TVW Temprals - Veil Walkers

Player and minmaxer guide

How the world AI thinks.

The AI does not solve the entire match like a chess engine. It reads the current hero turn, builds a tactical board, creates possible actions, avoids blocked or bad fights where its board can see them, promotes urgent goals, then commits the strongest justified option. In the game UI, AI status is explicit: you can see when it is thinking, when it acted, when no legal step was found, and when a local combat rule resolved a stalled encounter.

Same movement limits One route read per hero Objective biased Last rival standing Bad fights mean prepare
The AI decision funnel from map reading to chosen action. AI turn funnel Read board AP, claims, goals Create options move, fight, recruit Apply rules deny or promote Rank urgency first Act justified option

Mental model

Think of the AI as a tactical funnel, not an oracle.

Each hero gets a short, practical reading of the board. The AI is good at local priorities: resolving claims, taking reachable value, following the current objective, and avoiding obviously bad fights. It is not a perfect long-term economist that sacrifices ten turns for a future trap.

It sees the active turn. Movement left, action budget, local value, open claims, reachable routes, and current strategic phase.
It creates practical options. Move, recruit, recover, build, pick up value, visit a site, answer a claim, or start a prepared fight.
It ranks urgency before greed. A claim or decisive objective can beat an easy nearby reward. Local conversion beats wandering away.

Victory grammar

First it asks: what kind of win is this map asking for?

The AI is not chasing one universal number. Each map names the kind of victory that matters, then the AI bends its route choices toward that finish. Normal skirmish should read as a heroic map duel: grow, take power, force a decisive encounter. Scenario maps may ask for a hold, a relic sweep, or a score race instead.

VS Last rival standing

Beat every rival hero. This is the default skirmish fantasy: build a stronger company, choose the fight, and end the map in a decisive clash.

R Relic sweep

Capture every relic objective. The AI treats relic routes as more urgent when the board clearly presents a relic race.

H Core hold

Stand on the core and keep control long enough. If another player takes the core, the hold streak breaks and the siege story restarts.

VP Score race

Earn enough victory points from authored sources like objectives and economy. This is for maps that really want a cumulative race.

1Read map goal
2Find useful anchors
3Pick the next legal step

Friendly names like "heroic win" or "last one standing" describe the player read. The actual map rules underneath are the four families above: eliminate rivals, capture relics, hold the core, or reach a target score.

The board it reads

The tactical board is the AI's memory for one decision.

Before scoring anything, the AI compresses the map into a board for the current hero. This is why the same hero may chase an objective one turn, recruit the next, then answer a claim: the board changed.

The tactical board with actor, reach, pressure, objectives, and local value. Tactical board one hero, one decision not the whole future Actor AP, powers, army Reach routes and costs Pressure claims, races, holds Mission phase town, company, objective

Route planner

On large maps, it now reads the road once and reuses the answer.

The important trick is simple: for one hero decision, the AI builds a small route memory using the same movement law as the player preview. Then every relic, core, town, pickup, claim, or guarded objective asks that memory: "Can I get there, what is the first step, how many days, and what blocks me?"

A
Build reach cloud

Terrain, AP, day horizon, claims, and blocked tiles become one route map.

B
Ask every target

Relics, cores, towns, pickups, and enemy pressure all query the same evidence.

C
Keep the first step

The AI chooses a legal next tile. It does not teleport to the final objective.

D
Expose blockers

If a claim or terrain law blocks the path, that same blocker informs scoring.

A hero route cloud reused by multiple targets. Hero route memory Relic Core Guard one search claim blocker reused
Why this matters: the AI stays smarter and faster on huge maps because every target uses the same route evidence. That means fewer repeated long-road searches, fewer inconsistent choices, and cleaner reasons when a path is blocked.

Scoring

Urgency beats raw score.

The final choice is not just "highest number wins". First the AI removes illegal options. Then it compares priority bands. Only after that do score, type of action, distance, and cost break ties.

1. LegalCan this hero actually do it now?
2. UrgentDoes it resolve a claim, fight, or objective?
3. ValuableHow much progress or power does it create?
4. EfficientHow close, cheap, and stable is the option?
Candidate ranking from illegal options to the chosen action. Option A illegal Option B nearby reward Option C objective route Option D claim response Final ordering legal first, then urgency band, then score and efficiency 1

Combat and preparation

The AI wants objectives, but it does not love bad fights.

A protected objective becomes a two-step question: "Is the objective important?" and "Can I win the fight?" If the answer to the second question is no, the AI looks for power first.

How the AI decides between attacking and preparing. Guarded goal high value target Fight looks winnable? Attack route claim, bridge, combat Prepare first town, recruit, recover, site Gain power then retry

Minmax reading

How to predict and exploit it.

What you see Likely AI move Player counterplay
It stands on a useful town. Recruit, recover, build, or transfer before leaving. Create a stronger urgent threat if you want it to leave early.
It is weak near a protected goal. Delay the attack and seek preparation. Deny nearby towns, creature sites, and easy power pickups.
You contest a strategic value. Respond to the claim even if another reward is close. Use claims to pull tempo when its other options are slow.
No urgent pressure exists. Take nearby value or position toward the best visible target. This is the best window for baiting or racing elsewhere.
Version francaise Back to manual