1.4.1's Broken Diplomacy - and what we can learn from it

[Sidrath]

As has been reported in the 1.4.1 Beta Patch Notes thread and Customer Support forums, AI diplomacy in the 1.4.1 beta branch is badly broken. The AI will immediately declare war on any faction the player joins, often resulting in ~4 simultaneous wars. The other factions do not interact with each other in any way: they will stay locked in their original wars, never make peace with each other, and never initiate new wars. This is a major AI bug and will hopefully be addressed soon.

It occurred to me that there is a silver lining to this glitch. In any fresh playthrough, AI factions will stay locked in their original wars indefinitely, barring player intervention. This provides precise insights into each faction's relative capabilities against a single neighbour. Having a safe border (as Vlandia and Khuzait do, on the edges of the map) is immaterial: every border in 1.4.1 is safe apart from the sole warfront prescribed by the game's starting conditions. These are: Khuzait vs Northern Empire, Vlandia vs Sturgia, Battania vs Western Empire, Aserai vs Southern Empire.

Starting Conditions

As a reminder, these are the starting conditions of the game:



Simulations

These are the outcomes after 4 separate simulations in which the player takes no action, for a period of 4 years ( Spring 1088 ). In effect, these are the outcomes of the 4 wars forming the game's starting conditions: Khuzait vs Northern Empire, Vlandia vs Sturgia, Battania vs Western Empire, Aserai vs Southern Empire.









Results

• The fact that Northern Empire borders with Sturgia, Battania, Western Empire, Southern Empire, Khuzait does not matter. They will only be fighting Khuzait, and lose 1v1 in repeated simulations.
  
  
• The fact that Sturgia could struggle to reinforce far-away Tyal against Khuzait, horrible geography and travel times, etc, is mostly irrelevant in 1.4.1. They will only ever be fighting over the Varcheg/Omor region, to hold off Vlandia, their locked foe - and they will lose 1v1 in repeated simulations.
  
  
• Aserai vs Southern Empire appears evenly matched, very slightly favouring Southern Empire.
  
  
• Battania vs Western Empire appears evenly matched.

What can we glean from this?

• This will not come as a shock to anyone who's been playing the game for a while, but Khuzait is far too strong. I wouldn't say Northern Empire is necessarily weak: the other Emps have the same troops and manage just fine against their respective foes.
  
  
• Either Vlandia is too strong, or Sturgia is too weak; perhaps a bit of both but I'm inclined to blame Sturgia's infantry focus and lack of cultural campaign mobility to offset it. Importantly: these outcomes do not stem from factions being surrounded, or war-crazed leader AI (Raganvad can't get dogpiled in 1.4.1 unless you join Sturgia).
  
  
• Some modest proportion of the game's snowballing woes since launch could perhaps be attributed to starting conditions that give Vlandia and Khuzait free land to nibble on from their comparatively weaker starting rivals. Those early wins go on to dictate the tempo of the entire game, with Vlandia and Khuzait building on early victories (more wealth, more fiefs, higher tier armies).

Please note that factions with their backs to a safe wall, like Vlandia and Khuzait, have a very real advantage which will manifest again once diplomacy is fixed. The purpose of these experiments was to take advantage of 1.4.1's unique opportunity to evaluate faction balance in a 1v1 setting. Sturgia and Northern Empire's baseline inferiority to Vlandia and Khuzait stacks with their geographical weaknesses of stretched supply lines (Sturgia) and vulnerable open borders (Northern Empire), resulting in familiar outcomes across playthroughs.

Recommendation

The next beta branch should consider:

• Nerfs to Khuzait's campaign performance (e.g. fix their campaign movement speed bonus to only affect cavalry, or drastically reduce the amount - in practice Khuzait culture currently gives a free 0.4-0.6 'Feats' speed bonus permanent effect, even in all-infantry armies)
  
  
• Buffs to Sturgia's campaign performance (giving them their culture bonus, so they can successfully pick their battles at least on home turf, might be enough).

 

[Ishtarin]

I'd rather have the Khuzait trait affect only the cavalry bonus (something like 30% extra speed if an entire army is mounted) rather than giving a straight 10% buff to the entire party if a single horse exists.

 

[Nakh]

Well, this is old news, generally

Norther Empire always lose Amprella to Khuzaits
Western Empire almost every time lose Lageta to Battanians
Sturgians always lose Varcheg to Vlandria, most of times lose Revyl and generally first to collapse in every game.

I'm not sure campaign speed will solve Khuzait problem, they still will have huge cavalry advantage in field battles. Especially against empire factions and Sturgians, which don't have proper cavalry in meaningful quantity.
As for Sturgia, looks like suicidal AI is the main problem. AI rarely care to defend own cities.

 

[Dabos37]

Vlandia is in the same boat than Khuzaits. Vlandia start with 8 towns and settlements are pretty close to each other. Plus cavalry is still OP in autocalc and this faction has probably the best or at least one of the best map start position. Improving autocalc and maybe giving some slightly buffs/nerfs for factions in campaign map would be enough IMO.

On the other hand, I remember to read that Vlandia is one of the strongest factions at start of campaign but internal conflicts would harm this faction. I suppose that this will be the case once rebelions/revolts get implemented.

(And yes, the current Diplomacy AI is pretty sad ?)

 

[Maggot Toes]

The AI seems to behave on power bars, taking into account the troop quality and not raw numbers. Are we saying the power bars are not correct and therefore the AI does not understand it will lose in autocalc? From the behavior I've see I tend to think it knows the results but I haven't separated troop tiers from cavalry when engaging. A recent example: I just had a 600 man AI army refuse to engage my 300, and then when I attacked sure enough my power bar was greater, and my army is about 50% cavalry.

 

[Sidrath]

Small addendum: please note the precise banners flying over conquered lands in all 4 simulations. Lots of Derthert fiefs, lots of Monchug fiefs. Ruler greed is still prevalent in 1.4.1, the nerfs to Influence gain (nerfing forums Influence, slowing down construction) have had no impact.

 

[froggyluv]

The AI seems to behave on power bars, taking into account the troop quality and not raw numbers. Are we saying the power bars are not correct and therefore the AI does not understand it will lose in autocalc? From the behavior I've see I tend to think it knows the results but I haven't separated troop tiers from cavalry when engaging. A recent example: I just had a 600 man AI army refuse to engage my 300, and then when I attacked sure enough my power bar was greater, and my army is about 50% cavalry.

This kinda depress me. We really, really need a robust Auto-calc system that we can trust is giving thorough results. Defeats the point of a strategy game if the battle determinations are too simplistic

 

[OTuphlos]

Here are a couple more results

 

[federally]

This kinda depress me. We really, really need a robust Auto-calc system that we can trust is giving thorough results. Defeats the point of a strategy game if the battle determinations are too simplistic

The patch notes mention they improved it for 1.4.1, but I haven't seen anyone mention figuring out how they improved it.

I know prior to this patch it was pretty pathetic in how simple it was