FFA Runaway Leader: When the v35 Nations Stop Helping the Big Player and Start Ganging Up

Direct answer: the nations gang up the moment you are far enough ahead, and you change it by staying under the trigger

Mechanism	Trigger on the leader	What it means for your next move
Runaway-leader detection	Tile count >= 3x (Medium) / 2x (Hard) / 1.5x (Impossible) the runner-up; none on Easy	Stay under the ratio or the whole board starts pressuring you
Crown nuke gap	30% / 20% / 10% share gap (doubles if the AI does not border you) on Medium / Hard / Impossible	Keep your non-fallout share close to the top so the AI is not a clear crown target
MIRV steamroll	Leader above 20 (Easy) / 10 (Medium, Hard) / 8 (Impossible) cities and >= 1.5 / 1.25 / 1.15x the runner-up city count	City count is the hard floor the city-gap steamroll needs; holding a slim city lead avoids the multi-target MIRV
Victory-denial threshold	Leader at >= 75% (Easy) / 65% (Medium) / 55% (Hard) / 40% (Impossible) of remaining non-fallout land	Above these shares the AI can no longer be beaten; you must have already converted the lead into a win

The short answer: in a Free-for-All on Medium and above, the nation AIs stop treating you as a neutral and start pressuring you the instant you become the runaway leader — which the game defines as a human player who owns at least three times as many tiles as the runner-up on Medium, two times on Hard, and one and a half times on Impossible. Once that ratio is true, the weakest nations start aiming nukes at you the moment you are twenty to thirty percent ahead of them, any nation aims a MIRV at you if your city count is far enough above the runner-up, and the nations refuse to help you grow through the alliance door. Easy is the exception: it never registers a runaway leader in this band, though every difficulty still nukes whoever crosses fifty percent of the map. You change the outcome by either keeping your tile ratio under the trigger, or by splitting the lead so no single player is the clean target. That is the whole decision, and the rest of this guide prices every part of it against the v0.35.2 source.

This is the mechanic the v0.35.0 release notes name directly: the nations now “in FFA gang up on a runaway leader,” and the release’s nations section adds explicit “pressure on a runaway FFA leader.” The community describes the same thing from the player side — the big player suddenly gets swarmed by several AI nations at once, and the debate is whether it is personal hatred or a targeting rule. The answer is the targeting rule: there is a single function, findRunawayLeader, that every nation behaviour calls, and it returns you as the leader only when the tile ratio clears the difficulty-specific multiple. There is no randomness in the trigger itself; the randomness is in what a nation does after it sees you. Understanding the trigger tells you the exact moment the table turns hostile, and that moment is a number you can watch on your own tile count.

The reason this matters for a player who is ahead is that the gang-up is the single largest new cost of being the leader in v35. In earlier builds the leader’s problem was the human field coordinating a focus; now the AI adds a third, reliable attacker that reacts to a pure number on the board. That changes the late-game decision from “can I survive the humans” to “can I survive the humans and the AI that is watching my tile ratio.” The counter-moves are all about the same ratio and the same gaps, so the guide is built around reading them: the tile ratio that makes you the leader, the percentage gap that makes a nation nuke you, the fifty percent line that makes every nation nuke you, the city gap that makes a nation MIRV you, and the alliance door that closes when you are the leader. Every one of those is a value you can read in your own game, and every one of them has a concrete counter.

Version boundary: which build has the gang-up and the five constants that drive it

Every number in this guide is read from the OpenFrontIO source at the v0.35.2 tag, the latest non-TEST release of the v35 line, and the nation behaviour files it touches — NationUtils.ts, NationNukeBehavior.ts, NationMIRVBehavior.ts, AiAttackBehavior.ts, and NationAllianceBehavior.ts — are byte-identical between v0.35.2 and the upstream main branch, so the facts here are bound to a real released build rather than to unreleased main. The mechanic itself landed in v35.0, whose release notes (published 2026-10-16) describe the “smarter nations” change and name the runaway-leader gang-up explicitly; v35.1 and v35.2 are follow-up releases in the same line, and the v35.2 tag is the one this guide cites because it is the newest released tag and its nation code matches main. If you are playing a v34 build or earlier, the gang-up simply does not exist: the nations do not run findRunawayLeader, do not crown-nuke a runaway leader, and do not reject alliances because you are the leader. If you are on an older v35 beta, the constants are the same, because the mechanic and its numbers shipped together in v35.0 and were not changed in the patch releases.

The five constants that drive the whole decision are: the tile-ratio trigger (three, two, and one and a half times the runner-up on Medium, Hard, and Impossible, with Easy returning no leader at all); the crown nuke gap thresholds (thirty percent on Medium, twenty on Hard, ten on Impossible, each doubled when the leader does not share a border with the nation); the near-win hard target (any leader above fifty percent of the non-fallout land is a crown nuke target at every difficulty, including Easy); the MIRV steamroll (leader above ten cities — twenty on Easy — and at one and a half, one and a quarter, or one and a fifth times the runner-up’s city count on Medium, Hard, and Impossible); and the victory-denial threshold (seventy-five, sixty-five, fifty-five, and forty percent on Easy, Medium, Hard, and Impossible, the share a lone player or team must approach before a nation fires a MIRV to deny the win). All five are difficulty-keyed, which is why the same board produces a different answer on Medium versus Impossible, and all five are read from the configuration and the switch statements at the tag rather than estimated.

Two boundary notes matter before you apply anything. First, the runaway-leader trigger and the crown nuke gap both return no result on Easy: findRunawayLeader returns null for Easy, and the crown nuke’s gap branch returns null for Easy, so on Easy the nations never gang up on a leader in the thirty-to-fifty percent band. The only leader-targeting that survives on Easy is the fifty percent near-win nuke, which does not check difficulty at all — meaning an Easy nation still nukes a player who has crossed half the map even though it will not touch a player at forty percent. Second, the runaway-leader function only counts human players when it ranks the leader and the runner-up; bots are skipped, so a bot that owns a huge amount of land does not displace a human from the runner-up slot, and the ratio is always computed between the two largest human holdings. That is why a FFA full of AI nations still has a well-defined runaway leader even when the nations themselves hold a lot of the map.

What makes you the runaway leader: the tile-ratio trigger

The trigger is a ratio, not a share, and that is the first thing to internalise. findRunawayLeader walks the human players, ranks them by tiles owned, and takes the largest as the leader and the second-largest as the runner-up. It then compares the leader’s tile count to the runner-up’s tile count multiplied by the difficulty factor. On Medium the factor is three, on Hard it is two, and on Impossible it is one and a half; the function returns the leader only if the leader’s tiles are at least that product. In other words, on Medium you are the runaway leader the moment you own three times as many tiles as the second-place human; on Hard the bar is two times; on Impossible it is one and a half times. The lower the difficulty, the bigger the gap the game demands before it decides you are far enough ahead to deserve pressure. This is the opposite of the instinct most players have, which is that the hardest AIs are more aggressive and therefore more likely to gang up on you; the actual rule is that the hardest AIs gang up on you at the smallest lead, because one and a half times on Impossible is a much smaller margin than three times on Medium.

Because the trigger is a ratio, it is scale-invariant in an important way: it does not depend on the map size, it depends on the shape of the top of the leaderboard. A player at one hundred tiles with a runner-up at forty is the runaway leader on Medium (one hundred is at least three times forty is false, so no; three times forty is one hundred and twenty, and one hundred is below that, so not a leader), but the same player at one hundred with a runner-up at thirty is the leader on Medium (three times thirty is ninety, and one hundred is above it). The map having five hundred usable tiles or two hundred usable tiles does not change that comparison; only the relative gap does. What map size does change is the second, percentage-based half of the mechanic, which is covered in the nuke section: the crown nuke fires on a share gap, and a share is relative to the total non-fallout land, so on a small map the same number of tiles is a bigger share and the nuke gap closes sooner. The tile ratio and the share gap are two different triggers that both point at the leader, and the ratio is the one that decides whether the runaway-leader flag is set at all.

The practical consequence of the ratio trigger is that the thing you control is the runner-up, not just yourself. Because the trigger divides your tiles by the runner-up’s tiles, you can stop being the runaway leader either by reducing your own growth or by letting the runner-up grow. In a late FFA the second option is usually available, because the runner-up is a human player who is also building, and if you stop over-extending and let the second-place player close the gap, the ratio falls below the factor and the gang-up disengages. That is the single most useful counter in this guide, and it is a direct consequence of the ratio being a division rather than an absolute threshold. A leader who is three times the runner-up is under pressure; the same leader who is two point five times the runner-up is not, on Medium. The band between two point five and three times is dead space where the AI has disengaged but the humans have not yet coordinated, and a skilled leader can live in that band for several minutes. Reading your own ratio against the runner-up, and deciding whether to push it up to win or pull it down to survive, is the core decision this whole mechanic forces you to make.

The three ways the nations strike a runaway leader

Once findRunawayLeader returns you, three independent nation behaviours read that result and act on it, and they act on different clocks and with different weapons. The first is the nuke path, in the nation’s nuclear targeting, which decides whether a nation fires a standard nuke at the crown. The second is the MIRV path, in the nation’s strategic-weapons targeting, which decides whether a nation spends its missile budget on a MIRV at the leader. The third is the ground and alliance path, split between the general AI attack logic that decides whether a nation sends troops at the leader, and the alliance logic that decides whether a nation will form or hold an alliance that would help the leader grow. These three are independent, which is why the player experience of a gang-up looks like several things happening at once: a nuke on one border, a MIRV on a distant city, a ground push on a second border, and a silence where an alliance should have been. No single difficulty setting turns all three on or off together; each one has its own difficulty key, and that is why the same board can produce a nuke without a MIRV, or a MIRV without a ground attack.

The nuke path is the most common first signal, because standard nukes are cheap relative to MIRVs and because the gap threshold is the lowest of the three triggers. The MIRV path is the most expensive and the most decisive, because a MIRV that lands on the leader’s largest city cluster can knock the leader back several tiles in one blow and remove the buildings that were producing the lead. The ground path is the most sustained, because a nation that has decided to attack the leader keeps pressing the border while the leader is the runaway leader, and the release notes describe this as the nations “pressure on a runaway FFA leader.” The alliance path is the quietest but the most strategic, because it does not cost you anything directly; it just removes a tool. When the nations refuse to ally with you because you are the leader, you lose the ability to buy a lull and coordinate with them, which is exactly the tool a leader would want late, because the leader’s best late move is to ally with a second-strong player and win as a pair. The AI closing that door is the reason the leader’s late-game options shrink in v35 compared to v34.

Reading the three paths together gives the player a checklist for the moment they become the leader. First, check the nuke path: is any nation close enough behind you, share-wise, that the gap threshold is crossed? If the weakest non-allied nation is more than the difficulty gap behind you, that nation will nuke you, and the bordering ones are the ones that act first. Second, check the MIRV path: is your city count far enough above the runner-up’s, and above the minimum, that a nation will spend a MIRV on you? Third, check the ground path: is the leader a bordering target for a Hard or Impossible nation, and are there already troops pressing you? Fourth, check the alliance path: have you tried to ally and been refused, or has the field’s AI already stopped offering you deals? Answering those four tells you which of the three weapons is currently aimed at you, and the counter-moves in the failure section are keyed to exactly those four answers. The gang-up is not one threat with many faces; it is four separate triggers, and you can defuse any one of them independently.

The nuke path: crown targets and the thirty / twenty / ten percent gap ladder

The nuke path is the workhorse of the gang-up, and it has two distinct branches that players often conflate. The first branch is the near-win hard target: if the leader owns more than fifty percent of the non-fallout land, the leader is a crown nuke target at every difficulty, and the check does not look at the difficulty at all. That fifty percent constant is CROWN_NEAR_WIN_SHARE, and the share is computed as the leader’s tiles divided by the total land tiles minus the fallout tiles. Fallout matters here in both directions: nuked land is removed from the denominator, so a leader who has nuked a lot of the map can have a higher share than their raw tile count suggests, and a field that has nuked the leader’s land can push the leader’s share down. The practical meaning of the fifty percent line is that it is a no-escape target: once a player is above half the usable map, every non-allied FFA nation has a nuke aimed at them regardless of how far behind they are, and the only defence is to not be the player who crosses that line, or to be allied with the nations that would nuke you.

The second branch is the runaway gap ladder, which applies only when the leader is at or below fifty percent and is the result of findRunawayLeader. Here the nation compares the leader’s share to its own share and fires if the difference exceeds a difficulty threshold: thirty percent on Medium, twenty on Hard, and ten on Impossible. Two modifiers sharpen this. The first is the border rule: if the nation does not share a border with the leader, its threshold is doubled, so a non-bordering Medium nation only nukes the leader when the leader is more than sixty percent ahead of it, while a bordering Medium nation nukes at a thirty percent gap. The second is the relative ordering: the nations comment that neighbours react first and everyone else reacts once the gap is twice as big, which is exactly what the doubled threshold encodes. The result is a ladder in which the weakest nations — the ones farthest behind the leader — are the first to have their gap cross the threshold, so the first nukes in a gang-up come from the small players, not from the runner-up. A leader reading their board can see which nations are close enough to be in the ladder and which are still below it, and the weakest non-allied player on the map is the one most likely to nuke first.

The nuke path has a clean decision rule for a player who is the leader: keep the gap to the weakest non-allied nation under the difficulty threshold, and keep the gap to any bordering nation under the doubled threshold. On Hard, that means the leader wants to be no more than twenty percent ahead of the weakest non-allied player and no more than forty percent ahead of any bordering player, at or below the fifty percent line. On Impossible the numbers are tighter still: ten and twenty percent. The counter-moves follow directly. A leader who is about to cross the ladder either reduces the gap by letting the weakest player build, or moves the weakest player into the leader’s alliance so they stop being a non-allied nuke source, or stays on Easy, where the ladder does not exist at all and only the fifty percent near-win nuke applies. The nuke path is also the one that interacts most directly with the human field: the humans and the AI aim at the same crown, so a leader who is in the nuke ladder is simultaneously in the humans’ focus and the AI’s nuke ladder, and the combined pressure is what makes the v35 endgame feel like a wall that was not there in v34.

The MIRV path: city-count steamroll and the victory-denial threshold

The MIRV path is the strategic-weapons half of the gang-up, and it runs on a different variable than the nuke path: it reads city counts and share, not the tile-ratio trigger. There are two distinct MIRV triggers that point at a strong player. The first is the steamroll trigger, which fires when the leader’s city count is above a minimum and above the runner-up’s city count by a difficulty multiplier. The minimum city count is twenty on Easy, ten on Medium and Hard, and eight on Impossible. The multiplier is two on Easy, one and a half on Medium, one and a quarter on Hard, and one and a fifth on Impossible. So on Medium, a leader with at least ten cities who has one and a half times the runner-up’s city count is a valid MIRV target, and a nation with a missile silo and enough gold will spend a MIRV to knock that city cluster down. On Impossible the same logic trips far more easily, because the minimum is eight cities and the multiplier is one and a fifth, so even a modest city lead makes the leader a MIRV target. The steamroll is called that because it is the AI’s answer to a leader who is clearly rolling: the AI does not wait for the tile ratio, it reads the city count, which is a leading indicator of where the lead is going.

The second MIRV trigger is the victory-denial threshold, which fires when a lone player or a team is approaching the win bar. The threshold is a share of the total land: seventy-five percent on Easy, sixty-five on Medium, fifty-five on Hard, and forty on Impossible. When a lone player’s tiles times one hundred reaches the total land times the threshold, or a team’s combined territory times one hundred reaches it, the nations target the largest member with a MIRV to deny the imminent win. This trigger is scale-based rather than ratio-based, so it does not depend on the runner-up at all; it depends only on how close the player is to winning. On Impossible the forty percent threshold means the AI starts MIRV-denying a leader who has only forty percent of the map, which is very early, and on Easy the seventy-five percent threshold means the AI only reacts right before the game would end. The two triggers therefore cover different parts of the endgame: the steamroll covers the mid-to-late window where the leader is clearly ahead but not yet near the win bar, and the victory-denial covers the final window where the leader is about to cross eighty percent.

The player-facing meaning of the MIRV path is that the city count is a dial you can control, and the AI reads it. A leader who wants to avoid the steamroll keeps their city count within the multiplier of the runner-up’s city count and, on Medium and above, does not let the gap open past one and a half or one and a quarter times. The practical counter is to stop building cities at the same rate the runner-up is, or to redirect growth into tiles rather than cities so the city ratio stays flat while the tile lead grows. A leader who wants to avoid the victory-denial MIRV on Hard or Impossible has to understand that the AI will start denying them at fifty-five or forty percent, which means the leader cannot coast to the bar on those difficulties without expecting a MIRV in the final stretch. The counter there is to force the AI’s MIRVs to be spent earlier, so that by the time the leader is at the bar the field has already used its strategic budget, or to win before the denial threshold by keeping the map small enough that the threshold share is reached quickly. Both triggers are gold-gated, because a nation only fires a MIRV if it has a silo and enough gold, so the MIRV path only bites when the nations are actually wealthy enough to spend it, which is itself a late-game condition.

The ground path and the alliance rejection: pressure and the closed door

The ground path is where the release notes’ phrase “pressure on a runaway FFA leader” lives, and it is handled by the general AI attack logic rather than the nation-specific nuke or MIRV logic. The function that decides whether a nation sends troops at the leader is findCrownTarget, and it has a difficulty structure that is easier to misread than the others. On Hard and Impossible, a nation always considers the runaway leader as a ground target when the leader is a bordering enemy and the nation shares a border with the leader; the actual decision of whether the attack is worth it is then made by the troop-weakness check, which compares the stack the nation can send plus the troops already attacking the leader against the leader’s defence. On Medium, and by the same branch on Easy, a nation only joins an attack that is already under way: the leader becomes a ground target only if the troops already attacking the leader are at least ten percent of the leader’s troops. The consequence is that on Medium the AI does not start the ground pressure on its own; it joins the humans’ pressure once the humans have committed enough troops. On Hard and Impossible the AI starts the pressure itself, because it does not wait for the humans to be at ten percent.

This distinction is the reason the community experience of the gang-up is so different by difficulty. On Hard and Impossible, a leader who becomes the runaway leader feels the AI’s own boots on their border within a turn or two, because the AI is independently deciding to press. On Medium, the leader feels the AI only after a human has already started an attack, so the AI is a follow-on rather than a initiator. The ten percent rule is also a useful shield: a Medium leader who is being attacked by a small human force that is less than ten percent of their troops is safe from the AI joining, because the AI sees the attack as too small to back. The counter-moves for the ground path are therefore difficulty-specific. On Hard and Impossible, the leader cannot prevent the AI from pressing; the best the leader can do is make the pressing unattractive by keeping the leader’s defence thick so the troop-weakness check fails, or by not being the runaway leader in the first place. On Medium, the leader can keep the AI out of the fight by ensuring no single human attack reaches ten percent of the leader’s troops, which means the leader’s defence has to be large enough that any human attack is a small fraction of it.

The alliance path is the quiet half, and it is handled by the alliance behaviour reading the same findRunawayLeader result. The function isRunawayLeader returns true when the other player in a proposed alliance is the runaway leader, and the alliance logic uses that to refuse to help the leader grow, which is the release notes’ “don’t help a runaway leader grow.” The practical effect is that a leader who tries to form an alliance for protection is refused, or the alliance is not extended to include the leader, because the AI would not want to buy a lull for the very player it is pressuring. For a player who is not the leader, the alliance path is an opportunity: the AI will not refuse to help the runner-up or a small player grow, so the small players are the ones who get the AI’s alliance offers late in the game. The strategic reading is that the alliance door is the mechanism by which the AI keeps the field balanced: it refuses the leader and it helps the small players, which is exactly the behaviour of a referee who will not let one player snowball. The leader’s counter is to avoid being the clean runaway leader, so that the AI does not have a clean reason to refuse them; a leader who is close to the runner-up is not the runaway leader and so is not refused, which is why the ratio band between the trigger and a comfortable lead is where the leader’s diplomacy still works.

Scenario one: staying the leader on Hard without tripping the two-times trigger

Concrete board, stated assumptions first. A ten-player Hard FFA on a mid-size map with about four hundred and eighty non-fallout land tiles. You are the leader at one hundred and ten tiles, which is twenty-three percent of the map. The runner-up is a human at fifty tiles, which is ten and a half percent. The weakest non-allied human is at twenty tiles, four percent. You have nine cities; the runner-up has six. Read the triggers. The tile ratio: one hundred and ten divided by fifty is two point two times, and the Hard factor is two, so one hundred and ten is at least two times fifty, which is one hundred, so you are the runaway leader on this board. That single fact is what turns the table hostile, and it is the number to watch.

Now read what that buys the AI against you. The nuke ladder on Hard is a twenty percent gap, doubled to forty percent for non-bordering nations. Your share is twenty-three percent. The weakest non-allied player at four percent has a gap of nineteen percent to you, which is just under the twenty percent threshold, so that player is one tile of growth away from nuking you if they border you, and safe if they do not. The runner-up at ten and a half percent has a twelve point five percent gap, safely under. So the first nuke in your gang-up would come from the weakest player, and only if the gap crosses twenty percent. The MIRV steamroll on Hard needs you above ten cities and at least one and a quarter times the runner-up’s city count; you have nine cities, below the ten-city minimum, so the MIRV steamroll is not armed yet, even though nine cities is one and a half times the runner-up’s six. The ground path on Hard means the AI is independently pressing your border right now, because Hard nations always consider the bordering runaway leader. So on this board you are being ground-pressured and are one tile from the nuke ladder, but not yet MIRV-pressured.

The decision here is to stay the leader while keeping the triggers from arming. The ground pressure you cannot stop on Hard, so you answer it with defence on the pressed border. The nuke ladder you stop by not growing past the twenty percent gap to the weakest player: as long as the weakest player builds a little, your share stays at twenty-three percent and their share rises, closing the gap. The MIRV steamroll you avoid by not building a tenth city while the runner-up still has six, because nine to six is one and a half times and the minimum is the binding constraint, not the multiplier. The concrete counter is a three-part hold: defend the border the AI is pressing, pause city construction at nine, and let the weakest player build so the nuke gap stays under twenty percent. If you instead push to one hundred and fifty tiles, the ratio to the fifty-tile runner-up becomes three times, you are now far above the trigger, the weakest player’s gap widens past twenty percent, the nuke ladder arms across the field, and the game becomes a multi-directional nuke and ground focus against you at once. That is the difference the two-times trigger makes: it is the line between being the leader with manageable pressure and being the leader with the whole table against you.

Scenario two: a Medium lobby where you are ahead and the AI stays out

Second concrete board, different difficulty. An eight-player Medium FFA on a similar map, four hundred and eighty non-fallout tiles. You are the leader at seventy tiles, four percent… let us state it cleanly: you are at seventy tiles, which is fourteen and a half percent of the map. The runner-up is at twenty four tiles, five percent. You have seven cities; the runner-up has four. Read the tile ratio first, because on Medium that is the gate that matters. The Medium factor is three. The runner-up’s tiles times three is twenty four times three, which is seventy-two. You are at seventy, which is below seventy-two, so you are not the runaway leader on this board. The AI has not engaged at all: no crown nuke from the ladder, no MIRV steamroll from the ratio, no independent ground pressure. You are simply the biggest player, and the AI treats you as a normal big neighbour.

Now read what happens if you grow. The moment you reach seventy-two tiles, the ratio crosses three and you become the runaway leader, and the Medium machinery arms. The nuke ladder on Medium is a thirty percent gap, doubled to sixty for non-bordering nations. Your share at seventy-two tiles is fifteen percent. The weakest non-allied player at, say, six percent has a nine percent gap, far under thirty, so even after you cross the trigger the nuke ladder is not yet armed on this small, well-distributed board; the gap has to open to thirty percent before the weakest player nukes, which means the leader can be the runaway leader for a while before the nuke path fires. The ground path on Medium is the immediate change, though: a Medium nation will join an attack on you once the humans have committed at least ten percent of your troops to it, so the moment a human starts a real push, the AI adds to it. The MIRV steamroll on Medium needs you above ten cities and at one and a half times the runner-up; you have seven cities, below the minimum, so not armed.

The decision on Medium is therefore different from Hard. The AI is a slower, more conditional threat, and the human field is the primary pressure. Your best play is to use the dead space between seventy and seventy-two tiles: grow to seventy-one, where you are the biggest player but not the runaway leader, and let the humans fight the runner-up while the AI stays neutral. If you are confident you can win, cross to seventy-two and accept that the AI will now back whatever human attacks you, which means your defence has to be sized for a human-plus-AI front. The counter is the same ratio logic as on Hard but with a wider band: on Medium you have a full one-third of the runner-up’s growth as buffer before the trigger, so you can be more aggressive with your growth and still stay below the line. The two scenarios together show the central lesson of the mechanic: the difficulty does not change what the AI targets, it changes how small a lead is enough to become the target, and the counter is always to manage the ratio to the runner-up rather than the absolute size of your territory.

Failure modes and counters: when the gang-up catches you

The first and most common failure is crossing the fifty percent near-win line and becoming the no-escape crown target. A player who grows carelessly into the second half of the map triggers CROWN_NEAR_WIN_SHARE at every difficulty, including Easy, and the moment they do, every non-allied FFA nation has a standard nuke aimed at them regardless of the gap. The counter is not to cross that line as the clean leader. If you are the player who is about to cross fifty percent, the move is to keep a second human challenger alive at a real share, so that the leader identity is contested and the AI’s crown target is ambiguous, or to have allied with the nations that would nuke you so their nukes are not aimed at you. A player who is not the leader but is being pressured because the AI is ganging up on their ally should remember that the same fifty percent line protects them: the AI will nuke the leader, not them, as long as they stay below the line and are not the runaway leader themselves.

The second failure is the city-count steamroll catching a leader who has been building cities faster than the field. A leader on Medium or Hard who has eleven cities while the runner-up has six is at one and a half times the runner-up’s count and above the ten-city minimum, so the MIRV steamroll is armed and a wealthy nation will spend a MIRV to knock the city cluster. The counter is to hold the city ratio flat: grow tiles while the runner-up is growing cities, so the tile lead does the work and the city ratio stays inside the multiplier. A leader who is already in the steamroll band can buy time by not building the next city, because the minimum city count and the multiplier are both thresholds, and the leader can sit just under one of them. The third failure is the victory-denial MIRV on Hard and Impossible, where the AI starts denying the leader at fifty-five and forty percent respectively. The counter there is to force the field’s MIRV budget to be spent earlier in the game, so that when the leader is at the bar the nations have no strategic budget left, which means creating pressure on other borders early so the AI spends its MIRVs on rival targets before the leader reaches the denial threshold.

The fourth failure is the alliance door closing and removing the leader’s best late tool. A leader who becomes the runaway leader and then tries to ally for protection is refused, because isRunawayLeader tells the AI not to help the leader grow. The counter is to not be the clean runaway leader at the moment you want to ally: sit in the band below the ratio trigger and make the alliance, because the AI does not refuse a player who is not the runaway leader. The fifth failure is the combined human-plus-AI front, which is the real endgame threat: the humans focus the crown and the AI’s nuke ladder, MIRV, and ground pressure all point at the same target at the same time. The counter to that is the only one that actually ends the game, which is to win before the focus converges, or to split the board so the focus has two targets. Every counter in this section reduces to the same two levers, the tile ratio to the runner-up and the share to the weakest player, and the player who can read both numbers on their own board can choose which failure they are in and pick the matching counter before the gang-up fully arms.

Mode and map adjustments: when the gang-up is weaker or stronger

The mode adjustment is the cleanest: the runaway-leader gang-up is Free-for-All only. findRunawayLeader returns null for any game mode that is not FFA, so in Team mode the tile-ratio trigger, the crown nuke gap ladder, and the ground findCrownTarget that reads the runaway leader are all off. A team that is far ahead does not become the runaway leader in the sense the FFA mechanic uses, and the nations do not gang up on the leading team through this path. That does not mean Team has no leader-targeting: the MIRV steamroll and the victory-denial threshold both read team territory, so in Team a nation still fires a MIRV to deny a team that is approaching the win bar, and the steamroll still fires on a city-count lead. The difference is that the FFA-specific gang-up, the combination of the ratio trigger with the nuke ladder and the ground pressure, is a FFA behaviour. If you are playing Team and feeling the AI press your team, the cause is the MIRV denial path, not the runaway-leader gang-up, and the counter is the victory-denial counter, not the ratio counter.

The map adjustment is about scale and about fallout. The tile-ratio trigger and the share thresholds are both ratios and shares, so they are scale-invariant in the pure sense: the same gap on a large map and a small map produces the same trigger. What changes with map size is the absolute number of tiles you need to reach a given share, and the amount of fallout in play. On a small map, a single large nuke removes a big share of the usable land, so fallout shifts shares quickly, and a leader who has nuked a lot of the map can have their share pushed up or down by the fallout more than on a large map. On a large map, the leader needs more absolute tiles to be the runaway leader, so the trigger is slower to arm in raw tile terms, but the percentage ladder arms at the same relative gap. The practical reading for map choice is that the gang-up feels the same on every map if you read the ratios, but on small maps the absolute tile numbers move faster, so the window where the leader can sit in the dead band between the trigger and the nuke ladder is shorter in real time. A leader on a small map has fewer turns to manage the ratio, and a leader on a large map has more, which is why the same player feels the gang-up bite harder on a compact FFA map than on a sprawling one.

The difficulty adjustment is the one that most players misread, so it is worth stating flat. The hardest difficulties trigger the gang-up at the smallest lead, because the tile-ratio factor is smallest on Impossible and the share thresholds are smallest on Impossible. That is the opposite of the common belief that harder AIs are simply more aggressive and therefore more likely to gang up; the actual rule is that the hardest AIs gang up on the smallest margin, because one and a half times on Impossible is a much smaller lead than three times on Medium. The easiest difficulty, Easy, does not gang up on a runaway leader at all in the band below fifty percent, because the factor and the gap both return null, so on Easy the only leader-targeting is the fifty percent near-win nuke. The net effect across the four difficulties is a gradient: on Easy the AI is a loose threat that only nukes a leader past half the map; on Medium it is a conditional threat that needs a three-times lead and joins human attacks; on Hard it is an aggressive threat that presses at two times and fires nukes at a twenty percent gap; and on Impossible it is an early threat that presses at one and a half times, fires at a ten percent gap, and starts MIRV-denying at forty percent. Choosing your difficulty is therefore choosing how small a lead is enough to become the target, and the player who wants to be the leader and still win should understand which difficulty’s bar they are trying to stay under.

Decision framework: read the board, set your gap, and pick your counter

The whole mechanic reduces to a four-step read that you can do on your own board in a few seconds, and then one decision. Step one: are you the runaway leader? Read your tile count and the runner-up’s tile count and compare to the difficulty factor. If your tiles are at least three times the runner-up on Medium, two times on Hard, or one and a half times on Impossible, the flag is set and the table is hostile; if not, the AI has not engaged and your problem is the human field, not the AI. This is the cheapest and most important number, and it is the one the community argues about most, because the player on the leader’s side thinks they are just big while the player on the field’s side can see the ratio has already crossed. Step two: if you are the runaway leader, read the nuke gap. Compare your share to the weakest non-allied player’s share and to each bordering player’s share, and check the difficulty threshold and the doubled bordering threshold. This tells you which nations have a nuke aimed at you and which are still in the ladder but not yet armed.

Step three: read the MIRV triggers. Compare your city count to the runner-up’s city count and to the minimum, and check the steamroll multiplier; then compare your share to the victory-denial threshold to see if the AI is close to denying your win with a MIRV. Step four: read the ground and alliance paths. On Hard and Impossible assume the AI is independently pressing your border; on Medium assume the AI joins a human attack once it reaches ten percent of your troops; and assume the alliance door is closed to you while you are the runaway leader. The decision then has three shapes. If you are not the runaway leader, the AI is not your problem and you play the human endgame. If you are the runaway leader and you want to survive, your counter is to pull the ratio down by letting the runner-up build, keep the nuke gaps under the thresholds, and hold the city ratio flat. If you are the runaway leader and you want to win, your counter is to win before the focus converges, keep a second challenger alive so the crown target is ambiguous, and spend the field’s MIRV budget early so the denial path has nothing left at the bar. The framework is the same on every map and every difficulty; the difficulty only changes the numbers in the four reads, and the map only changes the speed at which the tile numbers move. A player who can do the four reads can always say which failure they are in and pick the matching counter, which is the entire value of knowing the mechanic instead of guessing at the AI’s hatred.

Where this guide fits with the rest of the FFA content

This guide covers the AI side of the leader problem, and the player-side pieces live in the guides it links, which is why the links are load-bearing rather than decorative. The player-side version of the same endgame is the focus guide, which prices when the human field should ally and commit to take down the snowballing leader, and it is the mirror of this guide: this guide explains why the AI joins that focus, and the focus guide explains how the humans time it. The pressure guide covers the general question of how the nations target and how to avoid being the best safe prize, which is the ordinary targeting behaviour that this guide’s runaway-leader behaviour layers on top of; the runaway-leader path is a special case that the pressure guide does not own. The ranked FFA guide covers the same gang-up in the ranked queue, where the ELO-matched lobby means the field is usually strong enough that the AI’s pressure is the deciding factor late, and it is the natural next read once you understand the mechanic on the casual maps.

The nuke calculator and the MIRV guide are the tools side of the decision, because the counters in this guide are all about timing and spending nukes and MIRVs, and the calculator prices the nuke sequence while the MIRV guide prices the strategic budget. The nation difficulty guide is the reference for what each difficulty’s nations do in general, and this guide is the FFA-specific slice of it, so the two should be read together: the difficulty guide tells you the temperament, and this guide tells you the numbers. The alliance break timing guide covers the player-side cost of the alliance the leader would want, and the closed-door section of this guide explains why the AI will not be the ally the leader needs, so the two explain the alliance problem from the player and the AI sides respectively. The beachhead doctrine guide covers how the v35 nations open and press a border, which is the ground half of the pressure this guide describes, and reading it alongside this guide shows how the ground path and the nuke path combine into the full gang-up. Together, the linked guides and this one cover the complete FFA leader problem: the human focus, the AI focus, the tools to price both, and the difficulty and mode boundaries that change the answer, and the single number that ties them all together is the tile ratio to the runner-up that this guide spends its length explaining.

Read them together rather than in isolation: this guide explains why the AI joins a focus, and the FFA focus crown endgame guide prices when the human field should commit to it; the nations pressure guide covers the ordinary targeting behaviour this runaway-leader path layers on top of; the ranked FFA guide carries the same gang-up into the ELO-matched lobby; the nuke calculator prices the nuke sequence the leader has to dodge; the MIRV guide prices the strategic budget behind the steamroll and denial triggers; and the nation difficulty guide is the reference for what each difficulty’s nations do in general. The alliance break timing guide and the nation beachhead doctrine guide close out the player-side alliance cost and the ground half of the pressure, respectively.