How the AI Treats Your Alliances in v0.35.3: Tribes Attack Traitors, Hard/Impossible Renew

Quick answer

v0.35.3 changed two things about how the AI handles alliances. First, tribes now break their alliance with a neighbour who turned traitor and attack that neighbour. If you flip an AI tribe against another AI tribe, you can now face both of them instead of losing only the support of the one you betrayed. Second, Hard and Impossible nations stop refusing every alliance renewal in small games. In a short game on a higher difficulty, a nation that would previously lock you out of re-signing your alliance will now renew it, so you can keep an AI partner alive into the late game. Both changes matter only when you are playing on a version that includes v0.35.3; on any older build the pre-v35 rules below still apply.

The two changes pull in opposite directions, and that is the whole shape of the patch. The tribe change makes a diplomatic move more expensive, because flipping one AI tribe can now turn two neighbours into two enemies. The nation renewal change makes a diplomatic move cheaper, because re-signing an AI nation on a higher difficulty in a short game is no longer a coin flip. So on v0.35.3 you should expect to spend less diplomacy on the nations you want to keep and to spend more care on the tribes you are tempted to flip. The rest of this guide is the expansion of that one sentence, section by section: the baseline decision tree first, so you know what did not change, then each of the two changes, then the framework that combines them, then two concrete scenarios with troop numbers, then the failure modes and the mode and map adjustments. If you only read one section, read the framework, because it is the part that changes your in-game decision rather than just your understanding. The tribes guide and the nations pressure guide are the player-side companions to this guide, and they cover the same diplomatic levers from the side you control.

The most important thing to understand before you go any further is that these two changes do not touch the part of the AI decision you feel most often, which is the traitor rejection. A nation or tribe you have betrayed will still refuse you most of the time on v0.35.3, exactly as it did before the patch, so the patch does not make it easier to win back an AI partner after you have switched sides. What the patch does is change what happens to the AI factions around you when you switch, and it changes whether you can re-establish a partnership you have not broken. That distinction is the whole difference between the two changes, and it is why the guide treats them as separate decisions even though they both appear in the same release note. If you have been playing on an older build and the new client shows v0.35.3 in the version field, you can expect the two behaviours described here from the first match you play on the new version, because both changes are engine-side and take effect as soon as the new engine is running. If you are on a server or a client that has not picked up v0.35.3, the older behaviour is still in force and the guidance in this guide only applies to the parts that did not change.

The pre-v35 baseline: how AI nations decide on alliances

Before you can see what v0.35.3 changed, you need the baseline, because the new changes are thin layers on top of a decision tree that has not changed. When an AI nation receives an alliance request it walks a fixed order of checks. The first is a random confusion step that scales down with difficulty: Easy nations sometimes ignore the strategy and accept or reject at random, while Impossible nations have no such randomness. The second check is whether the requestor is a traitor; nations reject a requestor who betrayed them about ninety percent of the time, which is why a nation that already lost you to a switch rarely takes you back. The third check caps the number of alliances a nation will hold: on Hard it avoids a requestor who is already allied with at least half of the eligible players, and on Impossible it avoids one allied with a quarter, because the engine wants enough enemies left to oppose the leading player. The fourth check is threat detection, and it works in the opposite direction from what most players expect: the nation accepts an alliance from a requestor that is too strong to fight, reasoning that it is safer to be allied with the threat than to be attacked by it. The fifth check is the game mode, and here the rejection chance climbs steeply with difficulty, from about a quarter on Easy up to a guaranteed refusal on Impossible in a team game. The sixth is the diplomatic relation: a relation worse than neutral rejects, while a friendly relation can trigger an immediate accept. The seventh is a per-difficulty alliance ceiling, with Impossible nations usually holding only two or three alliances. The eighth is an early-game bonus, where a request in the first minute or two has an extra chance of being accepted, highest on Easy and lowest on Impossible. Everything the v0.35.3 change adds sits on top of this order, and nothing about it changes the traitor rejection rate, the alliance caps, or the threat rule. The baseline is documented in the nation alliance decisions guide and in the nations mechanics page.

Check	What it does	Difficulty trend
Random confusion	Accepts or rejects at random	More on Easy, none on Impossible
Traitor	Rejects a requestor who betrayed the nation	About ninety percent of the time, all difficulties
Alliance count	Rejects a requestor already allied with too many	Caps tighten as difficulty rises
Threat	Accepts a requestor that is too strong to fight	Detection improves with difficulty
Team game	Extra rejection chance in team mode	Climbs from low on Easy to always on Impossible
Relation	Rejects below neutral, accepts when friendly	Same rule at every difficulty
Alliance ceiling	Stops once it holds enough alliances	Lower ceiling on Impossible
Early game	Extra acceptance in the first minutes	Higher on Easy, lower on Impossible

What changed for tribes: breaking and attacking a traitor neighbour

The tribe change is the more dramatic of the two, and it is a punishment rule rather than an availability rule. Before v0.35.3, if you were allied with an AI tribe and then signed a alliance with a neighbouring tribe, the first tribe would typically stop helping you; the alliance simply went cold. You lost a source of support, but the tribe you betrayed did not open hostilities against you, so the cost was a quieter one. v0.35.3 changes that cost into a visible one. Now, when a neighbour has turned traitor, the tribe breaks the alliance with that neighbour and attacks it. Read carefully: the tribe that attacks is the one whose neighbour turned traitor. If you sit between two AI tribes and you flip one of them against the other, the tribe on the other side of the betrayal is the one that turns on you, and you can end up fighting the side you just allied with as well as the side you just abandoned. The practical effect is that flipping an AI tribe against a neighbouring AI tribe is no longer a clean way to turn one enemy into an ally: it can become the move that turns two neighbours into two enemies at once. This matters most when your own expansion depends on one of those AI tribes holding the line while you build up, because the tribe you are leaning on is now the one most likely to open a second front. The tribes guide covers how to deal with a tribe that is already hostile; this section covers the new case where a tribe becomes hostile because of a decision you made. If you read the nations pressure guide you will see the same logic from the nation side: the engine wants to keep pressure on the leader, and a tribe that just watched its neighbour switch sides is the engine keeping that pressure on you.

The new behaviour also has a timing effect that is worth planning around. The break-and-attack fires the moment the neighbour turns traitor, not at the next alliance renewal, so the punishment is immediate and you do not get a grace period to withdraw the flip before the second front opens. That means the window to avoid the second front is before you sign the neighbouring tribe, not after, and the only way to reverse it once the attack has started is to defeat the tribe that turned on you or to bring a third faction into the fight to split its attention. The punishment also applies in both directions: if the AI tribe is the one that flipped, not you, and it broke its alliance with a neighbouring AI tribe to attack that neighbour, you will see the same break-and-attack play out between two AI factions on the map, which is a useful tell for what the engine is doing to the factions around you when you have not personally made the move. Watching two AI tribes break and attack each other is confirmation that the new punishment rule is active on the version you are playing, and it is the same rule that will apply to you the moment you sign one of them against the other. The practical takeaway is that on v0.35.3 the cost of a tribe flip is no longer a quiet loss of support but an immediate, visible second front, and the only way to avoid that second front is to not make the flip unless your own army can absorb it.

What changed for Hard and Impossible nations: small-game renewals

The second change is quieter but it fixes a real lockout that small-game players on higher difficulties ran into. In a small game, the window for diplomacy is short: you sign an alliance early, you fight in the middle, and by the time the endgame pressure builds the alliances you started with have either been broken or are about to expire. Before v0.35.3, on Hard and Impossible, a nation could refuse every alliance renewal, which meant that once a nation decided not to re-sign you could never get that nation back, no matter how useful it was to hold as a partner. In a short game that refusal was almost fatal, because you had only one or two chances to establish your diplomatic map and a single refusal locked you out of a partner for the rest of the match. v0.35.3 removes that blanket refusal. Now, in a small game, a Hard or Impossible nation will not refuse every renewal; it will renew under the same conditions the rest of the decision tree gives it, which means a nation that is not a traitor relationship, is not over its alliance ceiling, and has a workable relation can re-sign. The change does not make Hard and Impossible nations friendly; it makes them willing to continue a relationship they had already started, as long as the baseline checks still pass. If you betrayed the nation, the about-ninety percent rejection from the traitor rule still applies, so the renewal change does not let you re-sign a nation you switched. The change is specifically about the renewal window: the moment after an alliance has lapsed where the only question is whether to re-establish the one you had. That window now exists on the higher difficulties in small games, and it is the single most important new tool for holding an AI partner into the late game on a short map.

There is a subtle but important boundary to this change that players miss, and it is the boundary between a renewal and a fresh request. The change applies to the renewal window, which is the moment after an existing alliance has lapsed and the only question is whether to re-establish the one you had. It does not apply to a fresh alliance request to a nation you have never been allied with, and that fresh request still walks the full decision tree from the top, including the alliance-count cap and the threat rule. So the renewal change is a tool for keeping a partner you already have, not a tool for acquiring a new one, and if you try to use it to sign a nation you have never allied with on a small Impossible game, you will still be subject to the same refusals that existed before the patch. The practical consequence is that the best version of a small-game plan on v0.35.3 is to establish the one nation alliance you want in the first minute or two, when the early-game acceptance bonus is highest, and then renew it at every lapse rather than trying to sign a second nation later, because the second nation is a fresh request and the first nation is a renewal. That ordering, sign early and renew often, is the single most important diplomatic habit the patch enables, and it is the habit that separates a small-game plan that works on v0.35.3 from one that still fails the way it failed before the patch.

The decision framework: when to flip an AI tribe, when to hold a nation

The two changes point in opposite directions, so the framework is a pair of rules. The first rule is about tribes: do not flip an AI tribe against a neighbouring AI tribe unless you can afford to fight the tribe you flipped. The pre-v35 cost of a flip was the loss of one support source; the v0.35.3 cost can be two enemies at once. So the test before you flip is whether your own military is strong enough to absorb a simultaneous attack from the tribe you just allied with and the tribe you just abandoned. If you are building up and your capital is still vulnerable, the flip is the wrong move even if the tribal enemy on the other side is currently weak, because the allied tribe can now turn on you while you are exposed. The second rule is about nations on Hard and Impossible: hold a nation alliance you want and renew it in a small game, because the renewal window that used to close on you is now open. The test before you let a nation alliance lapse is whether that nation is the partner that is covering your flank while you push. If it is, you should actively renew rather than let it expire, because the pre-v35 risk was that the renewal would be refused and you would be left with no partner in the endgame. The framework collapses to one sentence: on v0.35.3, be more conservative about flipping AI tribes and more proactive about renewing AI nations on higher difficulties in short games. Both halves of that sentence follow directly from the two release changes, and neither of them changes the baseline traitor rejection or the alliance caps, so you do not need to re-learn the whole decision tree to use them.

A third element rounds out the framework, and it is the element that tells you which of the two rules to apply first when a match is unfolding. The element is the identity of the faction you are most afraid of. If the faction you are most afraid of is an AI tribe, the flip rule is the binding one, because the risk to you is that a flip turns that tribe into an enemy or turns the tribe you are leaning on into an enemy, and the framework tells you to be conservative. If the faction you are most afraid of is an AI nation, the renewal rule is the binding one, because the risk to you is that a lapsed alliance leaves you alone in the endgame, and the framework tells you to be proactive. If the faction you are most afraid of is another player, both rules are less binding, because the AI factions are not the main threat, and you can afford to be more aggressive with both flips and renewals since the cost of a mistake is lower. The framework therefore collapses further to a three-step test: identify the faction you are most afraid of, decide which rule is binding based on whether that faction is a tribe or a nation, and apply that rule with the appropriate level of caution. The whole framework is a way of deciding, in the middle of a match, which of the two v0.35.3 changes is the one that matters for your next diplomatic move, and that decision is the one that changes your in-game play more than anything else in this guide.

Scenario one: flipping a tribe between two neighbours on a large map

Imagine a large-map FFA where you are the leader and two AI tribes sit on either side of you. Tribe A is to the east and is currently allied with you; tribe B is to the west and is currently enemy. Tribe A and tribe B share a border, and tribe B is the tribe you are most afraid of because it has the most troops on the map. Your instinct is to sign tribe B and turn it against tribe A, because that would give you a strong ally on the west and turn your east into an open flank. Under the pre-v35 rules that would have worked the way you expected: tribe A would stop helping you, but it would not attack you, and you would be left with tribe B as a new partner and a free easterly expansion. Under v0.35.3 the move is different. When tribe A sees that its neighbour, tribe B, has turned traitor by allying with you against it, tribe A breaks the alliance with you and attacks you. So instead of losing one support source you now face tribe A, which has troops, and you have just given tribe B a reason to be less useful to you because it is busy fighting alliance flip.

The numbers that matter are the troop counts: if tribe A has, say, a hundred thousand troops and you are at sixty thousand with your capital still exposed, the flip is a loss, because you traded a quiet loss of support for a two-front war you cannot win. The correct version of the move is to only flip tribe B if your own troop count is high enough that you can absorb a tribe-A attack while tribe B is still distracted; if you are not strong enough, the better move is to keep tribe A allied and apply pressure to tribe B through trade or a nuke instead of a full alliance flip.

The same move works in your favour when the numbers tilt the other way. If tribe B, the enemy tribe you want to sign, is the one with the larger army and tribe A, your current allied tribe, is smaller, then flipping tribe B against tribe A is the move that saves you, because you are trading a weak ally for a strong one and the tribe-A attack you now have to absorb is coming from the smaller faction. The test is always the comparison of your own troop count against the troop count of the tribe that will now attack you, not against the tribe you are signing. A common mistake is to compare your troops to tribe B and decide the flip is safe because tribe B is large, when in fact the relevant comparison is your troops against tribe A, the side that will turn on you. The second comparison that matters is your capital exposure: a flip is safe when your capital is defensible even with a single front, and unsafe when your capital only survives if one of the AI tribes is holding the line. On a large map you usually have time to build that defensive capital, which is why the large-map scenario is the one where a careful flip can still pay off, even though the small-map version of the same move is almost always a mistake.

Scenario two: holding a nation partner in a small Hard game

Now imagine a small-map game on Hard where you are not the leader and you need one AI nation as a partner to survive the endgame. You signed the nation early, and it has been holding the line while you build. By the middle of the game your alliance has lapsed, and the question is whether to renew. Under the pre-v35 rules, on Hard, the nation could refuse the renewal outright, and in a small game you have no time to find a replacement, so the safest assumption was that you were now on your own. Under v0.35.3 the assumption changes. Because you did not betray the nation, the traitor rule does not fire; because you are not the leader, the alliance-count cap that protects the leader is not the binding constraint; and because the relation is workable, the renewal goes through. The decision is now about whether the nation is worth holding, not about whether you can even hold it. If the nation is covering your weakest flank, the answer is yes, and you should renew immediately rather than wait for the next lapse, because a second lapse on a small map is often the end of the game. If the nation is on your strongest flank and you have enough troops to defend that side on your own, the answer is no, and letting it lapse is fine because you do not need the partner. The change matters because it converts a coin flip into a choice: before v0.35.3 the renewal was sometimes refused no matter what you did, and now it is refused only when one of the baseline checks fails, so you can plan around it.

The renewal decision also depends on the state of the nation’s own military, which the pre-v35 player could not influence but can now factor in. If the nation you are renewing is itself losing troops to a third faction, renewing it is a bet that the nation will survive long enough to be useful, and the bet is weaker the more of the nation’s army is already gone. The right move in that case is to renew only if the nation is the partner covering your most exposed flank and the third faction attacking the nation is not also attacking you, because a nation that is about to be destroyed by a faction that is also attacking you is a partner that will not save you in the endgame no matter how fresh the renewal is. The opposite case is the nation that is winning its own fights: renewing that nation is a strong bet, because the nation is likely to survive and to bring its winning momentum into the endgame, and the early-game acceptance bonus that got you the alliance in the first place is now working in your favour at the renewal too, because the engine is willing to keep a strong partner around rather than let it walk. The practical summary is that the renewal window on v0.35.3 is open, but the decision of whether to use it still depends on the state of the nation you are renewing, and the patch removed the worst case without removing the need to judge.

Failure modes and countermeasures

The most common failure after the tribe change is the double-flip: a player who signs tribe B against tribe A, gets turned on by tribe A, and then, in panic, signs another neighbour against tribe A, which turns that neighbour on them too. The countermeasure is the test from the framework: before you flip, confirm your own troop count is high enough that you can take one AI attack without losing your capital, and never chain two flips in the same game. The second failure is the renewal mistake: a player who lets a useful nation alliance lapse on a small Hard game, assumes the pre-v35 refusal, and never renews, only to be alone in the endgame. The countermeasure is to check the baseline before you let a lapse stand: if you did not betray the nation and the relation is workable, renew it, because the blanket refusal that made renewal a gamble is gone. The third failure is the false confidence: a player who reads the renewal change and assumes Hard and Impossible nations are now friendly, signs too many of them, and ends up over-extended. The countermeasure is to remember that the renewal change only opens the window; the traitor rejection, the alliance cap, and the threat rule are all still in force, so a nation you betrayed will still refuse you about ninety percent of the time and a nation that is already allied with half the map will still refuse you on Hard. The fourth failure is the mode mistake: in a team game on Impossible the team-game rejection is a guaranteed refusal, so no amount of renewing will re-establish a nation alliance you lost, and the only lever left is the diplomatic relation and the threat rule. The fifth failure is the difficulty assumption: on Easy the random confusion is high, so a tribe that breaks your alliance may do it for reasons unrelated to your betrayal, and you should not read the confusion-driven break as a punishment you can negotiate your way out of.

A sixth failure, and the one that is easiest to make because it is a reading error rather than a play error, is the one where a player reads the tribe change as a general increase in AI aggression and decides the whole version is more hostile, so they play a defensive match that is not warranted. The tribe change is a specific punishment for a specific move, the flip of one AI tribe against a neighbouring AI tribe, and it does not make the AI more aggressive in every other situation. A player who is not flipping AI tribes is not exposed to the new second front, and playing a defensive match out of fear of a punishment that does not apply to their play is a cost the patch does not actually impose. The same reading error shows up on the nation side, where a player reads the renewal change as a sign that the AI is now more cooperative and signs more nations than the map can support, which is the false-confidence failure again, and the countermeasure is the same: the patch changed two specific decision paths, not the overall temperament of the engine, so the only situations where the new behaviour matters are the two it names.

Mode and map adjustments

The two changes interact with the map size and the game mode in ways worth separating out. On a large map the tribe change is the one that matters most, because a large map gives you enough room to be caught between two AI tribes, and it also gives you enough time to recover from a bad flip if your troop count is high. On a small map the nation renewal change is the one that matters most, because the diplomacy window is short and a lapse is almost the end of the game, so the open renewal window is the difference between holding a partner and being alone. In a team game the nation change is almost useless on Impossible, because the team-game rejection is a guaranteed refusal, so the renewal window is closed by the mode before the difficulty rule can open it; on Easy and Medium the team-game rejection is only a chance, so the renewal change still applies. In FFA both changes apply fully, because there is no team-mode rejection to interfere, so FFA is where the v0.35.3 diplomacy is most visible and most worth planning around. The nations mechanics page has the mode-specific rules for nations that this section builds on, and the changelog for v35 lists the full set of changes in the v35 series so you can see the v0.35.3 entries in context. The practical summary is that on a large FFA map you should be more careful about flipping AI tribes, and on a small Hard or Impossible game you should be more proactive about renewing AI nations, and in a team game on Impossible the nation change gives you very little, so you should plan the match without relying on a re-signed nation.

One more adjustment is worth naming explicitly: the map size also changes which of the two failures you should fear most, and therefore which one you should plan for. On a large map the failure to watch for is the double-flip, because a large map gives you the room to be caught between two AI tribes and the time to make the mistake twice, so your plan should include a hard rule about never chaining two flips in the same match. On a small map the failure to watch for is the lapsed renewal, because a small map gives you no time to recover from a missed renewal and no room to find a replacement partner, so your plan should include a hard rule about renewing a useful nation the moment its alliance lapses. The mode changes which of those two plans is the binding one: in FFA both plans are in force at the same time, in a team game on Impossible the renewal plan is mostly useless so you should lean on the double-flip plan, and in a team game on Easy or Medium the renewal plan is still worth keeping because the team-game rejection there is only a chance rather than a certainty. The difficulty changes how aggressively each plan has to be applied: on Easy the random confusion makes both plans less reliable, so you should plan for a tribe that may break your alliance for reasons unrelated to your betrayal and for a nation that may renew it for reasons unrelated to your merit; on Impossible the plans are more reliable because the engine is less random, so you can count on the renewal window opening when the baseline checks pass and you can count on the traitor punishment firing when a neighbour switches sides.