multiplier
- Pop Quiz half-damage vs full dodge on success
- Honor Roll cooldown reduction per stack
- Final Exam radius
Rhythm BrawlerMikeConcept
A street fighter who dances through combat; hit the beat and the combos flow on their own.
| Slot | Ability | What it does | From |
|---|---|---|---|
| Basic | Jab (Left Click) | In combat mode, left click attacks toward the cursor (Mike). Hit quality depends on how well the falling arrows are being hit. | Mike |
| Q | Haymaker or Dodge (Right Click) | Right click is either a dodge or a heavy haymaker (Mike left it open). Recommendation: tap for dodge roll, hold for haymaker. | Mike; Agent: Characters (tap/hold split) |
| W | Bust a Move | Out-of-combat dash that drops the Brawler straight into combat mode with a free perfect streak of 4. | Agent: Characters |
| E | Power Move (hold Left Click) | Hold left click while still hitting the arrows to charge a power move; release for a big knockback strike (Mike). | Mike |
| R (ultimate) | Drop the Beat | Tempo doubles for 8 seconds. Every perfect note sends a small shockwave. A miss ends it early. | Agent: Characters |
| Passive | Groove | Entering combat mode starts a DDR-style lane of falling arrows on WASD (Mike). Mouse controls movement while in combat mode (Mike). Hitting arrows builds a Groove multiplier that increases damage and plays automatic combos (Mike: the player does not choose combos). Missed notes slow attacks but never stop them. | Mike; Agent: Characters (Groove multiplier, misses never stop attacks) |
Ability pool: candidates, slot TBD
| Slot | Ability | What it does | From |
|---|---|---|---|
| TBD | Dance-Off | Challenge a nearby player to a 10-second rhythm duel on both their screens; the loser takes a big knockback. Works great on the lava platform. | Claude |
| TBD | Hype Meter | Passive idea: a streak of perfect hits fills a meter that speeds up the music and buffs the next combo. | Claude |
Open questions
- Which abilities from the pool make the final kit, and in which slots?
- Is the music global (everyone hears it) or only for the Brawler?
- Can two Brawlers dancing near each other share a beat for a duel?
▸ Forest vs arena, fun, risks, tuning
Forest phase: Mixed. Rhythm fatigue over a 20-minute forest phase is real. Normal WASD movement outside combat keeps travel comfortable.
Lava arena: Risky. Mouse-driven movement near a lava edge is dangerous, and knockback spells interrupt the flow. Good dodge plus a strong knockback power move make it workable.
Skill ceiling: Holding a perfect streak while positioning with the mouse, charging power moves without dropping notes, and timing Drop the Beat when the opponent cannot escape.
Inspiration: Mike's design; Hi-Fi Rush, Crypt of the NecroDancer, Dance Dance Revolution.
Why it's fun
- Nothing else in the genre plays like this; it has strong novelty.
- Automatic combos look great to spectators even when the player is just hitting arrows (Mike's point).
- A clear expression of skill: the Groove meter shows how well you are doing.
Risks
- Two control schemes in one character are confusing.
- Highest skill floor on the roster.
- Rhythm timing breaks if frame rate is unstable.
- Fighting creeps in rhythm mode every time becomes a chore.
Fixes
- Combat mode only triggers near enemy players or elite creeps; regular creeps use a simple auto-attack mode.
- Movement in combat mode: the Brawler walks toward the cursor only when it is outside a small dead zone, so it can stand still.
- Difficulty charts (Easy, Normal, Expert) selectable by the player, with Expert giving more damage.
- Judge timing on the local client and send results to the server; add a latency calibration step like Crypt of the NecroDancer.
Tuning knobs
- Song BPM and note density per difficulty
- Timing window (perfect / good)
- Groove multiplier cap
- Combat-mode trigger radius
- Dead zone size
Company Squad (U.S. Army)MikeConcept
A Sergeant leading a growing blob of WWII soldiers, tanks, and artillery that moves as one.
| Slot | Ability | What it does | From |
|---|---|---|---|
| Basic | Squad Fire | Units auto-target what the Sergeant attacks. No per-unit micro (Mike). | Mike |
| Q | Grenade | Infantry toss a hand grenade at a target spot (Mike). | Mike |
| W | MG Nest | Machine-gun team sets up facing a direction; enemies in the cone are suppressed (slowed and damage reduced), as in Company of Heroes. | Mike (MG unit); Agent: Characters (suppression) |
| E | Artillery Deploy | Artillery unpacks into siege mode and bombards an area at long range; slow to pack up (Mike, like SC2 siege tank). | Mike |
| R (ultimate) | Airstrike | A WWII fighter plane strafes a line and drops a bomb on the end point (Mike). | Mike |
| Passive | Requisition | Spend gold and XP to buy units into the blob: infantry, MG team, artillery, tank (Mike). Unit slots are capped; a tank costs 3 slots. Units are the Sergeant's HP buffer; lost units can be reinforced for gold. | Mike (buying units); Agent: Characters (slot cap, reinforce) |
Ability pool: candidates, slot TBD
| Slot | Ability | What it does | From |
|---|---|---|---|
| TBD | Medic! | A medic unit that slowly revives fallen squad members if the squad stays out of combat. | Claude |
| TBD | Smoke Screen | Grenade a smoke cloud that hides how many units are in the squad, so enemies can't read its strength. | Claude |
Open questions
- Which abilities from the pool make the final kit, and in which slots?
- Does the Sergeant die when he dies, or when all units die? (Recommendation: when he dies; units protect him.)
- How many units can 15 players field at once? (Mirror handles a few hundred on LAN; cap each squad at 8.)
▸ Forest vs arena, fun, risks, tuning
Forest phase: Strong and fun. The blob snowballs as the game goes on, which is the payoff Mike wants. Weak early, strong late.
Lava arena: Problematic. Each unit getting knocked into lava separately wipes the squad, and a large blob on a small platform crowds everyone. Needs an arena variant.
Skill ceiling: Positioning artillery before a fight starts, MG suppression angles, and timing the airstrike line through a fleeing group.
Inspiration: Mike's design; Company of Heroes (suppression, reinforcement), SC2 siege tank; Lone Druid and Meepo show the 'one player, many units' readability problem in Dota.
Why it's fun
- Watching the army grow is a satisfying progression.
- Calling in an airstrike is a top-tier spectacle.
- Accessible RTS fantasy without RTS micro (Mike's call).
Risks
- A large blob is unreadable in a crowded fight.
- Units bunched on a tree-heavy map may path badly.
- Snowball: a fully built squad could dominate the late forest.
Fixes
- Formation movement (units hold offsets around the Sergeant) instead of each unit pathing separately.
- Arena: squad capped at 3 units plus the Sergeant; knockback applies to the whole formation as one mass; lost units return next forest phase.
- Units become weaker as their number grows (each extra unit adds less), so the blob scales sub-linearly.
Tuning knobs
- Slot cap and per-unit slot cost
- Unit prices and reinforce cost
- Suppression strength
- Artillery deploy/pack time
- Arena unit cap
Gravity LordMikeConcept
A living black hole who bends the battlefield, drags everyone around, and tears holes in space.
Visual effects: Lean hard on Interstellar's black hole (Gargantua): a dark core ringed by a glowing disk of light that wraps over and under it (gravitational lensing). Gravity wells and anti-gravity zones show a glowing grid on the ground that sinks into a funnel toward the center for a pull and bulges up into a hill for a push. That grid picture is the 'rubber sheet' or 'spacetime curvature' diagram from physics explainers. (Mike, 10/14 and 10/15.)
| Slot | Ability | What it does | From |
|---|---|---|---|
| Basic | Mass Driver | A small orbiting rock flung at the cursor; light knockback. | Agent: Characters |
| Q | Gravity Well | A zone that pulls characters toward its center, or toward each other (Mike). Grid visual dips into the ground (Mike). | Mike |
| W | Anti-Gravity Zone | A zone that pushes everyone away from a spot (Mike). Grid bulges up (Mike). | Mike |
| E | Event Horizon | Teleport (Mike). Leaves a brief gravity pull at the departure point. | Mike (teleport); Agent: Characters (pull at origin) |
| R (ultimate) | Shift the Axis | Gravity tilts in one direction for 5 seconds; everyone slides that way (Mike's 'everyone flows to the right'). A 1.5-second arrow warning appears on every screen first. | Mike (gravity shift); Agent: Characters (1.5s warning) |
| Passive | Wormholes | Wormholes passively appear on the map; anyone can use them. Some grant a boon, others invert controls for up to 5 seconds (Mike). The Gravity Lord sees which is which. | Mike (wormholes); Agent: Characters (Lord sees types) |
Ability pool: candidates, slot TBD
| Slot | Ability | What it does | From |
|---|---|---|---|
| TBD | Orbit | An enemy is caught circling the Gravity Lord for a few seconds. | Agent: Characters |
| TBD | Spaghettify | A target is visibly stretched toward a well and takes extra damage while stretched. | Agent: Characters |
| TBD | Slingshot | Use your own gravity well to fling yourself across the map. | Agent: Characters |
| TBD | Time Dilation | Interstellar-themed: inside his field, enemies move and cast at half speed while he moves normally ('one hour here is seven years'). | Claude |
Open questions
- Which abilities from the pool make the final kit, and in which slots?
- Can the global tilt affect projectiles too (curving shots)?
- Do wormholes persist into the arena?
- Should the warped grid show for every player whenever a well is up, or only inside its radius?
▸ Forest vs arena, fun, risks, tuning
Forest phase: Good. Wells group creeps for AOE farming; wormholes create chaos for the whole map; the global tilt is a funny event-sized ultimate.
Lava arena: Dominant, maybe too much. Pulls, pushes, and tilts on a lava platform are all instant-kill tools. This is the most 'Warlock-native' character.
Skill ceiling: Combining a Well and Anti-Gravity to flick someone into lava, using Event Horizon's pull as bait, and timing Shift the Axis to send the whole lobby toward an edge.
Inspiration: Mike's design (including the grid-bending visual); Enigma Black Hole (Dota 2), Sigma (Overwatch), Portal.
Why it's fun
- Black hole visuals and bending grids are a strong identity (Mike).
- Wormholes give everyone a shared chaos toy.
- Shift the Axis is a room-wide moment.
Risks
- Inverted controls feel awful if they come with no warning.
- Pull plus push combos could chain-kill in the arena with no counterplay.
- Kit could feel like 'more push' if every ability is a knockback variant (Mike's 'won't just end up boring' worry).
Fixes
- Inverted controls always show a screen tint and 'INVERTED' banner, capped at 5 seconds.
- The Gravity Lord is affected by his own zones at half strength, so he has to position.
- Add non-push verbs: Orbit (an enemy circles you), Spaghettify (a stretched target takes more damage), Slingshot (use your own well to fling yourself).
Tuning knobs
- Pull and push strength
- Zone radius and duration
- Tilt strength and warning time
- Wormhole spawn rate and boon/curse ratio
- Self-effect strength
XenomorphMikeConcept
A fast, fragile alien hunter that kills in seconds and burns anyone who hits it back up close.
| Slot | Ability | What it does | From |
|---|---|---|---|
| Basic | Claws | Melee only, very high DPS (Mike). | Mike |
| Q | Pounce | Leap onto a target and pin them for 0.75 seconds. Knockback immune while airborne. | Agent: Characters |
| W | Vent Crawl | Enter a vent (any tunnel entrance, wormhole, or dense tree cluster) and exit from another nearby. Could absorb Mike's tunneling idea. | Agent: Characters |
| E | Tail Stab | A short-range lunge with a little extra reach, so the Xeno is not helpless against kiting. | Agent: Characters |
| R (ultimate) | Facehugger | Attach a facehugger to a target. If the target does not kill a creep or heal within 8 seconds, it bursts for heavy damage and spawns a small hostile chestburster. | Agent: Characters |
| Passive | Acid Blood | Melee attackers take damage over time (Mike). Taking damage also leaves small acid puddles on the ground. | Mike (acid blood); Agent: Characters (puddles) |
Ability pool: candidates, slot TBD
| Slot | Ability | What it does | From |
|---|---|---|---|
| TBD | Hive Sense | Briefly sense every low-health player on the map, like a predator smelling blood. | Claude |
| TBD | Wall Crawl | Climb straight over cliffs and dense tree clusters that other characters must path around, so it can ambush from places nobody watches. | Claude |
Open questions
- Which abilities from the pool make the final kit, and in which slots?
- Should a second Xeno player be able to spawn a hive?
- Does the chestburster count as a creep for XP?
- Is acid blood on every melee hit, or only when the Xenomorph is below half health?
▸ Forest vs arena, fun, risks, tuning
Forest phase: Excellent. Ambush from tree cover, quick kills on farmers, fast escapes. Very low HP means a mistake is fatal.
Lava arena: Weak. Melee-only with low HP is easy to kite and knock into lava. Flagged as the opposite of the train: breaks in the arena by being too weak.
Skill ceiling: Choosing engagements, Pounce angles that avoid crowd control, and never being hit back.
Inspiration: Mike's design; Alien franchise (copyright ignored for now), Rengar and Kha'Zix (League), Evolve's monster.
Why it's fun
- Simplest controls on the roster, so a good beginner pick.
- Acid Blood punishes melee brawlers who fight it head-on; a clear lesson.
- Facehugger creates a panic timer the victim can solve.
Risks
- Ambush deaths from an unseen fast melee can feel unfair to farming players.
- Melee-only plus low HP is a trap in the arena.
- Acid puddles add visual clutter.
Fixes
- A hissing audio cue when the Xeno is close.
- Arena: Grip passive, the Xeno takes 40% less knockback (claws dig into the platform).
- Facehugger has clear counterplay (kill a creep or heal), so it is a puzzle, not a death sentence.
Tuning knobs
- HP pool
- Claw DPS
- Acid DoT value
- Pounce range and pin time
- Arena knockback reduction
The BarkeepAgent: CharactersConcept
A tavern keeper who heals anyone who catches a drink, including enemies, and gets the whole lobby drunk on command.
| Slot | Ability | What it does | From |
|---|---|---|---|
| Basic | Bottle Toss | Thrown bottle; light damage. | Agent: Characters |
| Q | Pour One Out | Throw a drink that lands as a pickup. Whoever grabs it first heals, friend or foe. | Agent: Characters |
| W | Bouncer | A shove that knocks a target back hard. Strong in the arena. | Agent: Characters |
| E | Bar Tab | Mark a player. The next time they heal (from anything), the Barkeep earns gold. | Agent: Characters |
| R (ultimate) | Round on the House | Global: every other player gets wobbly aim for 6 seconds, with the cursor drifting around where they point. Home for Mike's orphan 'wobbly aim' ability. | Mike (wobbly aim); Agent: Characters (as Barkeep ult) |
| Passive | Regulars | Players the Barkeep has healed deal 10% less damage to him for 30 seconds. | Agent: Characters |
Ability pool: candidates, slot TBD
| Slot | Ability | What it does | From |
|---|---|---|---|
| TBD | Bar Fight | Everyone near the Barkeep is forced to attack the nearest other player for 3 seconds. Instant tavern brawl; the Barkeep slips out the back. | Claude |
Open questions
- Which abilities from the pool make the final kit, and in which slots?
- Does Round on the House affect allies in team formats?
▸ Forest vs arena, fun, risks, tuning
Forest phase: A support in a free-for-all: healing creates temporary alliances and grudges. Bar Tab makes healing profitable.
Lava arena: Good. Bouncer is a reliable knockback, drinks become contested pickups that pull people toward risky spots, and the global wobble is chaos.
Skill ceiling: Placing drinks so only you or your current ally can reach them, using drinks as bait near lava, and timing the wobble for a key arena moment.
Inspiration: Brewmaster (Dota 2), Ana (Overwatch), the clan's own LAN-party bar.
Why it's fun
- Fills the roster's support gap in a way that fits free-for-all.
- Drinks as pickups make people fight over heals.
- Round on the House is a party gag the whole room feels.
Risks
- Healing an enemy by accident feels bad.
- Global wobble on everyone could annoy spectators of a 1v1.
Fixes
- The drink lands where thrown, so placement is the skill.
- In 1v1 arenas the wobble only affects the opponent.
Tuning knobs
- Heal amount
- Drink lifetime
- Wobble amplitude and duration
- Bouncer knockback
The HarpoonerAgent: CharactersConcept
A whaler with a harpoon gun who hooks enemies across the map and reels them into danger.
| Slot | Ability | What it does | From |
|---|---|---|---|
| Basic | Spear Shot | Long-range skillshot with moderate damage. The roster's ranged poke. | Agent: Characters |
| Q | Harpoon | A hook that pulls the target to you, or you to terrain. | Agent: Characters |
| W | Net | Thrown net that roots for 1 second. | Agent: Characters |
| E | Slip Away | Short dash backwards. | Agent: Characters |
| R (ultimate) | Kraken Call | Tentacles erupt in an area and fling everyone in random directions. | Agent: Characters |
| Passive | Deadeye | Long-range hits deal bonus damage. | Agent: Characters |
Ability pool: candidates, slot TBD
| Slot | Ability | What it does | From |
|---|---|---|---|
| TBD | Grapple | Harpoon terrain instead of a player to reel yourself to it, for mobility and escapes. | Claude |
Open questions
- Which abilities from the pool make the final kit, and in which slots?
- Can the harpoon hook trees to swing around corners?
▸ Forest vs arena, fun, risks, tuning
Forest phase: Solid. Safe ranged farming and picks on careless players.
Lava arena: Strong. A hook is the most classic arena tool: pull them over the lava.
Skill ceiling: Hook prediction and angles that drag targets across lava.
Inspiration: Pudge (Dota 2), Roadhog (Overwatch), Blitzcrank (League), Warlock's classic grapple.
Why it's fun
- Landing hooks is the most understood skill expression in the genre (Pudge, Roadhog).
- Simple to learn, deep to master.
Risks
- Hook chains can feel oppressive if the cooldown is short.
Fixes
- Long hook cooldown and a visible windup.
Tuning knobs
- Hook range, speed, cooldown
- Net root time
- Spear damage falloff
Pinball WizardAgent: CharactersConcept
A carnival showman who turns the battlefield into a pinball table and the enemies into the ball.
| Slot | Ability | What it does | From |
|---|---|---|---|
| Basic | Flipper | Short melee swat that knocks the target in a straight line. | Agent: Characters |
| Q | Bumper | Place a bumper. Anyone who touches it bounces off at high speed. | Agent: Characters |
| W | Tilt | Shake the screen area: small push on everyone nearby. | Agent: Characters |
| E | Launch Plunger | Charge and launch yourself forward. | Agent: Characters |
| R (ultimate) | Multiball | Seal an enemy in a ball and fire them around the map, bouncing off trees, walls, and bumpers for 3 seconds; low damage (Mike's orphan pinball ability). | Mike (pinball idea); Agent: Characters (character) |
| Passive | High Score | Each bounce a victim takes adds points; points become bonus gold. | Agent: Characters |
Ability pool: candidates, slot TBD
| Slot | Ability | What it does | From |
|---|---|---|---|
| TBD | Extra Ball | Once per arena round, the first time you would fall into lava you bounce back onto the platform. | Claude |
Open questions
- Which abilities from the pool make the final kit, and in which slots?
- Should Multiball instead be a rare shop item for anyone?
▸ Forest vs arena, fun, risks, tuning
Forest phase: Fun. Trees make a natural pinball table; Multiball is mostly comedy damage.
Lava arena: Very strong. A ball with no walls on a lava platform is a kill. Needs an arena rule (see fixes).
Skill ceiling: Bumper placement that sets up combos and Multiball angles.
Inspiration: Mike's pinball idea; Peggle, Rounds' bouncing bullets.
Why it's fun
- Mike's pinball gag gets a whole character around it.
- Bumpers are a shared map hazard that creates chaos.
Risks
- Being tossed around is loss of control for the victim.
- Arena Multiball is an instant kill.
Fixes
- Multiball capped at 3 seconds and the victim can steer slightly (nudge).
- In the arena, the ball bounces off an invisible rail at the platform's edge, then the rail breaks.
Tuning knobs
- Ball duration and speed
- Bumper count and lifetime
- Victim steering strength
Dam BeaverAgent: CharactersConcept
A giant engineer beaver who chews down the forest, builds dams, and floods everyone out.
| Slot | Ability | What it does | From |
|---|---|---|---|
| Basic | Tail Slap | Melee with small knockback. | Agent: Characters |
| Q | Chomp | Fell a tree; it falls in a line, damaging and blocking. | Agent: Characters |
| W | Dam | Build a dam wall from felled trees that stores water. | Agent: Characters |
| E | Water Jet | A cone of water that pushes enemies back (Mike's water push). | Mike (water push); Agent: Characters (as Beaver ability) |
| R (ultimate) | Flash Flood | Break the dam: a wave rolls over the area, pushing everyone and slowing them (Mike's flooding). | Mike (flooding); Agent: Characters |
| Passive | Lumber | Each felled tree gives lumber; dams and walls cost lumber. | Agent: Characters |
Ability pool: candidates, slot TBD
| Slot | Ability | What it does | From |
|---|---|---|---|
| TBD | Log Roll | Roll a felled log that bowls players over like pins. | Claude |
Open questions
- Which abilities from the pool make the final kit, and in which slots?
- Does water put out lava for a short time, creating safe paths?
▸ Forest vs arena, fun, risks, tuning
Forest phase: Excellent. It is the only character that uses the forest as a resource, and it changes the map.
Lava arena: Good. Water Jet and Flash Flood are knockback tools. Without trees it relies on stored lumber, a nice cross-phase decision.
Skill ceiling: Timing a flood to wash a group into lava and building dams that funnel enemies.
Inspiration: Mike's water and flooding idea; Tidehunter (Dota 2), Timbersaw (Dota 2).
Why it's fun
- Physics chaos Mike wanted, with water rolling across the map.
- Felling trees changes the map for everyone.
Risks
- Fluid simulation is expensive to network.
Fixes
- Fake the water: a moving push volume with a water shader, not real fluid.
Tuning knobs
- Flood width and push strength
- Lumber per tree
- Dam HP
Abilities looking for a home
| Ability | What it does | Proposed home | Notes |
|---|---|---|---|
| Wobbly Aim (global) | Everyone else's aim gets wobbly; their cursor drifts around where they are aiming. Not debilitating, just harder. | barkeep | Barkeep's ultimate, Round on the House. Could also be a rare global event ('Happy Hour'). In 1v1 arenas, only the opponent is affected so spectators are not punished. |
| Body Swap / Take Control | Take control of an opposing player (or a random one); as the cost, they control your body. Must be visually clear, they cannot spend your money, and there must be no instant-kill exploit. | shop item | Targeted swap is a rare forest-only consumable ('Freaky Friday Scroll'); the random-player version is a global event ('Soul Shuffle'). Safeguards: 10 seconds max; gold, shop, and item selling locked; both bodies take 75% reduced damage from the swapped player; ultimates locked; neither body can be walked into lava (it acts as a wall); swapped players glow and have a tether line plus swapped name tags. Too strong and too situational to sit on one character's kit. |
| Pinball Ball | Seal an opponent in a ball and shoot them around the map, bouncing off walls and trees, not lethal. | pinball-wizard | Pinball Wizard's ultimate, Multiball. In the arena an invisible edge rail stops one bounce, then breaks, so it is not an instant kill. |
| Water Push / Flooding | Shoot water that pushes players back; water rolls over the map as physics chaos. | dam-beaver | Dam Beaver's Water Jet and Flash Flood. A forest 'Flash Flood' global event is a second home. Fake the fluid with a moving push volume. |
Roster analysis Agent: Characters
Gaps
- Support or healer: none of Mike's ten heals. Proposed: Barkeep (free-for-all heals as contested pickups).
- Ranged poke and simple skillshots: only the Math Wizard and Train artillery attack from range. Proposed: Harpooner.
- Beginner-friendly picks: half of Mike's characters use non-standard controls (Train, Motorcycle, Rhythm Brawler, Math Wizard, Squad). New players need at least three plain WASD + mouse picks.
- Arena specialists: Warlock's DNA is knockback, but several characters (Xenomorph, Squad, Beastmaster) are weak on the platform. Gravity Lord, Harpooner, Pinball Wizard, and Dam Beaver cover it.
- Characters that use the forest as a resource: only the Train (plows trees) and Tunneler (ore). Dam Beaver adds lumber.
Overlaps
- Motorcycle and Steam Train: both momentum vehicles. Keep both, separated by weight (light drifting raider vs heavy siege), and share one momentum controller in code.
- Tunneler, Gravity Lord, and Xenomorph all have portal or tunnel travel. Give the Tunneler deliberate infrastructure, the Gravity Lord random wormholes, the Xeno opportunistic vents.
- Beastmaster, Company Squad, and the Merchant's minions are all 'army' characters. Beastmaster is an unruly pack with hunger, the Squad is a disciplined blob with called-in abilities, the Merchant only rents.
- Gravity Lord and Pinball Wizard both push people around. Gravity uses zones; Pinball uses objects and single-target launches.
Beginner-friendly picks
- xenomorph
- harpooner
- barkeep
- merchant
Five of Mike's ten characters change the control scheme, which is great for variety but a heavy onboarding load at a LAN party. Each character should get a one-line 'how to play' on the pick screen. Suggested prototype order: Xenomorph (cheapest, tests melee and acid), Gravity Lord (tests arena knockback), Steam Train (builds the shared momentum controller), Math Wizard (cheap UI, big laughs), then Squad, Beastmaster, Merchant, Tunneler, Motorcycle, and the Rhythm Brawler last. Every ability should have a forest value and an arena value from day one, because half the kits change character between phases.
Engine and tech Proposed
The earlier Unity + Mirror pick still holds: host-authoritative, with a custom plain-C# physics sim. Start on Unity 6.3 LTS (6.0 LTS support ends Oct 2026), and replicate mobs, squads and minions as batched swarm streams, not one networked object per unit, which is the change the bigger scope forces.
| Layer | Pick | Why |
|---|---|---|
| Engine | Unity 6.3 LTS (6000.3), URP | Unity 6.0 LTS support ends October 2026; 6.3 LTS is supported to December 2027. Free under Unity Personal. Mirror added Unity 6 support in v93; verify 6.3 on day 1. |
| Networking | Mirror (latest), KCP transport, NetworkDiscovery for LAN, Hex Spatial Hashing interest management | MIT-licensed with no restrictions, mature, LAN discovery built in. 15 players is far below its proven range (production games at 1,000-5,000 CCU). FishNet (free, no CCU cap) is the Plan B if host CPU becomes the bottleneck. |
| Networking | Custom SwarmStream replication: packed arrays of unit states (~10 bytes per unit) per owner or region at 20 Hz; bitsets for trees, water and lava; RPCs for one-shot events | One NetworkIdentity/NetworkTransform per unit would cost ~1.5 MB/s per client at 600 units (~20 MB/s host upload). Batching plus spatial interest brings it to ~30-120 KB/s per client. |
| Physics | Custom plain-C# 2D solver on a fixed 30 Hz host tick: circles, a few capsules, static tree/cliff colliders, force fields (gravity well, anti-gravity, flood push), Warlock DP knockback | Exact Warlock knockback feel, chaos features defined as data, no PhysX nondeterminism. A few hundred bodies is trivial with a spatial-hash broadphase. Unity 6.3's Box2D v3 low-level API is a possible later backend for collisions. |
| AI | Sleeping creep camps (no tick, no replication until a player is in leash range) + budgeted NavMesh path queries + leader/formation steering for squads and minions; flow fields held in reserve | Keeps active AI to a few dozen units at a time. NavMesh handles hundreds of agents if repaths are staggered; local avoidance is the real cost with clumped groups, so separation forces live in our own solver. |
| Other | Rendering: GPU-instanced units with vertex-animation textures and unit LOD | Hundreds of animated units must run on the weakest clan PC; hundreds of SkinnedMeshRenderers will not. |
| Other | Body swap via a host-side input-slot table (controllerOf[pawn] = connection); economy keyed to account, not pawn | Server-authoritative design makes swap a two-entry table change with no authority transfer. The same table covers mob mind control, disconnect-to-bot and spectating. |
| Other | Arena as a far-off region of the same world scene (teleport = position set); forest sim paused during small-format duels | Avoids scene-load and spawn race issues; Hex Spatial Hashing isolates traffic. Additive scenes with Mirror's Scene interest management are the fallback if arena art gets heavy. |
| Other | UI mini-games (math, rhythm, mazes) generated by host from seed + deadline, played and timed client-side, result validated by host | Network jitter never affects rhythm timing; the simulation only consumes the outcome (fizzle, crit, half damage). |
Top risks
| Risk | Severity | Mitigation |
|---|---|---|
| Host PC overload: it runs the sim, AI and replication for hundreds of units while its owner also plays | high | Swarm batching, interest management, sleeping creep camps, 30 Hz sim with 20 Hz unit sends, struct-of-arrays sim ready for Burst jobs. Spike measures it first. Fallbacks: Burst, then FishNet, then a dedicated headless host on a spare PC. |
| Client rendering of hundreds of animated units plus chaos VFX on the weakest clan PC | high | GPU instancing with vertex-animation textures, unit LOD, VFX budget. Benchmark on the slowest clan machine during the spike. |
| Chaos-feature interactions (body swap during arena teleport, train through a flooded gravity well) produce edge-case bugs and scope creep | medium | One plain-C# sim with unit tests; features as data; a bot-driven 'chaos test' scene that fires random interactions; ship a small core before adding features. |
| LAN-party Wi-Fi: packet loss and limited host upload | medium | Wire the host at minimum; KCP with unreliable deltas; budget host upload under 3 MB/s. |
| Version drift: Mirror on Unity 6.3 is not explicitly listed (README says 6000.1) | low | Verify on day 1 of the spike and pin package versions. |
First technical spike
- Create a Unity 6.3 LTS project with the latest Mirror; confirm NetworkDiscovery finds a host across two physical PCs on the LAN.
- Write the plain-C# sim core: fixed 30 Hz tick, circle bodies, Warlock DP knockback, one gravity-well force field.
- Spawn 600 dummy units: 400 creeps in 40 sleep/wake camps, 10 squads of 15 following a leader, 50 minions; pathing via budgeted NavMesh queries.
- Replicate two ways and record both: (a) naive NetworkIdentity + NetworkTransform per unit, (b) batched SwarmStream arrays with Hex Spatial Hashing interest.
- Load test with 15 headless bot clients (3-5 per spare PC) sending random WASD and casts; measure host sim ms/tick, host frame time while playing, bytes/s per client, total host upload.
- Add a 5,000-tree forest with a destroyed-tree bitset and the arena as a far-off region with a teleport-all trigger.
- Prototype body swap through the input-slot table, including a swap that is active when the arena teleport fires.
- Apply pass/fail gates: host sim < 8 ms/tick at 600 units, < 150 KB/s per client, host upload < 3 MB/s, 60 FPS on the slowest clan PC. If they fail, try Burst, then FishNet, before considering Netcode for Entities.
Idea log
Every idea, in the order it came up: Mike's, Claude's, and the research agents'. Use the ▾ buttons to filter a column, like Excel. The topic sections above are the current truth; this log is the history.
Roadmap
| # | Milestone | What it covers | Status |
|---|---|---|---|
| 0 | Research | Warlock dossier, MOBA/Deadlock, fun mechanics, character workshop, engine re-check: all done 10/15 | Done |
| 1 | Design pass | Lock design principles, the character shortlist, core loop and economy | To do |
| 2 | Engine and netcode spike | Done: Unity 6.6 URP project, headless builds, custom fixed-tick sim, Mirror LAN multiplayer (v0.2). Next: two-PC test, then the 600-unit / 15-bot stress test | In progress |
| 2a | v0.1 walkable forest | Windows .exe: one warlock walking a seeded low-poly forest, tree collision, knockback test key | Done |
| 2b | v0.2 LAN multiplayer | Host or join (LAN discovery or IP), colored wizards with name tags, bumping, Space shove with knockback | Done |
| 2c | v0.3 lava arena | Hex platform eaten by a snake, lava, HP and Warlock knockback, rounds and scores, Wizard (Fireball, Teleport) and Xenomorph (Claw, Pounce), test bots, works over LAN | Done |
| 3 | Arena prototype (proposed first build) | v0.3 done (see 2c). Next: verify numbers against the Warlock dossier, shop between rounds, host settings screen | In progress |
| 4 | Forest prototype | Map, mobs, XP, arena triggers, return to forest | To do |
| 5 | First characters | Two or three concepts built out end to end | To do |
| 6 | LAN playtest | First real session with the clan | To do |
Task board
| Task | Owner | Status | Notes |
|---|---|---|---|
| Create GitHub repo | Claude | Done | Private repo video-game-projects |
| Warlock research dossier | Claude | Done | Google Doc linked below |
| Move portal to /NEDWaur | Claude | Done | Old /video_game_projects and /nedwar redirect here |
| MOBA, Deadlock and mob-design dossier | Claude (sub-agent) | Done | Google Doc — done 10/15, see Research and links |
| Fun and addictive mechanics dossier + proposed principles | Claude (sub-agent) | Done | Google Doc — done 10/15, see Research and links |
| Character workshop | Claude (sub-agent) | Done | Fills the Characters section — done 10/15, see Research and links |
| Engine and networking re-check for the bigger scope | Claude (sub-agent) | Done | Does Unity 6 + Mirror still hold? — done 10/15, see Research and links |
| Review the 4 research docs and the Characters section | Mike | To do | Then pick a character shortlist for the first prototype |
| Review proposed design principles | Mike | To do | Accept, edit, or reject each |
| Confirm arena-first build order | Mike | To do | Claude recommends building the Warlock arena first |
| Run the engine stress test (600 dummy units, 15 bots) | Claude | To do | From the tech doc; pass = host tick under 8 ms, under 150 KB/s per client |
| v0.1 playable build: character walking in a forest | Claude | Done | NEDWaur/Builds/Windows/NEDWaur.exe (rebuild: ProjectSetup.CI) · tree crowns fade when they hide the player |
| Add Mirror networking (host + clients over LAN) | Claude | Done | Mirror 96.11.3; host-authoritative, batched snapshots; LAN discovery + join by IP; tested with 2 instances on one PC (10/20) |
| Feature clip pipeline | Claude | Done | tools/make-clip.ps1 → docs/clips/*.mp4; first clip: v0.1-forest-knockback |
| Turn on NVIDIA Instant Replay (Alt+Z → Instant Replay on; Alt+F10 saves the last 30 s) | Mike | To do | One-time setup for capturing live moments |
| Two-PC LAN test (PC hosts, new laptop joins) | Mike | To do | Thursday 10/22 if the laptop is set up. Checklist: docs/lan-test-checklist.md |
| Stress test: 600 dummy units + 15 bot clients | Claude | To do | From the tech research; pass = host tick under 8 ms, under 150 KB/s per client |
| Check arena numbers (fireball, lava, shrink) against the Warlock dossier | Claude | To do | v0.3 uses marked approximations in SimConfig |
| Read and decide on the 20 proposed design principles | Mike | To do | When there's time; no build work is blocked |
Decision log
| Date | Decision | Why |
|---|---|---|
| 2026-10-11 | All gameplay numbers live in data, not code | Every value can become a host setting |
| 2026-10-11 | LAN-first, no accounts or online servers | Built for one room of friends |
| 2026-10-14 | The game is NEDWaur: forest phase plus Warlock arena | Grew from a straight Warlock rebuild into Mike's own concept |
| 2026-10-14 | Chaos and imbalance are acceptable for now | The audience is the clan; polish comes if it's fun |
| 2026-10-14 | Copyright deferred (e.g. the Xenomorph) | Not publishing yet |
| 2026-10-15 | Moved the portal to /NEDWaur; old URLs redirect | The project has a name now |
| 2026-10-19 | Engine: Unity, starting on 6.6 (installed by Unity Hub); move to the next LTS when it ships | 6.6 is newer than 6.3 LTS and a forward upgrade is routine; downgrading later is the painful direction |
| 2026-10-19 | 3D with a top-down camera | Better end state; no clear 2D art direction; matches Warlock |
| 2026-10-19 | Custom movement and physics code, not Unity physics | Mike wants wonky mechanics (knockback, gravity, momentum); our own fixed-tick sim makes them exact and tweakable |
| 2026-10-19 | URP render pipeline | Runs well on mixed clan PCs; HDRP is overkill |
| 2026-10-19 | Git LFS for binary assets | Keeps the repo small; painful to add after the fact |
| 2026-10-19 | Clips: two lanes. Claude makes scripted feature clips (tools/make-clip.ps1); Mike captures live moments with NVIDIA Instant Replay. No built-in video recorder; build a replay system later instead | Scripted clips are repeatable and free; NVIDIA's encoder beats anything we'd build; a sim replay system powers highlight reels and re-renders |
| 2026-10-20 | Networking: Mirror low-level messages, no NetworkManager or per-unit network objects | Host runs the sim and broadcasts one batched snapshot per tick; scales to many units per the tech research |
| 2026-10-20 | First characters: vanilla Warlock wizard (Fireball, Teleport) and a basic Xenomorph (Claw, Pounce, acid blood, Grip) | Wizard mirrors original Warlock for testing; Xeno is simple and Mike is less attached to it |
| 2026-10-20 | Arena: hex-tile platform eaten by a snake, outer ring first, spiraling inward | Mike's call; deterministic order means the network only sends a tile count |
| 2026-10-20 | Duels: prototype both forest-keeps-running and everyone-watches; lean forest-keeps-running | Mike wants to feel both before choosing |
Open questions
| Question | Status |
|---|---|
| Is the Warlock arena the first thing we build? Yes: v0.3 (10/20) | Decided |
| Should mobs or player kills pay more? (Research: score only arena placements; farming about 65% of income, kills about 35%, kill bounty scales with the victim's wealth) | Open |
| During a 1v1 or 2v2, does the forest keep running? Mike (10/20): try both, leaning yes. Idea: arena keeps running too, with mid-fight joiners and platform regrowth | Open |
| Which characters go in the first prototype? Decided 10/20: vanilla Wizard + Xenomorph | Decided |
| Which arena triggers and formats make the cut? | Open |
| Which Warlock version sets the arena numbers: WC3 1.02, Community Edition 1.10F, or the Dota port defaults? | Open |
| Art direction: forest aesthetic and theme | Open |
| Who on the crew wants to build vs just play? | Open |
| Steam launch? Research (10/21): $7.99 list with launch discount, premium plus earnable cosmetics, no loot boxes, Next Fest mid-2027. Decide after the January LAN | Open |
Research and links
- Warlock Research Dossier — LAN Rebuild
- NEDWaur — Good Games, MOBAs & Deadlock Dossier
- NEDWaur — What Makes Games Fun, Rewarding & Addictive
- NEDWaur — Character Workshop
- NEDWaur — Engine & Networking Feasibility
- NEDWaur — Steam Launch: Pricing, Monetization & Marketing
- Mike's NEDWaur working doc
- GitHub:
michaelpaulfitz-blip/video-game-projects (private)
Google Docs are private to Mike's account.