First, pin down the rules
Before discussing tactics, here are the confirmed facts that form the foundation. Every conclusion below is grounded in these rules alone.
- Runs 21:00–02:00, five hours total
- Two win conditions. (1) Hold for 2.5 continuous hours for an instant win. (2) If (1) doesn't happen within five hours, the top cumulative hold time wins. The continuous counter resets to zero the instant the hold breaks, even for one second. Given reasonable matchmaking, (1) rarely lands because attacking is favored — so it's effectively decided by (2)
- The score is total occupation time. Only the length of time held is compared between alliances
- A rally is one march, per person. One person can field only one march at a time. Landing time is a choice of 5 or 10 minutes
- March time scales with distance to the castle (closer is faster)
- You cannot rally a tile your own alliance is holding
- A rally in flight is only visible to whoever currently holds the tile. The side about to attack can't see it yet
- Reinforcement is outside the march limit — it doesn't consume your one march. But the troop count garrisoned at the castle has a capacity ceiling
The single equation that decides occupation time
In practice the result is almost always decided by top cumulative hold time. Since the score is total occupation time, your share is captured completely in one line.
Total time (5 hours) is fixed. So there are only two ways to grow your share: (A) extend the time you already hold (defend successfully), or (B) cut the time the opponent holds (retake it faster). Every tactic reduces to one of these two.
And every individual fight is decided by the same comparison.
Leader power is the fixed strength baked in by development and formation. Troop count is the number of troops committed — a variable that rises and falls with each fight. Here's the key: launching a rally cuts the opponent's troop count, and reinforcing restores your own. This correspondence is the shared vocabulary for everything that follows.
A structural fact: you always concede the first 5 minutes
Consider the moment the castle is taken from you. Even if you rally to retake it immediately, landing takes 5 minutes at the earliest. So whoever loses the castle concedes at least 5 minutes, no matter what. This has nothing to do with skill or power — it's a floor created by the rally landing delay.
This "5 minutes you always concede" is the starting point for every tactic. Everything from here is about shrinking the time after that opening 5 minutes.
Attack and defense are asymmetric — perfect defense doesn't exist
You have to grasp that defense has a structural ceiling. Attack and defense are asymmetric from the very way marches are fielded.
Attack = as many rallies as you have people
A rally merges many people into one march. Even at one march per person, an alliance of 100 can field that many rallies in parallel (a 10-person rally means 10 of them). The ceiling on attack volume is effectively "how many people you can mobilize" — which is very high.
Defense = capacity is the ceiling
The defender, on the other hand, is capped by one garrison holder's troop capacity. Reinforcement only restores up to that ceiling. So if same-tier rallies arrive in numbers exceeding capacity — say ten 1M-power rallies hitting a 1M garrison — no amount of queue management via insertion can physically hold (assuming no large power gap).
Defense has an absolute ceiling; attack does not. So perfect defense is impossible by the rules. Maximizing occupation time can't be done by defending alone — it's necessarily a combination of "defend to stall" and "retake fastest when lost."
The tool to extend a hold: reinforcement insertion
(A) The concrete way to extend a hold is "insertion" — a probabilistic technique that stalls within capacity by exploiting the order the queue is processed in.
Landings are processed one at a time, in queue order
Even when the on-screen clock shows the same second, landings on a garrison are processed internally one at a time, in order (likely managed down to the millisecond). Slipping a reinforcement into that order is the insertion.
When an enemy lands two rallies back to back, the default flow is "first one shreds troop count → second one takes the castle." But if you slip one reinforcement between them, you get "repel #1 → reinforcement restores troop count → repel #2," and you hold.
The reinforcement's queue position decides success
The catch: the clock only shows seconds, so deliberately slotting into the "gap" is nearly impossible to time. In practice it becomes a probability game — many people pile reinforcements onto the same second, hoping one lands in the gap.
Syncing rally landing times used to take real hands-on skill; a rules change has since made that easy. Timing a reinforcement to that same instant, however, still takes skill — another way the defender is at a disadvantage.
The origin of everything: grab it first
Consider the most idealized case, and the value of the first grab comes into sharp focus. This is a thought experiment purely to surface that value — as the previous sections established, in reality attacking is favored and no perfect equilibrium occurs.
In the ideal equilibrium, the first grab alone decides it
Suppose both sides are exactly equal in power and, with zero delay, run the same tactic (the pre-launched relay, below). Occupation then trades cleanly every 5 minutes, settling into a steady state where 5 hours = 60 slots are split alternately.
In this state occupation time is perfectly even (30 slots each), and the only difference is who took the very first slot. Whoever grabbed first wins by exactly that much.
No equilibrium exists in reality. But the first grab is always positive
Of course, perfect equilibrium never happens. Reaction delay, execution precision, capacity gaps, power gaps always intrude. Even so, the value of grabbing first can never be zero or negative by the rules. The reason splits into two independent layers.
Grab first, and all the time until the opponent retakes it is pure score for you. Their retake carries at least a 5-minute landing delay, so at least 5 minutes is unconditionally yours. Whoever the opponent is, whatever the power gap — you always get it.
Grab first and you become the "defender," the opponent the "attacker." The defender can stall with reinforcement (outside the march limit, endlessly repeatable); the attacker, once repelled, stalls while waiting for the march to return and heal. In other words, you force the disadvantaged role onto them. This edge only holds while their volume stays within capacity — beyond that it vanishes. But Layer 1 never does.
After it's taken: retaking within your own alliance
(B) Cutting the opponent's hold means shrinking the gap between losing the castle and retaking it. Without any cross-alliance coordination, there are only two ways to do it entirely within your own alliance.
Method 1: the pre-launched 10-minute rally (relay)
One person fields only one march. So the pre-launch can't be run solo — it's a relay where several members split the marches. Here's the mechanic.
A member other than the one retaking keeps a longer (10-minute) rally running ahead of time. Once the castle flips to you, that rally becomes invisible to the enemy — only the holder can see it. By the time the enemy retakes, this arrow already has less time left, so it lands sooner — you've pre-paid the reaction delay. It can be canceled, and with enough people you can stage several arrows at different remaining times.
Method 2: solo capture
Skip the rally delay and take it directly with a single march. This works when your total power (leader power + troop count) beats theirs solo — for instance, catching the moment right after their retake, before reinforcement has topped up, when troop count is thinnest. Since there's no rally to assemble, it's the fastest possible move.
Playstyle by power tier
Given all the above, the moves actually available to you step down according to where your alliance's relative power sits. Which tier you're in is roughly readable from the event ranking.
Tier 2 in depth: three signals to watch in the near-even tier
In the near-even tier, defending involves the luck of insertion and attacking won't break through on one rally. So continuously judging "what to target right now" is itself the difference in occupation time. Three axes drive that call.
- ① The occupation leader's current state: if the leader is garrisoned, it's a rally target — but before firing, always check whether other alliances are also rallying
- ② Other alliances' rallies and landing timing: e.g. if another alliance's rally lands and strips the leader, and you read that the leader will re-rally, overlay your own rally on that re-rally to cut the leader's hold. Snipe the moment they come back
- ③ Occupation-time gap to the alliances above and below: whether it pays to hit the leader or to pull away from #2 changes with the gap. Always pick the best move while watching the differential
Update these three minute by minute and move your rally target accordingly. In the near-even tier, the result is decided less by individual combat strength than by this board reading and speed of judgment.
Reality
Over a five-hour haul, the same conditions rarely hold the whole way through.
Troop count — the variable in the power equation — is garrison headcount on defense and rally headcount on attack, and it drops as time passes. Only so many people can stay up past midnight and keep grinding. So rallies that fielded three drop to two, and garrisons can't heal back the troops the turrets chip away fast enough. For five straight hours a commander has to track a situation that shifts minute by minute, fighting off sleep while operating — it's a genuinely grueling event.
And most participants are Tier 3, described above (at least on my state and generation). In that tier, plenty of people just farm the points they need for rewards and head home. Still, it's undeniably a lively event where, once every two weeks, the whole server piles into the center for a chaotic scrum — so however you enjoy it, it's worth showing up.