OpenFront Multi-Front Defense: Which Front to Hold, Which to Concede

Direct answer: defend the front you can actually win, concede the one you can only slow

When two or more players push you at the same time, the instinct to split your army and meet every front is the losing move, because the defender’s per-tile loss makes two-front attrition compound faster than any split stack can replace. The single decision that separates a recoverable multi-front attack from a collapse is an allocation choice made before the contact: pick the one front where your stack is at or above the defender’s troops-per-tile line, hold that front with a defense post or terrain, and actively concede the other fronts by pulling the border inward a controlled amount so the attacker pays a full magnitude of troops per tile for every conquest. Conceding is not defeat; in the v0.34.24 model a conceded plains tile still costs the attacker roughly eighty magnitude-scaled troops and drains you of your average troops-per-tile, so a deliberate retreat on a weak front is a positive trade when that same troops-per-tile would be spent dying on a front you cannot hold. The counter-play that breaks a genuine ganged pair rather than absorbing it is to identify the weaker member of the pair and overload that member’s front while the stronger member ties down your main stack, which turns the pair’s coordination against it and forces them to split in response. This whole section compresses into one test: if you can name the front you will hold, the front you will concede, and the specific unit that will do the conceding, you have a plan; if you cannot, you are already the front.

The defender’s troops-per-tile line is the single number that decides whether a front is holdable, and you compute it before the fight, not during. It is your current total mobile troops divided by your current land tiles: at one hundred twenty thousand troops on one hundred fifty thousand land tiles your line is eighty troops per tile, and any attacker whose committed stack sits below that figure is already losing the local duel before terrain or posts enter the calculation. That is why the hold front is not the front with the most troops in absolute terms but the front where your committed stack meets or beats that eighty and a post or highland can push the effective number higher still. If you cannot name the number you are not deciding; you are guessing which front feels bigger, which is exactly the split-to-meet-every-front failure this guide exists to stop. The operational step that turns the principle into a plan is to say the three names to yourself before the second attacker is within march range: the hold front, the concede front, and the specific named unit that will do the conceding, because the instant a unit is doing the conceding the front stops being a place you might lose and becomes a lever you pull deliberately.

Detection: is this a real ganged pair or just opportunistic jumping

The first thing to resolve before allocating is whether you are facing a coordinated team or two players who both happen to be attacking you at the same moment, because the counter-play differs. In the community threads this is the most argued point, and the working distinction is that real teaming has a pre-battle signature while opportunistic jumping does not. A real ganged pair typically spawns adjacent or within a short march, places cities that face each other’s borders rather than the outside, and keeps a spare stack on the coast or flank that is ready to rotate the instant one of the pair is engaged. The tell is not that they attack together once but that their production is pointed at the shared border instead of at their own outside expansion. Opportunistic jumping has the opposite signature: each attacker’s cities face outward, each is still expanding into empty or bot territory, and the second attacker only shows up because you are already bleeding and look cheap. The practical consequence is that you treat a detected pair as a single threat with two fronts and run the allocation rule below, while you treat opportunistic jumping as two independent single-front problems that you can defeat one at a time if you are willing to concede the second temporarily. A third signature worth checking is multi-accounting, where two avatars act in perfect sync from one human; the sync is so tight that the two fronts move on the same tick and their stacks are the same size, which you can test by watching whether both fronts pause and resume together. If the two fronts act as one brain, defend them as one front and break the cheaper of the two; if they act as two, you may be able to beat them sequentially.

The binding constraint on detection is the time window, because the counter-play only works if you detect the pair before the second attacker’s cities face your border, and by the time the second stack is visible on your border the detection question has already resolved in the wrong direction. So you run the check continuously rather than once: the moment a neighbour places a city that faces your border instead of their outside, or holds a stack at the coast or flank instead of expanding, you re-run the pre-battle signature. The ambiguous case is the most dangerous, where one attacker shows a real-team signature and the other still looks opportunistic, and the correct handling is not to average the two into a middle guess but to work from the stronger signal and treat the weaker as a check, meaning you allocate as if facing the real pair and watch the weaker attacker for the moment they commit to the shared border. If the weaker attacker stays outward-facing while you fight the detected pair you have been over-allocating to a two-front threat that was really one and a half and you can fold the spare stack back into the hold front; if the weaker attacker turns to face your border the signature confirms and the break slot becomes mandatory. The rule that keeps this from becoming analysis paralysis is that a detected second signal must be acted on within one production cycle, because the cost of treating an opportunist as a pair is a slightly over-cautious hold while the cost of treating a pair as an opportunist is the whole collapse.

The per-tile drain: why two fronts compound faster than two stacks

The mechanical heart of the multi-front problem is that the defender’s loss is charged per conquered tile, per attacking front, not per attacker. In the v0.34.24 attack logic, every tile an attacker conquers removes the defender’s average troops-per-tile from the defender, computed as total defender troops divided by defender land tiles, and this charge is independent of how big the attacker’s stack is. That means a single attacker conquering two of your tiles drains you the same total as two attackers each conquering one of your tiles, even if the second attacker is a thin stack of a few thousand troops. The compound happens on two axes at once. On the loss axis, two fronts double the rate at which your army is removed, so your troop cap regenerates against a drain that is twice the single-front rate and you fall below the line faster. On the speed axis, each attacker’s conquest is slowed by its own ratio against the specific front it is pressing, so two fronts let the attackers pick the two weakest points of your border and ignore the strong point, instead of one attacker being forced into your best-defended sector. The large-territory bonus deepens this late game: when both your neighbours are large, each gets a reduced per-tile cost to push from and into, so the two-front squeeze arrives earlier and bites harder in the mid-to-late game than it does in the opening. The defence-side mirror is that a Defense Post in range of a tile multiplies that tile’s attack magnitude by 5 and its speed term by 3, so a defended front is the one front where you can make the attacker’s per-tile cost high enough that conceding becomes more expensive than holding. Every allocation decision in the next sections follows from this arithmetic: you defend the front where your local force plus a post plus terrain makes the per-tile cost high, and you concede the front where you cannot.

A worked micro-example makes the compounding concrete. Suppose your line is eighty troops per tile and two attackers press at once. Attacker A takes one plains tile and you are drained your average of eighty on it; at the same instant attacker B takes one tile and you are drained eighty again, so two tiles and one production cycle and your line has dropped from eighty to roughly seventy-six, and that lower line is the number A is now measured against on its next tile, so a stack that was borderline at eighty is now clearly over the line. The single-front version takes one tile per cycle and your line falls to about seventy-nine, which cap regeneration largely restores; the two-front version falls to seventy-six and the regeneration cannot keep up because it is being charged at double the tile rate, so the gap between your line and the attackers’ committed stacks widens every cycle instead of holding steady. That widening gap is the actual kill mechanism of a ganged attack, and it is precisely why the concession and the break exist: you concede a front to stop one of the two charges and you break the pair to remove an entire attacker’s contribution to the drain.

The allocation framework: hold one front, bleed two, break the pair

The framework has three named slots, and you must fill all three before the contact. Slot one is the held front: the single front where your stack is at or above the defender’s troops-per-tile line and where you can place or already have a defense post within thirty tiles, or where the terrain is highland or mountain so the attacker’s per-tile cost is raised. You commit the bulk of your mobile army to this front and you let it absorb the strongest of the attackers. Slot two is the conceded front or fronts: every other border, where you deliberately pull the border inward a controlled amount, keep the local stack thin enough that the attacker must commit a real stack to advance, and let each conquered tile pay the full eighty-to-twelvey magnitude-scaled cost while draining you only your average troops-per-tile. The goal of the conceded front is not to stop the attacker but to make the advance slow and expensive enough that by the time the attacker reaches your core you have already won the held front. Slot three is the break: if the threat is a real ganged pair, the break is the act of overloading the weaker member’s front with a dedicated stack while your main stack ties down the stronger member, which forces the pair to either split their coordination (destroying it) or accept that the weaker member is being deleted (removing one front). The order of operations is: detect, then hold, then bleed, then break, and you do not advance to breaking until the hold is stable and the bleed is paying. If you cannot fill the hold slot because no front is at or above your troops-per-tile line, the honest answer is that you are overextended for this lobby and the correct play is to concede the whole border in a planned retreat that protects the core, not to split and lose on all fronts.

When the threat is three attackers rather than two the framework does not change, only the arithmetic gets stricter. The hold front is still the single front at or above your line with a post or highland and you still commit the bulk of your mobile army to it, but now two conceded fronts are draining you at twice the single-concede rate, so the planned retreat must run deeper and the re-entry point must be chosen deliberately rather than optimistically. The break becomes both more valuable and more dangerous: removing one of the three deletes a front, but breaking the wrong member can delete your dedicated force without removing any front, so the break target must be the smallest stack committed into you specifically, measured against your line, not the weakest of the three in absolute terms. In practice a three-front squeeze usually means you are already overextended, and the honest answer is often the full retreat that protects the core and fights from a smaller defensible area, because three simultaneous charges on a mid-territory player is the one case where even a well-run hold-concede-break can lose the line faster than cap regeneration restores it.

Scenario one: land and sea at once, the coastal pincer

Concrete setup. You are at roughly three hundred thousand tiles in a public FFA, two human neighbours and one bot. Your northern neighbour is pushing you by land through plains at a stack of about ninety thousand troops, and your western coast is receiving a landing from your southern neighbour’s transport at about forty thousand troops. You have one destroyer that is currently on the eastern coast, your army is at about one hundred twenty thousand troops against a cap of about one hundred fifty thousand, and you have one defense post already built near the northern plains route. The allocation: the held front is the north, because you have the defense post within thirty tiles of the route, the post multiplies the defended tile’s magnitude by 5 and its speed by 3, and the plains base magnitude of eighty becomes four hundred on defended tiles, which is enough that your one-hundred-twenty-thousand stack at or above the defender’s troops-per-tile line holds. The conceded front is the west coast, but you do not concede it for free: you move the destroyer to the west before the landing commits, because the single most common failure in the community threads is the naval unit being on the wrong coast when the landing lands, and a destroyer that arrives before the transport unloads converts the west from a second army into a naval duel you can win locally. The break is not available here because this is opportunistic jumping, not a coordinated pair, so you hold the north, win the west locally with the destroyer, and then have a free army to turn on the weaker of the two attackers. The numbers that make this work: the west landing at forty thousand against your destroyer and a thin shore stack loses a large share of its stack in the naval exchange, the north attacker pays four-hundred-scaled magnitude per defended tile and is slowed to roughly a third of its plains speed, and your average troops-per-tile drain on the conceded western tiles is small because you keep the west thin. If the destroyer is not on the west in time, the west becomes a real second army and you must concede the west in a planned retreat that pulls the coast inward, accepting the lost tiles at their full cost to save the core.

If the destroyer reaches the west a full production cycle early the west collapses into a naval duel and the north hold becomes the entire fight, which is exactly the single-front problem you were trying to avoid, so the destroyer timing is the difference between a two-front fight and a winnable one.

Scenario two: two large neighbours, the late-game squeeze

Concrete setup. You are at about four hundred thousand tiles in the late game, with two large human neighbours both at three hundred thousand or more. Both are pushing you at once, the east through highland and the south through plains, because the large-territory bonus has made both of their attacks cheaper and faster to push from and into. Your army is at about two hundred twenty thousand troops against a cap of about two hundred forty thousand, you have no defense posts built yet, and you are one push from the endgame. The allocation: the held front is the east through the highland, because the highland base magnitude of one hundred and its higher tile cost mean the attacker pays more per tile on the highland than on the plains, and you can build one defense post inside thirty tiles of the highland route to raise the defended magnitude to five hundred and slow the front by three times, which is the cheapest front to defend in the game and the best front to hold against a large attacker. The conceded front is the south plains, where you pull the border inward and let the plains attacker pay its full eighty-scaled cost per tile while draining you your average troops-per-tile, which you can afford because the highland hold is stable. The break here is the real pair break: the two neighbours are large enough that they are effectively a team even if they are not allied, so you identify the one whose stack into your south is smaller and overload that stack with a dedicated force of about sixty thousand troops while your main two-hundred-twenty-thousand stack ties down the east, which deletes the south attacker’s committed force and removes one front. The numbers that make this work: the highland hold with the post costs the east attacker five-hundred-scaled magnitude per defended tile and slows it to a third, the south conceded front pays eighty-scaled cost and drains you only the average, and the sixty-thousand break stack against a smaller south stack wins the local duel because it is at or above the defender’s troops-per-tile line for that front. If both neighbours are genuinely equal and you cannot find a smaller stack to break, the honest play is the full retreat that protects the core and fights from a smaller territory, because two equal large neighbours pushing at once in the late game is the case where the per-tile drain on both fronts outpaces your regeneration no matter how you split.

Defense posts as the front-asymmetry tool

A Defense Post is not a generic wall; in the multi-front problem it is the tool that lets you make one front dramatically more expensive than the others, which is exactly what the allocation needs. The v0.34.24 constants are a thirty-tile range, a 5x multiplier on the defended tile’s attack magnitude, a 3x multiplier on the defended tile’s speed term, a build cost that runs from fifty thousand Gold for the first post up to a two-hundred-fifty-thousand cap, and a fifty normal-tick construction window. The decision is therefore never about whether a post is good in the abstract but about which front you are making expensive and whether the post can finish its fifty-tick construction before the attacker reaches it. The key asymmetry is that the bonus applies per defended tile within range, so a single post placed at the bend of a route can cover several of the tiles the attacker must take and multiply each of their magnitudes by 5, which is what turns a plains front with an eighty base into a four-hundred front on the defended tiles. The allocation consequence is that you spend the post’s fifty-to-two-hundred-fifty-thousand Gold on the front you have already chosen to hold, not spread across every front, because a post on a front you are conceding only buys a little time at the cost of the Gold you needed to make the held front winnable. The fifty-tick construction window is the binding constraint: the post must start before contact, which means in a detected multi-front attack you build the post on the held front at the moment you detect the second attacker, not after the first attacker is already at your border. The failure mode is the overlapping-post illusion, where two posts on the same route feel like a 25x defense but the attack logic uses a qualifying post per tile and does not stack the multipliers, so the second post is only worth its Gold if it covers a different entrance or the next defensive line. In a two-front attack, the single most valuable use of post Gold is the one post on the held front that converts your best terrain into a front the attacker cannot take in the time the second attacker takes to overrun your conceded front.

A useful field test is to count how many of the tiles on the held front’s route fall inside the post’s thirty-tile range before you build; if only one or two route tiles are covered the post is a thin speed bump, but if the bend puts four or five route tiles in range the post converts the whole held front from a holdable line into a front the attacker cannot take on time.

Concession, terrain, and the retreat that is not a loss

Conceding a front is a deliberate economic action, not a failure, and the v0.34.24 model makes it a positive trade when done on the right front. When you concede a front, you pull the border inward a controlled amount so that every tile the attacker takes costs them their full magnitude-scaled troops and drains you only your average troops-per-tile, which on a thin front is a small number. The terrain choice is the lever: you concede through the cheapest terrain for you to give up and the costliest terrain for the attacker to take, which is highland or mountain, where the base magnitude is one hundred or one hundred twenty and the tile cost is twenty or twenty-five, so the attacker is slowed and made to pay more per tile than on plains. The retreat that is not a loss has three properties. First, it is planned before contact, so you know in advance which tiles you will give and you have moved the core’s production and the army’s reserve to the line you will defend, instead of discovering the retreat while already losing. Second, it is local to the conceded front, so you give the front’s border but keep the held front intact and keep the core’s tile count high enough that your troops-per-tile average, which is the per-tile drain number, stays low and the concession stays cheap. Third, it has a named re-entry point, the specific sector where, after you win the held front, you push back into the conceded front with the free army to reclaim the tiles at the attacker’s weakened troops-per-tile. The failure mode is the retreat that is a loss, which happens when the retreat is reactive rather than planned, when it pulls the core’s production into the conceded front so the troops-per-tile average rises and the drain accelerates, or when it has no re-entry point and becomes a permanent shrink. In a two-front attack, conceding one front in a planned retreat that keeps the core and the held front intact is almost always better than holding both fronts at half strength, because the per-tile drain on two half-held fronts compounds faster than the full drain on one cleanly conceded front, and the free army you save by not splitting is the army that wins the held front and enables the re-entry.

The depth of the concession is set by the number you can afford to lose before the held front is won: give enough that the attacker’s advance is slow and expensive, but not so much that your core’s tile count falls and the troops-per-tile average climbs high enough to make the concession itself unaffordable.

Mode and map adjustments: FFA, ranked 2v2, naval, and small maps

The allocation rule is the same in every mode, but the detection and the break change. In public FFA, the threat is usually opportunistic jumping or a detected pair, so the break is often not available and the core play is hold-one-bleed-two with a planned retreat; the FFA-specific adjustment is that the map is large enough that a conceded front can be re-entered from the side after you win the held front, so the re-entry point is usually available and the retreat is a true positive trade. In ranked 2v2, you are the team, so the multi-front problem inverts: you are the ganged side and the question is which of your two fronts to overload the enemy’s weaker member with, which is the attacker-side mirror of the break and the same arithmetic, because the enemy defender faces your two fronts and must allocate the way you allocate against a pair. The 2v2 adjustment is to spawn and place cities so that the enemy’s two-front problem is real, which means the two of you face two different sectors of the enemy border so the enemy must split, exactly the geometry you defend against in FFA. The naval adjustment is the one the community threads flag most: a coastal pincer is the worst multi-front case because the sea front does not respect your land defense posts, so the held front must be the land front and the sea front must be won locally with a warship or destroyer that is on the right coast before the landing commits, which is why the destroyer position is the first thing to fix in any coastal multi-front. The small-map adjustment is that there is less room to concede, so the planned retreat has fewer tiles to give and the troops-per-tile average rises faster, which makes the held front and the post even more load-bearing and makes the break more urgent, because on a small map you cannot afford to concede a whole front and the only way out is to delete one of the attackers locally. Across all modes the single invariant is the same: defend the front where your local force plus a post plus terrain makes the per-tile cost high, concede the front where you cannot, and break the pair by overloading its weaker member.

Across modes the one skill that transfers unchanged is naming the line early: the moment you can state your troops-per-tile figure out loud and match a front to it, the allocation, the concession, and the break all follow, and the modes that defeat you are almost always the ones where that number was never stated.

Decision table, failure modes, and the reading path

The decision table below is the compressed form of the whole guide: each row is a front state and the action, and the rule is that you fill the hold, the bleed, and the break slots from these rows before contact.

The failure modes, in order of how often they show up in the community threads, are the split-to-meet-every-front failure (splitting the army so no front is at the troops-per-tile line and losing on all of them), the wrong-coast naval failure (the warship on the wrong coast when the landing lands), the reactive-retreat failure (discovering the retreat while already losing instead of planning it before contact), the overlapping-post failure (spending two posts on one route that the attack logic does not double-stack), and the no-re-entry failure (conceding a front with no named point to take back, so the retreat becomes a permanent shrink). The reading path for the next step: read Defense posts for the single-build decision this guide uses as the held-front tool, read Land combat for the local force and terrain and counterattack choices on the held front, and read FFA trapped middle for the specific case where the two-front squeeze is the map shape rather than a teaming choice. The official version boundary for this article is the OpenFront v0.34.24 release, checked against the tag source for the per-tile defender loss, the defense post constants, and the troop-cap fix that closed the non-bordering exploit. If a later formal release changes the per-tile loss, the defense post multipliers, or the large-territory bonus, re-run this allocation framework instead of carrying the v0.34.24 numbers forward.