FFA Endgame Commit: When to Spend the Last Reserve with Two or Three Players Left

Direct answer

In a two- or three-player Free-For-All final the win is the first player to cross 80 percent of usable tiles, and the only decision that separates a player who holds too long from a player who commits too early is whether the next spend buys a named tile objective and still leaves a reserve that can absorb the counterstrike. Commit the last reserve or the last ready MIRV when you can name the specific border or core the spend converts and the reserve that remains after the spend is at least as strong as the rival’s current army; hold and reposition behind the two big players’ mutual-nuke standoff when you cannot name that objective but you can fund a ring of high-level SAM launchers that intercepts the last 150 tiles of a warhead arc before it reaches your core; hold liquid and let the threshold settle only in a lobby where OVERTIME is enabled, because with the timer off the bar never sinks on its own and holding is a losing default. The 80 percent threshold is the constant; the timer, the SAM ring, and the traitor window are the three variables that change which of the three branches wins the final. A player who reads the three branches as a preference ladder, choosing the one that feels most comfortable, loses the final the same way as the player who holds too long, because the branches are conditions, not preferences, and the condition that names your branch is the number you can see on the board, not the reserve you would like to have. The commit branch is named by the point gap and the named tile objective, the reposition branch is named by the MIRV ladder and the last-150-tiles SAM window, and the hold branch is named by the OVERTIME lobby setting and your position as the leading side. Reading the three numbers before choosing the branch is the whole skill, and the rest of this guide is the three numbers worked out so that a player in the final two or three players can name the branch and the number in one move.

The most common losing move in a FFA final is not the wrong branch, it is running the right branch inside the wrong window. The player who commits the last reserve while their own defense is halved for the next thirty seconds, because they just broke the last alliance, hands the third player the exact moment to absorb the commitment and collapse the backline. The player who holds the reserve until a rival commits first, expecting to react, usually finds the third player already across the 80 percent line because the bar only moves one way. This guide treats the final two or three players as a decision problem with three branches and a set of numbers that say which branch is available right now, rather than a feeling about how strong you are, because in a FFA the players you cannot see are the ones who win by default.

Why the final two or three players are a different game

The midgame of a Free-For-All is a target-selection problem: you are choosing who to expand against, who to ally with, and who to deter, and the midgame hold, expand, and ally decision is where most of the game is decided. The endgame is a different animal, and the difference is that the player count has dropped to two or three, which means the deterrence posture that kept you safe while six or eight players were still in the game has either become your entire game or stopped working at all. With three players left, the two biggest rivals are usually mutually threatening with nuclear fire, and the third player is the one who can still be moved: they are the only one whose economy is not already locked into a standoff, so they are the only one who can still choose which front to open. With two players left, the deterrence question disappears entirely and the game becomes a pure 80 percent race: the first player to cross 80 percent of usable tiles wins, and the only question left is who spends their last reserve first and whether that spend survives the counterstrike.

The reason the final two or three players force a commit decision rather than a continuation of the midgame is that the 80 percent threshold is a single fixed line and the map is a fixed pool of usable tiles. In a two-player final the two players’ territory shares must sum to 100 percent of usable tiles, so if one player is at 62 percent the other is at 38 percent and the winner is the one who converts the remaining 18 points, which means the game ends the moment one of them commits enough to push past 80 and the other cannot stop it. In a three-player final the three shares must sum to 100 percent, so the third player’s share is the variable that the two big players are fighting over, and the third player is the one who decides whether the two big players’ mutual nukes cancel each other out or whether one of them gets a second front. The deterrence posture guide is explicit that it does not own the endgame, and the reason is exactly this: the four-input deterrence question (banked Gold, ready Silo tubes, visible troop strength, territory exposure) that a midgame player answers to stop being the easy hit is the wrong question once the player count has dropped, because with two or three players the cost of attacking you is no longer the variable, the 80 percent line is.

The practical consequence is that a FFA player who keeps playing the midgame script into the final two or three players is playing a different game than the game that is actually on the board. They keep asking “how expensive do I make it to attack me” when the real question is “do I have a named tile objective that the last reserve buys, and does the reserve that remains survive the counterstrike.” This is why the community’s recurring regret in a FFA final is hoarding the reserve until a rival commits first: the reserve was held as a deterrence posture, but a deterrence posture in the final two or three players is a reserve that is waiting for a rival to make the first move, and the rival who makes the first move is the one who crosses 80 percent first. The guide that follows builds a three-branch test from the source numbers so that a player in the final two or three players can name the branch they are in and the number that decides it, instead of feeling it.

Version boundary: the one number that never changes

The version boundary for this guide is the v0.34.21 tag, commit d7075bface, which is the latest non-TEST release as of 2026-10-01 and is published in src/data/_meta.json as the upstream cursor. Between v0.34.19 and v0.34.21 the released changes are infrastructure, lobby, and login fixes; none of them touch the traitor rule, the MIRV price, the SAM interception window, or the win threshold, so the numbers in this guide are stable across that window. That matters because the FFA endgame is one of the few parts of the game where the community’s mental model is out of date: several players still assume the win threshold is two-thirds of territory or that the OVERTIME timer is always on, and both assumptions are wrong in v34. The one number that never changes across any mode, map, or lobby is PERCENT_TILES_OWNED_TO_WIN, which is 80 in Config.ts, and it is the number the entire guide is built around.

The 80 percent threshold is the share of usable tiles, not the share of all tiles on the map. Usable tiles exclude water and the tiles that are not part of the contested pool, so on a larger FFA map the 80 percent line is a larger absolute territory than on a compact map, but it is always 80 percent of the usable pool. The consequence is that a player’s path to the 80 percent line depends on the map, but the player’s position relative to the line does not: a player at 62 percent on a large map and a player at 62 percent on a compact map are both 18 points from the line, and the 18 points is the number that decides the commit. The winning overtime guide covers the OVERTIME threshold timing and the OVERTIME-only read, and it is the natural next step once a player has decided which branch of the commit test they are in, because the OVERTIME timer changes the hold branch from a losing default to the default.

The second stable number is the OVERTIME default set, which is { enabled: false, startMinutes: 30, dropPercentPerMinute: 2 }. The timer is off by default, which means that in a public lobby that has not enabled it, the 80 percent bar never sinks on its own and the game ends only when a player crosses the line. In a lobby where the timer is enabled, after the 30 minute mark the required percentage falls 2 points per minute with no floor, so a stalled game always ends and the leading side only has to wait for the bar to sink below their lead. The community’s recurring OVERTIME confusion, surfaced in the source threads, is exactly that a player does not know whether the timer is in their lobby, so the hold branch of this guide is conditional on that lobby setting. The third stable number is the traitor break debuff, which is defense x0.5 and enemy attack speed x0.8 for 30 seconds on the player who clicks break, and the alliance break timing guide carries the full rule. The fourth stable number is the MIRV price ladder, which is 25M at the first launch and rises 15M per game-wide launch, so five consecutive launches from zero cost 25M, 40M, 55M, 70M, and 85M, and the MIRV price ladder guide carries the same numbers against the v0.34.11 reference. The rest of the guide uses these four numbers to build the three-branch commit test.

The three-branch commit test

The decision in the final two or three players is a three-branch test, and the branch you are in is named by the number you can see, not by how strong you feel. The table below is the test: each branch is named by the concrete condition that makes it the right branch, the source number that decides it, and the failure mode that happens when the branch is run inside the wrong window. A player in the final two or three players should be able to read their own position against the table and name the branch, because the three branches are mutually exclusive in the source numbers and the only error a player can make is running one branch inside the window where another branch owns the decision.

The three branches are not a preference ladder; they are a condition ladder. The commit branch is the only branch that works when you can name a tile objective and the spend survives the counterstrike, and it is the only branch that is urgent, because the 80 percent line does not wait. The reposition branch is the only branch that works when you cannot name a tile objective but you can still fund a SAM ring, and it is the only branch that converts the two big players’ mutual nukes from a threat into a defensive advantage. The hold branch is the only branch that works when the OVERTIME timer is on and you are the leading side, and it is the only branch that is a default rather than a choice, because the timer does the work of sinking the bar. The next three sections work out each branch in full, with the numbers and the assumption that each branch requires, and the scenarios that follow work out the two concrete finals that the community keeps asking about.

Branch one — commit the last reserve or the last ready MIRV

The commit branch is the only branch that ends the game before the threshold would, and it is the branch the community runs both too early and too late. The condition that names it is simple: you can name the specific border or core that the last reserve or the last ready MIRV converts into your territory, and the reserve that remains after the spend is at least as strong as the rival’s current army. If either of those two facts is false, the commit is not a commit, it is a liquidation, because a spend that does not buy a named objective and does not leave a survivable reserve empties the home front and hands the win to the third player. The source number that decides the commit is the point gap between your territory share and the 80 percent line, because the commit only works if the spend closes that gap, and a spend that closes half the gap leaves the game to the counterstrike.

The two forms of the commit are the last-reserve commit and the last-MIRV commit, and they are different in the source numbers. The last-reserve commit spends the current Troops to break a border and convert the contested tiles on the other side into your territory; it is limited by the troop growth rate, which means the reserve you rebuild after the spend grows fastest right after the drain, so the window to commit the reserve is the moment your reserve is at its peak, not the moment you are low. The last-MIRV commit spends the last ready Silo tube to convert a rival’s core into unclaimed territory by destroying the City or the capital that anchors their share; it is limited by the MIRV price ladder, which means the fifth launch from zero costs 85M and the MIRV price ladder guide carries the full 25M / 40M / 55M / 70M / 85M numbers. The choice between the two is the choice between converting border tiles with Troops and converting core tiles with a warhead, and the assumption each one requires is named in the scenario sections. The one rule that both forms share is that the spend must leave a reserve that can absorb the next counterstrike, because a FFA final is won by the player who survives the counterstrike, not by the player who commits first.

The window that kills the commit branch is the traitor window. If the commit requires breaking the last alliance first, the player who clicks break is the traitor for 30 seconds, during which their defense is halved and enemy attacks land 20 percent faster, and the alliance break timing guide carries the full debuff. The 30 seconds is long enough for the third player to absorb the commitment and collapse the backline, because the third player is the only one whose army is not locked into the standoff and the only one who can move to your exposed backline while your defense is halved. The commit branch is therefore conditional on the traitor window: a commit that does not require breaking an alliance is available the whole time, and a commit that does require breaking an alliance is only available outside the 30 second window, which means the player who breaks the alliance to commit must time the spend so that the counterstrike force is already in position before the window closes. The failure mode of the commit branch is not the wrong spend, it is the spend made inside the traitor window, and the next section works out the reposition branch as the alternative when the commit is not available in the window.

Branch two — reposition behind the mutual-nuke standoff

The reposition branch is the only branch that works when you cannot name a tile objective but you can still fund a ring of high-level SAM launchers, and it is the branch the community runs when the two big players are mutually threatening with nukes and a smaller third player is the one who can still be moved. The condition that names it is that you can intercept the last 150 tiles of a warhead arc before it reaches your core, and the source numbers that decide it are the 25M / 40M / 55M / 70M / 85M MIRV ladder and the last-150-tiles SAM window that the nuke defense protocol guide carries. The reposition is not a hold; it is a deliberate move that repositions your economy so that the two big players’ mutual nukes cancel each other out while you consolidate the core and force the third front. The assumption the reposition requires is that your liquid Gold can fund a ring of at least two high-level launchers on the landing side before the first warhead enters the targetable window, because a ring that is not ready when the first MIRV flies is a ring that converts the standoff into a mutual kill.

The reason the reposition works is that a nuke is targetable only inside the last 150 tiles of its flight, which means that a ring of high-level SAM launchers on the landing side is the only thing that can intercept the warhead before it lands, and the level of the launchers matters because a level-N SAM Launcher has N slots and a real firing range of 150 minus 480 over level plus 5, so a high-level ring both has more interception slots and reaches further out into the arc. The two big players’ mutual nukes are a standoff because neither of them can clear the other’s ring and land a warhead on the other’s core before the other’s ring intercepts the return fire, and the third player is the one who is not in the standoff, because the third player’s economy is not locked into the ring and the third player is the only one who can move to a front that the two big players cannot see. The reposition is the third player’s move: hold the reserve, keep your own nukes quiet, and move your economy so that the two big players’ mutual nukes are the threat to each other, not to you, and the doomsday versus overtime guide carries the read on when the breaker is the one who is actually committing.

The window that kills the reposition branch is the time before the ring is ready. If the first warhead enters the targetable window before the ring is built, the standoff collapses into a mutual kill and the third player wins by default, because the two big players’ cores are both exposed and the third player is the only one whose core is not under fire. The reposition is therefore conditional on the build time of the ring, and the build time is the number that decides whether the reposition is available: if your liquid Gold can fund the ring before the first warhead enters the targetable window, the reposition is available, and if it cannot, the reposition is not a reposition, it is a hold, and the hold branch is the only branch left. The failure mode of the reposition branch is not the wrong move, it is the move made too late, and the scenario that follows works out the concrete numbers of a reposition that is ready in time and the reposition that is not.

Branch three — hold liquid and let the threshold settle

The hold branch is the only branch that is a default rather than a choice, and it is the only branch that works when the OVERTIME timer is enabled and you are the leading side. The condition that names it is that the lobby has OVERTIME enabled and you are ahead of the rival on the territory share, so that the bar sinks 2 points per minute and you only have to wait for it to sink below your lead. The source number that decides the hold is the OVERTIME default set, enabled: false, startMinutes: 30, dropPercentPerMinute: 2, and the winning overtime guide carries the full threshold-timing read. The hold is not a passive move; it is an active decision to not commit, to keep the reserve liquid, and to let the timer do the work of sinking the bar, because the 80 percent line does not wait and the timer is the only thing that changes the 80 percent line into a number that your lead can cross.

The reason the hold is conditional on the timer is that with the timer off the bar never sinks, and a hold with the timer off is a losing default because the rival who commits first crosses 80 percent first. The community’s recurring OVERTIME confusion is exactly that a player holds the reserve expecting the timer to sink the bar, and then finds the timer is off in their lobby and the bar stays at 80 percent, so the rival who commits the last reserve crosses the line and the player who was holding loses by default. The hold is therefore only a branch in a lobby where the timer is on, and the assumption it requires is that you are the leading side at the 30 minute mark, because a trailing side that holds with the timer on is the trailing side that the timer is sinking the bar toward, and the timer does not help the trailing side. The hold is the branch the community runs when they are ahead and the timer is on, and it is the branch that converts the 80 percent line into a number that the lead can cross without a commit.

The window that kills the hold branch is the moment the timer is off and the rival commits. If the lobby does not have OVERTIME enabled, the hold is not a branch, it is a loss, because the bar stays at 80 percent and the rival who commits the last reserve crosses the line. If the lobby has OVERTIME enabled but the player is the trailing side, the hold is not a branch, it is a loss, because the timer is sinking the bar toward the trailing side’s share. The hold is therefore conditional on both the lobby setting and the player’s position, and the assumption it requires is named in the scenario sections. The failure mode of the hold branch is not the wrong decision, it is the decision made in a lobby where the timer is off, and the next section works out the two concrete finals that the community keeps asking about, one where the commit is the right branch and one where the reposition is the right branch.

Scenario one — the named 18-point gap

The first concrete final is a three-player Free-For-All where player A controls 62 percent of usable tiles, player B controls 25 percent, and player C controls 8 percent. A is 18 points from the 80 percent threshold, and the only way A crosses it is by breaking C’s border and converting the unclaimed or contested tiles near C’s core into A’s. The commit question is whether A’s last reserve can break C’s border and still hold a counterstrike force against B, because if B collapses A’s second front the same moment A commits, A loses the home front and B wins by default. The numbers the scenario works are C’s 8 percent and the 18-point gap, and the assumption is that A’s last reserve is sized to break C’s border and leave a reserve at least as strong as B’s current army. The traitor window is the variable: if A’s commit requires breaking C’s alliance first, A is the traitor for 30 seconds with defense halved, and the 30 seconds is long enough for B to absorb the commitment and collapse A’s backline, so A must time the spend so that the counterstrike force is already in position before the window closes.

The commit is available in this scenario because A can name the specific border the spend converts, the 18-point gap, and the reserve that remains after the spend is at least as strong as B’s current army. The last-reserve form of the commit spends the current Troops to break C’s border and convert the 8 percent of C’s tiles into A’s, which pushes A from 62 percent to 70 percent, and the remaining 10 percent of the gap is closed by the counterstrike force that holds the contested tiles near C’s core. The last-MIRV form of the commit spends the last ready Silo tube to destroy C’s City, which converts C’s core into unclaimed territory and pushes A’s share past 80 percent, but the fifth launch from zero costs 85M and the MIRV price ladder guide carries the full ladder. The choice between the two forms is the choice between a troop commit that leaves a reserve and a warhead commit that leaves the core exposed, and the assumption each one requires is that the spend survives the counterstrike. The scenario is the community’s most common FFA final, and the reason it is a commit and not a reposition is that A can name the 18-point gap and the reserve that closes it, so the commit branch owns the decision.

Scenario two — the reposition and the SAM ring

The second concrete final is the same three-player structure, but the third player is the smaller one and the two big players are mutually threatening with nukes. Player A controls 55 percent, player B controls 30 percent, and player C controls 15 percent, and A and B are both at the last ready Silo tube, so neither can land a warhead on the other’s core before the other’s ring intercepts the return fire. The reposition question is whether C’s liquid Gold can fund a ring of high-level SAM launchers before A or B commits, because if C’s ring is not ready when the first MIRV flies, the standoff collapses into a mutual kill and C wins by default. The numbers the scenario works are the 25M / 40M / 55M MIRV ladder and the last-150-tiles SAM window, and the assumption is that C’s liquid Gold can fund a ring of at least two high-level launchers on the landing side before the first warhead enters the targetable window. The reposition is available in this scenario because C cannot name a tile objective that the last reserve buys, but C can fund the ring, so the reposition branch owns the decision and the nuke defense protocol guide carries the full ring build.

The reposition is C’s move: hold the reserve, keep C’s own nukes quiet, and move C’s economy so that A and B’s mutual nukes are the threat to each other, not to C, while C builds the ring and consolidates the core. The window that decides the reposition is the time before the ring is ready, and the build time of the ring is the number that decides whether C’s ring is available: if C’s liquid Gold can fund the ring before A or B commits the last warhead, the reposition is available and C wins by default when A and B’s mutual nukes cancel each other out. If C’s ring is not ready when the first warhead enters the targetable window, the standoff collapses into a mutual kill and C wins by default, because A and B’s cores are both exposed and C is the only one whose core is not under fire. The scenario is the community’s second most common FFA final, and the reason it is a reposition and not a commit is that C cannot name a tile objective that the last reserve buys, but C can fund the ring, so the reposition branch owns the decision and the hold branch is the only branch left if C cannot fund the ring.

Failure modes, mode and map adjustments, and next reading

The failure modes of the three branches are the wrong window, not the wrong branch. The commit branch fails when it is run inside the traitor window, so the spend is absorbed and the backline is collapsed. The reposition branch fails when it is run too late, so the first warhead enters the targetable window before the ring is ready and the standoff collapses into a mutual kill. The hold branch fails when it is run in a lobby where the timer is off, so the bar stays at 80 percent and the rival who commits first crosses the line. The one failure mode that all three branches share is the player who keeps playing the midgame script into the final two or three players, who keeps asking how expensive they make it to attack them when the real question is whether they have a named tile objective and a survivable reserve, and the ffa midgame strategy guide is the wrong guide for the endgame and this guide is the right one.

The mode and map adjustments are the same ones the existing endgame guides carry. On a larger FFA map the last-150-tiles SAM window is the same but the time before the alert fires is longer, so the reposition branch has more time to build the ring and the commit branch is less urgent because the 18-point gap is a larger absolute territory. On a compact FFA map the same standoff collapses faster, so the commit branch is more urgent and the reposition branch has less time to build the ring. In a lobby where OVERTIME is enabled, the hold branch becomes the default for the leading side and the commit and reposition branches are reserved for a named objective inside the timer’s remaining window. In a lobby where OVERTIME is off, the commit and reposition branches are the only tools and the hold branch is a losing default because the bar never sinks on its own. The one rule that does not change across any mode, map, or lobby is that in a two- or three-player FFA final the win condition is the 80 percent threshold, not the player count, and the player who commits first with a named objective and a survivable reserve is the player who crosses it first.

The next reading is the winning overtime guide once a player has decided which branch of the commit test they are in, because the OVERTIME timer changes the hold branch from a losing default to the default. The source boundary for this guide is the v0.34.21 tag and the four source numbers it carries: the 80 percent threshold, the OVERTIME default set, the traitor break debuff, and the MIRV price ladder. The guide does not own the midgame hold, expand, and ally decision, which the ffa midgame strategy guide owns, and it does not own the OVERTIME threshold timing, which the winning overtime guide owns, and it does not own the nuke defense ring build, which the nuke defense protocol guide owns. It owns the one decision that the final two or three players force: the player-count commit decision, and the number that decides it, which is the 80 percent threshold.