牧游社 牧有漢化翻譯

CK3 Dev Diary #104: AI AI AI

rageair, Design

Greetings!

大家好！

One of the trickiest bits about GSGs is making a good AI. You want it to pose a challenge and play well, but you don't want it to ruin the player's fun by acting out of character. We've read a lot of discussions and threads in various places across the community, and we understand that there is room for improvement. Over the summer we've been working on a few aspects of the AI, specifically the economy (and a little bit of diplomacy). We've not focused on warfare, except for one thing which we'll get to later.

在大戰(zhàn)略游戲的制作中，最棘手的問(wèn)題之一就是做出好的AI。制作者既希望AI能玩得夠好，對(duì)玩家形成挑戰(zhàn)，又不希望AI過(guò)于違和，毀掉玩家的游戲樂(lè)趣。我們?cè)谟螒蛏鐓^(qū)的各個(gè)地方讀到了很多討論，也認(rèn)識(shí)到在這方面仍有改進(jìn)的空間。在夏季期間，我們一直在改進(jìn)AI的幾個(gè)部分，尤其是經(jīng)濟(jì)（以及一些外交）。除了本期日志中會(huì)提到的一點(diǎn)之外，我們沒(méi)有涉及到其它關(guān)于戰(zhàn)爭(zhēng)的內(nèi)容。

All changes outlined in this Dev Diary will be released in a future update.

這篇日志中列出的所有改動(dòng)將會(huì)在未來(lái)的某次更新中實(shí)裝。

For the player to feel motivated to do well, the AI must show that it can do well alongside you. It doesn't have to be fantastic or play optimally (in fact, that's usually immersion-breaking), but it must be capable of progress. We want the AI in CK3 to be able to provide a good challenge and do well enough to motivate you to do even better. Also, we want it to do this without cheating.

為了讓玩家有動(dòng)力好好玩下去，AI必須表現(xiàn)出它有跟玩家旗鼓相當(dāng)?shù)膶?shí)力。AI并不需要極強(qiáng)，甚至做出最優(yōu)化的操作（實(shí)際上，那樣常常會(huì)破壞沉浸感），但是AI必須要有進(jìn)步的能力。我們希望CK3中的AI能夠造成一定的挑戰(zhàn)，激勵(lì)玩家做到更好。另外，我們也希望AI能在不作弊的前提下達(dá)到這個(gè)目的。

Before delving deeper into what we've done, I'd like to explain some of the challenges we set out to solve for the Economy:

在深入介紹我們所做的東西之前，我想首先解釋一下在經(jīng)濟(jì)方面我們著手解決的一些問(wèn)題：

Economically, all AIs currently play the same. We want their personality to affect how an AI decides to spend their hard-earned gold, and on what. You should be able to look at an AIs personality traits and, roughly, be able to tell what they’re up to.
目前在經(jīng)濟(jì)上，所有AI的玩法都是一樣的。我們希望AI在消費(fèi)自己辛苦賺來(lái)的錢(qián)時(shí)，所做的決定以及所購(gòu)買(mǎi)的東西能受到其個(gè)性的影響。假如你查看AI的個(gè)性特質(zhì)，就基本上能知道AI大概在干些什么。

A player understands that there are certain ways of using gold that are optimal and appropriate regardless of their situation; that there are certain risks worth taking. We want the AI to understand this, employing decisions that make them prosper and are suitable to the stage the game is in.
玩家們都很清楚，有一些最佳的花錢(qián)方式在各種情況下都很合適；也明白有些風(fēng)險(xiǎn)是值得冒的。我們希望AI也能學(xué)到這點(diǎn)，讓AI做出能讓自己致富，又適合游戲階段的決定。

We want the AI to build more buildings, so that the world becomes more and more developed over time. Right now, only players realize the value that buildings bring in the long run, while AIs spend most of their gold on wars.
我們希望AI能夠建造更多的建筑，這樣的話整個(gè)世界就會(huì)隨時(shí)間變得越來(lái)越發(fā)達(dá)。目前，只有玩家知道建筑能夠帶來(lái)的長(zhǎng)期收益，AI會(huì)把大部分金錢(qián)花費(fèi)在戰(zhàn)爭(zhēng)上。

The power of a ruler lies within their domain. The AI, much like the player, should strive to build a base of power and cultivate it over time.
統(tǒng)治者的權(quán)力建立在其直轄領(lǐng)之上。AI應(yīng)該像玩家一樣，努力經(jīng)營(yíng)一塊基本盤(pán)，并且慢慢對(duì)其進(jìn)行開(kāi)發(fā)。

…and here are some of the challenges we set out to solve for Diplomacy:

……以下這些是在外交領(lǐng)域我們希望解決的問(wèn)題：

AI realms tend to fall apart, and never reform. Vast seas of counts and dukes, unwilling to accept the overlordship of neighboring kings. We want the AI to be able to diplomatically expand, much like the player is able to. New realms with great monarchs should rise out of the ashes of former empires.
AI國(guó)家都很容易碎，而且碎了之后就回不來(lái)了。海量的獨(dú)立伯爵、公爵，都不愿意向相鄰的國(guó)王俯首稱臣。我們希望AI能夠用外交方式擴(kuò)張，就和玩家們一樣。在偉大君主的帶領(lǐng)下，新的國(guó)家將在舊帝國(guó)的廢墟中崛起。

A stable realm is a prosperous realm, and there are many tools for achieving this. The AI should not hold back when using available methods of stabilization, be they diplomatic schemes or underhanded council actions.
國(guó)家穩(wěn)定了，就會(huì)繁榮昌盛，要達(dá)到這點(diǎn)的方法有很多。AI應(yīng)該放開(kāi)手使用各種各樣的手段維穩(wěn)，包括外交密謀或秘密內(nèi)閣行動(dòng)等。

The AI should understand that Tyranny is sometimes appropriate. It's for the greater good of the realm. Within reason, of course - personality and relationships must play a large role here. As it stands, the AI will never willingly take on Tyranny, under any circumstance.
AI應(yīng)該學(xué)會(huì)，有時(shí)候暴政是合適的。都是為了國(guó)家的利益嘛。當(dāng)然了，AI的個(gè)性以及關(guān)系也是有很大影響的。目前，在任何時(shí)候AI都不會(huì)自愿進(jìn)行暴政行為。

…and for Warfare:

……至于戰(zhàn)爭(zhēng)：

In real life, the Mongols were an existential threat. Currently, in CK, they are not. We want them to be. There is one pivotal change that enables this: the ability to gather armies closer to their target.
在史實(shí)中，蒙古人是生死存亡的威脅。而在當(dāng)前版本的CK中卻完全不是這么一回事。我們希望他們能和歷史上一樣。因此我們做了一項(xiàng)至關(guān)重要的改動(dòng)：讓他們能在離敵人更近的地方集結(jié)軍隊(duì)。

The Economy, fools!

笨蛋，問(wèn)題是經(jīng)濟(jì)！

The key to success in CK3 is a strong and stable economy, and such a thing is driven by buildings (and, partly, MaA). We've set out to change how the AI uses its gold so that it builds up its realm more. We've taken considerations to make sure that the AI does this in a way so that it, for example, doesn't become too easy to defeat for the player. The backbone of this work rests upon a framework of Economic Archetypes.

CK3中，通向成功的關(guān)鍵是強(qiáng)健的經(jīng)濟(jì)，而經(jīng)濟(jì)是由建筑驅(qū)動(dòng)的（以及部分程度上也與兵士有關(guān)）。我們已經(jīng)著手改變AI花錢(qián)的方式，讓AI在國(guó)內(nèi)做更多的建設(shè)。我們也做了很多的考慮，比如說(shuō)，不能讓AI沉迷建設(shè)被玩家輕易打爆。這項(xiàng)改動(dòng)的基礎(chǔ)在于經(jīng)濟(jì)原型的框架上。

Economic Archetypes

經(jīng)濟(jì)原型

There are four economic archetypes that AI rulers can fall into, of which three are significant: Warlike, Cautious, and Builder. If a character falls into none of these archetypes, they will be Unpredictable and use aspects of other archetypes in a semi-random fashion.

所有AI領(lǐng)主分為四種經(jīng)濟(jì)原型，其中三種比較重要：好戰(zhàn)、謹(jǐn)慎、建設(shè)。如果一個(gè)角色不屬于這三種原型，他就屬于善變，以半隨機(jī)的方式使用其它原型的各個(gè)方面。

1. Warlike

1、好戰(zhàn)

[Image - Warlike Personality Example]

[圖中展示的是好戰(zhàn)個(gè)性的范例]

The Warlike archetype wants to be ready to declare war, always. This is the archetype that is most similar to how the AI behaves in the game right now. They will put their gold into a war chest before considering other options. If a Warlike archetype is at peace for an extended period of time, they will use the gold in their war chest to develop the realm - but they will first and foremost prepare for war, even in the early game.

好戰(zhàn)原型隨時(shí)準(zhǔn)備開(kāi)戰(zhàn)。當(dāng)前游戲中AI的表現(xiàn)就與這個(gè)原型最為相似。在考慮其它選擇之前，他們會(huì)優(yōu)先把自己的金錢(qián)投入戰(zhàn)爭(zhēng)金庫(kù)里。如果好戰(zhàn)原型的角色長(zhǎng)期處于和平之中，他們就會(huì)利用戰(zhàn)爭(zhēng)金庫(kù)里的錢(qián)去擴(kuò)展自己的疆土——哪怕是在游戲早期，他們也會(huì)首先整軍備戰(zhàn)。

2. Cautious

2、謹(jǐn)慎

[Image - Cautious Personality Example]
[圖中展示的是謹(jǐn)慎個(gè)性的范例]

Not to be confused with cowardly, the Cautious archetype wants to be prepared for having war be declared upon them. They are somewhat averse to declaring offensive wars, instead preferring a slow buildup. This archetype will save up a minimum buffer of gold, depending on their tier. When choosing to invest gold into buildings or MaA, they will evaluate the state of their military, how long they've been at peace, how many allies they have, and their level of dread - depending on these factors, they will feel 'safe' and invest more gold than they would otherwise, while keeping the aforementioned minimum buffer.

與懦弱不同，謹(jǐn)慎原型希望做好準(zhǔn)備應(yīng)對(duì)有人向他們宣戰(zhàn)的情形。他們?cè)谀撤N程度上不喜歡宣布進(jìn)攻性的戰(zhàn)爭(zhēng)，而喜歡緩慢進(jìn)行積累。謹(jǐn)慎原型會(huì)攢下最低限度的緩沖儲(chǔ)備金，數(shù)額取決于他們的頭銜等級(jí)。當(dāng)選擇將金錢(qián)投資于建筑或兵士時(shí)，他們會(huì)評(píng)估他們的軍隊(duì)狀況、已經(jīng)和平了多久、有多少盟友，以及他們的恐怖值等級(jí)——根據(jù)這些因素，他們會(huì)感到 "安全"，并在保持上述最低儲(chǔ)備的情況下，進(jìn)行更多金錢(qián)投資。

Non-bold AIs tend to be Cautious. Common traits include Patient, Calm, Craven, Paranoid, and Content. Rulers with the Fickle or Lunatic trait will never be cautious. Rulers with the Stoic Ethos or the Stalwart Defenders Cultural Tradition are also drawn to this archetype.

不大膽的AI往往是謹(jǐn)慎的。常見(jiàn)的特質(zhì)包括耐心、冷靜、膽小、偏執(zhí)和安于現(xiàn)狀。具有善變或瘋狂特質(zhì)的統(tǒng)治者永遠(yuǎn)不會(huì)謹(jǐn)慎行事。具有“堅(jiān)忍”教義或“堅(jiān)定的防御者”文化傳統(tǒng)的統(tǒng)治者也被納入到這個(gè)原型中。

3. Builder

3、建設(shè)

[Image - Builder Personality Example]
[圖中展示的是建設(shè)個(gè)性的范例]

The Builder archetype wants to invest all of its gold into the realm. They will empty their treasury in order to build buildings and MaA as quickly as they can, and will also tend to construct more economical buildings and new holdings than other archetypes. They are bold and a bit reckless, only saving up a war chest if they are under a direct threat (such as a strong faction). This is the rarest of all archetypes.

建設(shè)原型希望將其所有的金錢(qián)投入到領(lǐng)地中。他們會(huì)清空自己的國(guó)庫(kù)，以便盡快建造建筑和兵士，也會(huì)傾向于建造比其他原型收益更高的建筑和新的地產(chǎn)。他們很大膽，有點(diǎn)魯莽，只有在受到直接威脅時(shí)（比如一個(gè)強(qiáng)大的派系）才會(huì)積攢戰(zhàn)爭(zhēng)資金。這是所有原型中最罕見(jiàn)的一種。

Bold AIs with a propensity for building and no particular inclination for war fall into this archetype. Common traits include Calm, Patient, Diligent, Generous, Stubborn, Profligate, and Improvident. Where Warlike AI's want to expand their realm, the Builder wants to build up their domain. Rulers with the Domain Focus also tend to fall into this category.

大膽的AI，帶有建設(shè)的傾向，沒(méi)有特別地偏好戰(zhàn)爭(zhēng)，會(huì)被歸入這種原型。常見(jiàn)的特質(zhì)包括冷靜、耐心、勤奮、慷慨、固執(zhí)、揮霍和貪婪。好戰(zhàn)的AI想要擴(kuò)大他們的領(lǐng)國(guó)，而建設(shè)的AI則想要建設(shè)他們的直轄領(lǐng)。擁有領(lǐng)地重心的統(tǒng)治者也傾向于歸入這個(gè)類別。

4. Unpredictable

4、善變

[Image - Unpredictable Personality Example]
[圖中展示的是善變個(gè)性的范例]

The Unpredictable archetype doesn't have a propensity towards anything in particular, and might decide upon any strategy. They can decide to boom their economy much like the Builder archetype, but they do so randomly (weighed up by traits such as Diligent or Architect), and the amount will be modified by how bold they are. Otherwise, they will want to build up their realms to a minimum degree (higher than a Warlike archetype), keep a decent buffer (lower than a Cautious archetype), and go to war at a normal pace.

善變?cè)蜎](méi)有任何特別的傾向性，可能會(huì)依照任何策略進(jìn)行決策。他們可能像建設(shè)原型一樣決定增進(jìn)他們的經(jīng)濟(jì)，但這是隨機(jī)的（勤奮或建筑師等特質(zhì)會(huì)增加權(quán)重），而且投資數(shù)量將根據(jù)他們的大膽程度進(jìn)行修正。另一方面，他們也可能把自己的領(lǐng)地按最低程度進(jìn)行建設(shè)（比好戰(zhàn)原型的建設(shè)程度高一些），保持適當(dāng)?shù)木彌_儲(chǔ)備（比謹(jǐn)慎原型的儲(chǔ)備低一些），并以正常的節(jié)奏去打仗。

AIs who do not qualify for any of the above categories will fall into this archetype.

不符合上述任何類別的AI將屬于這種原型。

Economic Stages

經(jīng)濟(jì)階段

The AI divides its economic game into three stages; Capital Development, Domain Development, and Late Development. The first two of these are 'early' stages that should be all but over roughly 100 years into the game.

AI將其經(jīng)濟(jì)操作分為三個(gè)階段：首都發(fā)展、直轄領(lǐng)地發(fā)展和后期發(fā)展。其中前兩個(gè)是“早期”階段，應(yīng)該在游戲進(jìn)行到100年左右時(shí)全部結(jié)束。

Each of these stages modifies how the AI spends its gold, which is then again modified by the economic personalities above.

這些階段中的每一個(gè)階段，都會(huì)對(duì)AI如何花費(fèi)金錢(qián)進(jìn)行修正，然后由上述的經(jīng)濟(jì)個(gè)性再進(jìn)行修正。

1. Early Capital Development

1、早期首都發(fā)展

The most important holding to develop is your capital, I’m sure you all agree. The AI will strive quite hard to fill all of the empty slots in its capital before doing many other things, as first-level buildings have a significantly higher return of investment than upgrades do. At this stage, the archetypes matter less, but this phase is also over somewhat quickly.

我相信大家都同意，需要發(fā)展的最重要的地產(chǎn)，是你的首都。在做許多其它事情之前，AI會(huì)相當(dāng)努力地填補(bǔ)其首都中的所有空位，因?yàn)橐患?jí)建筑的投資回報(bào)率明顯高于升級(jí)建筑。在這個(gè)階段，個(gè)性原型不太重要，但這個(gè)階段也會(huì)很快結(jié)束。

This has a quite significant effect on the AI overall. By being able to kickstart their economy earlier than before, you will see them doing much better in the mid-late game. In the current version, you see a lot of provinces with open building slots well into the late game: even places of prominence, such as Paris, can sometimes be devoid of buildings - after these changes, this will be exceedingly rare.

這對(duì)AI整體的影響相當(dāng)大。由于能夠比以前更早地啟動(dòng)他們的經(jīng)濟(jì)，你會(huì)看到他們?cè)谟螒蛑泻笃诒憩F(xiàn)得更好。在目前的版本中，你會(huì)看到很多省份在游戲后期仍有空置的建筑槽位，即使是像巴黎這樣的重要地方，有時(shí)也會(huì)沒(méi)有建筑。在這次改動(dòng)后，這種情況將是非常罕見(jiàn)的。

2. Early Domain Development

2、早期直轄領(lǐng)地發(fā)展

In the second stage, the AI wants to develop its secondary counties by constructing first-level buildings there too. Truth be told, many AI's elect to buy MaA or keep developing their capital during this phase as the choice of what to do/construct is semi-random and weighted - but this really just results in more powerful rulers overall.

在第二階段，AI希望發(fā)展他們首都以外的次級(jí)伯爵領(lǐng)地，在那里也建造一級(jí)建筑。說(shuō)實(shí)話，許多AI也會(huì)在這個(gè)階段選擇購(gòu)買(mǎi)兵士，或繼續(xù)發(fā)展他們的首都，因?yàn)檫x擇做什么/建造什么是半隨機(jī)和加權(quán)的——但這實(shí)際上都會(huì)促成整體更強(qiáng)大的AI統(tǒng)治者。

When first-level buildings are present in an AI's entire domain, they move on to the Later Economic Behavior.
當(dāng)一級(jí)建筑出現(xiàn)在AI的所有直轄領(lǐng)地時(shí)，他們就會(huì)進(jìn)入后期經(jīng)濟(jì)行為。ious AI''s saving up buffers, etc.

在這個(gè)階段，個(gè)性原型正在顯現(xiàn)，好戰(zhàn)的AI在積攢戰(zhàn)爭(zhēng)資金（盡管預(yù)留可供開(kāi)銷的時(shí)間比后期階段的時(shí)間短?。?jǐn)慎的AI在積攢緩沖儲(chǔ)備資金，等等。

When first-level buildings are present in an AI’s entire domain, they move on to the Later Economic Behavior.

當(dāng)一級(jí)建筑出現(xiàn)在AI的所有直轄領(lǐng)地時(shí)，他們就會(huì)進(jìn)入后期經(jīng)濟(jì)行為。

3. Later Economic Behavior

3、后期經(jīng)濟(jì)行為

During this stage, the archetypes will manifest fully. Builder AI's will constantly build up their realm. Cautious AI's will want a buffer, allies, and a powerful military, Warlike AI's will only invest their gold if they've been at peace for many, many years - and so on.

在這個(gè)階段，個(gè)性原型將充分表現(xiàn)出來(lái)。建設(shè)AI將不斷建設(shè)他們的領(lǐng)國(guó)。謹(jǐn)慎AI會(huì)想要一個(gè)緩沖儲(chǔ)備金、許多盟友和一支強(qiáng)大的軍隊(duì)，好戰(zhàn)AI只有在他們已經(jīng)和平了很多很多年之后才會(huì)投資他們的金錢(qián)——以及其它諸多AI個(gè)性表現(xiàn)。

Interestingly enough, despite the AI being disincentivized from declaring frequent wars (previously all AIs were constantly gearing up for war), there are no fewer wars in the world. The wars that are declared are, however, more intense and exciting, as the AIs are generally more built up and have a stronger economic backbone.

相當(dāng)有趣的是，盡管AI被設(shè)置為不能經(jīng)常宣戰(zhàn)（以前所有的AI都在持之以恒地為戰(zhàn)爭(zhēng)做準(zhǔn)備），但世界上的戰(zhàn)爭(zhēng)并沒(méi)有減少。被宣布的戰(zhàn)爭(zhēng)卻更加激烈和激動(dòng)人心，因?yàn)锳I通常更加強(qiáng)力，并且有更強(qiáng)大的經(jīng)濟(jì)支柱。

While the Economy part of this update isn't the only thing contributing to what you're about to see, here's a comparison of the Live version of the game compared to the upcoming update. The sample size is 200 years, and there are no major deviations except for the Mongol Empire not having collapsed yet in the Live comparison.

雖然你看到的經(jīng)濟(jì)改動(dòng)并不是這次更新的全部改動(dòng)，但這里有一份當(dāng)前游戲版本和即將到來(lái)的更新版本的比較圖。樣本大小為200年，這里面也沒(méi)有什么大的偏差，除了在現(xiàn)有的比較中，蒙古帝國(guó)仍然沒(méi)有崩潰。

[Image - Various Comparisons between the Live Version (1.6.1.2) and the upcoming update (1.7.0)]

[圖例：許多當(dāng)前版本（1.6.1.2）和即將到來(lái)的更新版本（1.7.0）的比較]

Counts+ in this context means Counts, Dukes, Kings, and Emperors.
這里的Counts+表示男爵，公爵，國(guó)王和皇帝。

Diplomatic Matters

外交事務(wù)

One of the AI's faults is that it underuses the tools given to it, choosing to use them only sporadically and without much direction. The AI acting out its personality is all well and good, but in many instances, personality didn't even matter at all, for example when choosing to revoke a title. We've done many small things that together add up to a vastly improved experience.

AI的一個(gè)毛病是，它沒(méi)有充分利用給它的工具，只是零星地選擇使用它們，并且沒(méi)有明確的目的。AI表演它個(gè)性的功能和成果是好的，但在許多情況下，個(gè)性甚至根本不起作用，例如，在選擇剝奪頭銜時(shí)。我們做了許多小事，這些小事加在一起，使我們的體驗(yàn)得到了極大的提升。

Domain Consolidation

整合直轄領(lǐng)

The AI is now going to strive towards having a strong capital and domain, prioritizing first and foremost the consolidation of lands within their De Jure Capital Duchy. For example, the king of France will want to hold Isle-de-France and all counties within the duchy of Valois.

現(xiàn)在，AI將力爭(zhēng)擁有一個(gè)強(qiáng)大的首都和直轄領(lǐng)，最優(yōu)先最重要的是，鞏固其法理首都公爵領(lǐng)內(nèi)的土地。例如，法國(guó)國(guó)王將希望擁有法蘭西島和瓦盧瓦公爵領(lǐng)內(nèi)的所有伯爵領(lǐng)。

In order to do this, they are now willing to take on a modicum of Tyranny via revoking or retracting, modified by their personality. For reference, the AI has never before willingly taken on Tyranny and has always waited for a revocation reason. All AIs will want their capital county and at least a small domain (3 or so), regardless of personality, but Just and Generous AIs will stop after that point. Depending on the AI's rationality, they will be willing to take on more or less Tyranny. Wise and rational rulers will want to let their Tyranny decay completely before taking a tyrannical action again, while more irrational rulers are willing to keep their Tyranny high in order to consolidate their domain.

為了做到這一點(diǎn)，他們現(xiàn)在愿意通過(guò)承擔(dān)少量暴政值來(lái)剝奪頭銜或者回收領(lǐng)地，這會(huì)根據(jù)他們的個(gè)性有所修正。作為參考，AI以前從未愿意承擔(dān)暴政值，并且一直在等待一個(gè)回收領(lǐng)地的理由。無(wú)論是什么個(gè)性，所有的AI都會(huì)想要他們的首府伯爵領(lǐng)和至少一片小小的直轄領(lǐng)（3個(gè)或3個(gè)左右），但公正和慷慨的AI會(huì)在達(dá)成這點(diǎn)后停止。根據(jù)AI的理性程度來(lái)判斷他們將愿意承擔(dān)更多或者更少的暴政值。明智和理性的統(tǒng)治者將會(huì)在再次采取暴政行動(dòng)前讓他們的暴政值完全衰減，而更多非理性的統(tǒng)治者則愿意讓他們的暴政值保持較高水平，以此來(lái)鞏固他們的領(lǐng)地。

In order to consider revocation, the AI looks to its reserves of gold and military strength, making sure to not get into Tyranny wars it doesn't think it can win (of course, it still happens that a ruler is overthrown via Tyranny, but it's rarer!).

AI為了考慮回收領(lǐng)地，會(huì)審視他的金錢(qián)儲(chǔ)備和軍事力量，確保不陷入他自己認(rèn)為無(wú)法勝利的暴政戰(zhàn)爭(zhēng)（當(dāng)然，仍然有統(tǒng)治者通過(guò)暴政被推翻的情況，但這更少見(jiàn)?。?/p>

Of course, relationships are accounted for, so you won't see the AI revoking from their friends, lovers, children, etc.

當(dāng)然了，人際關(guān)系也是會(huì)被計(jì)算在內(nèi)，因此你不會(huì)見(jiàn)到AI剝奪他們的朋友們、愛(ài)人們或孩子們等關(guān)系角色的領(lǐng)地。

In practice, this means that the average AI ruler is much more powerful, and you won’t see things such as the king of France being content with having one single domain county in Brittany, just because they don’t have a revoke reason on the current ruler of Isle-de-France. In combination with the previously mentioned economic changes, this results in economically powerful AI rulers.

實(shí)際上，這意味著平均水平的AI統(tǒng)治者會(huì)更強(qiáng)力，你不會(huì)看到諸如明明是法國(guó)國(guó)王卻滿足于在布列塔尼擁有一個(gè)單一的伯爵領(lǐng)，僅僅只是因?yàn)樗麄儗?duì)法蘭西島現(xiàn)任統(tǒng)治者沒(méi)有回收領(lǐng)地理由的這類事情。結(jié)合前面提到的經(jīng)濟(jì)修改，這就產(chǎn)生了經(jīng)濟(jì)上強(qiáng)力的AI統(tǒng)治者。

[Image - Alba having consolidated their capital duchy]

[圖例：阿爾巴王國(guó)已經(jīng)整合了他們的首都公爵領(lǐng)]

The AI has also been taught to revoke secondary baronies in the counties it owns, which is especially useful for Clan rulers, as they can hold temples. This makes it so that much of the Clan sphere has a stronger economy, especially early on, and is able to match European and Indian feudal realms which generally have more favorable terrain.

AI也被教導(dǎo)會(huì)剝奪它擁有的伯爵領(lǐng)內(nèi)次級(jí)男爵領(lǐng)頭銜，這對(duì)氏族制統(tǒng)治者會(huì)特別有用，因?yàn)樗麄兛梢猿钟兴聫R。這使得大部分氏族制統(tǒng)治階級(jí)擁有更強(qiáng)力的經(jīng)濟(jì)——特別是在早期，并使他們能夠與歐洲和印度的封建王國(guó)相媲美（通常擁有更有利的地形）。

Vassalization

附庸化

One of the things that we've changed a lot is how vassalization works and how the AI uses it. This isn't a pure AI change, as we've also rebalanced all the modifiers that affect whether a character wishes to accept your overlordship or not. But, yes, simultaneously we’ve taught the AI to consider these facts and try to make their neighbors into vassals.

我們改變了很多東西，其中之一是附庸化的運(yùn)作方式以及AI如何使用它。這不是一個(gè)單純的AI改動(dòng)，因?yàn)槲覀円仓匦缕胶饬怂械男拚?xiàng)，那些會(huì)影響一個(gè)角色愿意或者拒絕接受你統(tǒng)治的所有修正項(xiàng)。但是，是的，我們已經(jīng)教會(huì)AI考慮這些事實(shí)，同時(shí)教會(huì)AI去試著把他們的鄰居變成附庸。

A sub-goal here was to make vassalization into a viable alternative to military conquest, while also dialing back the most powerful sources of vassal acceptance and adding more interactive modifiers. While I won't go into all of the detail here, I'll say that the modifiers from Diplomatic Court and True Ruler have been lowered, while modifiers such as Average Powerful Vassal Opinion (-20/+20) have been added to promote alternatives to only being militarily powerful. We've also tweaked the opinion modifiers for Clan rulers so that Feudal rulers don't have a fundamental advantage when it comes to vassalization (by offering low Contracts).

這里的一個(gè)子目標(biāo)是使附庸成為一個(gè)可行的替代軍事征服的選項(xiàng),同時(shí)回調(diào)影響附庸接受度的最強(qiáng)有力的來(lái)源，并且添加更多的交互修正。雖熱我不會(huì)在這里討論所有的細(xì)節(jié)，但我想告訴你們，來(lái)自外交宮廷和真正統(tǒng)治者的修正已經(jīng)降低，而像常規(guī)強(qiáng)力附庸好感度（-20/+20）已經(jīng)被添加，用來(lái)提供軍事強(qiáng)力以外的替代選項(xiàng)。我們還調(diào)整了氏族統(tǒng)治者的好感度修正，使得封建統(tǒng)治者在附庸方面沒(méi)有根本的優(yōu)勢(shì)（通過(guò)提供要求更低的契約）。

The AI will now much more frequently check if neighbors would accept vassalization, and they will actively seek to sway or befriend neighbors if the AI considers them targets for diplomatic expansion. This means that the player does not have a monopoly on vassalizations any longer; if you take too long, AI rulers will sweep in and offer overlordship to your small independent neighbors.

AI現(xiàn)在會(huì)更頻繁地確認(rèn)鄰國(guó)是否會(huì)接受封庸，如果AI認(rèn)為鄰國(guó)是外交擴(kuò)張的目標(biāo)，他們將積極尋求游說(shuō)或結(jié)交鄰國(guó)。這意味著附庸國(guó)家并不再被玩家所壟斷：如果你花費(fèi)的時(shí)間過(guò)長(zhǎng)，AI統(tǒng)治者可能會(huì)介入，并且向你獨(dú)立弱小的鄰居提供領(lǐng)主庇護(hù)。

In practice, this means that areas stay shattered and stagnant for less time, while also creating more interesting realms where different cultures (and even faiths) are represented in the vassal strata. As an example, you might see a great unifying Sultan appear somewhere in Persia and diplomatically sweep up the remains of a shattered Seljuk Empire.

在實(shí)際游戲中，這意味著地區(qū)保持破碎和停滯的時(shí)間會(huì)更短，同時(shí)也會(huì)創(chuàng)造更有趣的國(guó)度，其中包含不同的文化（甚至信仰）的封臣。舉個(gè)例子，你可能會(huì)看到一個(gè)巨大統(tǒng)一的蘇丹國(guó)出現(xiàn)在波斯區(qū)域，并且通過(guò)外交手段整合了過(guò)去的塞爾柱帝國(guó)的疆域。

[Image - Example of new vassalization modifiers]

[圖例 – 新的附庸化修正項(xiàng)]

總值：+2
基礎(chǔ)不情愿值：-50
你不是他真正的法理領(lǐng)主：-20
軍事力量差距：+13（最大20）
你的強(qiáng)力附庸評(píng)價(jià)：+12（最大20）
他屬于你的王國(guó)：+5
你會(huì)說(shuō)伊朗語(yǔ)：+5
偏執(zhí)：-20
埃米爾穆?tīng)柨撕箤?duì)你的好感度：+57

Holy Wars

圣戰(zhàn)

As it turns out, rulers of the same faith as the defender in a Holy War would never join in their defense, despite the game claiming they would. This is our bad: a set of triggers were looking at the wrong character, and thus no one would consider joining. This is now changed, and we've put some effort into ensuring that the rulers who join up against you feel right.

事實(shí)證明，與圣戰(zhàn)中的捍衛(wèi)者有同樣信仰的統(tǒng)治者永遠(yuǎn)不會(huì)加入他們的防御，盡管游戲聲稱他們會(huì)加入。這是我們的問(wèn)題：一連串觸發(fā)器指向了錯(cuò)誤的角色，因此沒(méi)有人會(huì)考慮加入?，F(xiàn)在這種情況改變了，我們做出了一些努力，來(lái)確保那些反對(duì)你的統(tǒng)治者也會(huì)正確地加入圣戰(zhàn)。

Protecting the boundaries of your faith is important, and the AI is now aware of this. Depending on personality and economic readiness, they will now assist their brothers-in-faith against encroaching heathens. Zealous and bold AI's are the first to join, while you can be certain that cynical and cowardly rulers will not. Vassals of your target won't join in, though vassals of neighboring rulers might - generally, you need only look within the immediate area as you declare wars. Extremely zealous rulers, though, might decide to join in from further away.

保護(hù)信仰的邊界很重要，AI現(xiàn)在意識(shí)到了這一點(diǎn)。根據(jù)個(gè)性和經(jīng)濟(jì)準(zhǔn)備程度的不同，他們現(xiàn)在將幫助他們的兄弟們抵御異教徒的入侵。熱心和大膽的AI會(huì)是最先加入的，你也可以確定憤世嫉俗和懦弱的統(tǒng)治者不會(huì)這樣做。你的目標(biāo)的附庸不會(huì)加入，盡管鄰近統(tǒng)治者的封臣可能會(huì)——一般來(lái)說(shuō)，你只需要在你宣戰(zhàn)時(shí)查看鄰近的地區(qū)。然而，極端熱心的統(tǒng)治者可能會(huì)決定從更遠(yuǎn)的地方加入進(jìn)來(lái)。

In practice, you have to prepare well before declaring a holy war, but taking stock of your neighbors will help you here - personalities are telling, so you shouldn't be overly surprised when someone aids your target.

實(shí)際上，你必須在宣布圣戰(zhàn)之前做好準(zhǔn)備，評(píng)估你的鄰居是否會(huì)幫助你——他的個(gè)性會(huì)告訴你，所以當(dāng)有人幫助你完成戰(zhàn)爭(zhēng)目標(biāo)時(shí)，你不應(yīng)該感到過(guò)于驚訝。

From a game perspective, this makes it much less likely for the Middle East to be completely overtaken by either Catholics or Orthodox Christians (which seemed to happen in at least 90% of all observer games that ran until the end date). Major religious upsets are now rarer, but when they happen they are all that much more likely to stick, as they've probably been launched by strong rulers during an opportune time.

從游戲的角度來(lái)看，這使得中東不太可能被天主教徒或東正教徒完全取代（在至少90%的持續(xù)到游戲結(jié)束的觀察者游戲中似乎都發(fā)生了這種情況）。主要宗教的混亂現(xiàn)在更少了，但當(dāng)它們發(fā)生時(shí)，它們就更有可能持續(xù)下去，因?yàn)樗鼈兛赡苁怯蓮?qiáng)大的統(tǒng)治者在適當(dāng)?shù)臅r(shí)機(jī)發(fā)起的。

Realm Stability

領(lǐng)國(guó)穩(wěn)定性

AI rulers should now be somewhat more aware of the state of their realm, and proactively try to improve its stability. There are many flavors of stability, some more underhanded than others…

AI統(tǒng)治者現(xiàn)在應(yīng)該更加了解自己的領(lǐng)國(guó)狀況，并且主動(dòng)努力提高其穩(wěn)定性。這里會(huì)有更多的穩(wěn)定度偏好，有的比其他的更為隱蔽……

Firstly, the AI is now much more likely to sway, and when they choose who to sway they are much more likely to pick important characters. For example, they will sway their realm priest, spymaster, and angry powerful vassals much more often, and also keep swaying them for a longer time. This also goes for the befriend scheme, should it be available. In the comparison screenshot at the end of the economy section, you'll note that there are significantly more AIs with bishops/realm priests who endorse them.

首先，AI現(xiàn)在更有可能游說(shuō)他人，當(dāng)他們猶豫選擇誰(shuí)時(shí)，他們就更有可能選擇對(duì)他們更為重要的角色。例如，他們會(huì)更頻繁地想要游說(shuō)主教、間諜主管和憤怒的強(qiáng)力封臣，還會(huì)繼續(xù)更長(zhǎng)時(shí)間地游說(shuō)他們。對(duì)于交友的計(jì)謀也是這樣，只要交友計(jì)謀可用的話。在經(jīng)濟(jì)部分末尾的比較截圖中，你會(huì)注意到有更多的主教/祭司支持他們。

AI rulers with a more… unsavory disposition will use the Find Secrets council task, which the AI previously didn't use. They'll tend to target their vassals or rivals, if the chance of being found out isn't too great. They'll then blackmail and use the ensuing hooks to improve vassal contracts. Rulers who have unlocked the Fabricate Hook scheme will also much more often use it on vassals whose contracts can be modified. Over the generations, this means that the contracts AI rulers have with their vassals tend to be much better than they used to.

有更多……令人討厭的性格的AI統(tǒng)治者，將使用尋找秘密的內(nèi)閣任務(wù)，這是AI之前沒(méi)有使用過(guò)的。如果被發(fā)現(xiàn)的機(jī)會(huì)不是太大，他們就會(huì)傾向于將他們的封臣或?qū)κ侄槟繕?biāo)。然后他們會(huì)勒索，并利用隨后的牽制來(lái)改進(jìn)附庸契約。解鎖了制造牽制的統(tǒng)治者也會(huì)更經(jīng)常地使用那些牽制來(lái)修改附庸契約。在過(guò)去幾代之后，AI統(tǒng)治者和他們的封臣之間簽訂的契約條件要好得多。

[Image - Sample vassal contracts, France 1389]

[圖例 - 契約示例，1389年的法蘭西]

In general, we've taken a pass on when AIs will use certain Council Tasks. They will avoid certain things completely (using bestow royal favor on vassals-of-vassals), and more proactively switch to tasks that support them in the short term (support schemes when any hostile scheme is active, Organize Army when gold is running very low, etc).

總體而言，我們已經(jīng)放手讓AI們決定它們會(huì)在何時(shí)使用某個(gè)特定內(nèi)閣任務(wù)。它們將會(huì)完全避開(kāi)某些特定事項(xiàng)（對(duì)封臣的封臣使用賜予帝王恩惠），并會(huì)更加主動(dòng)地采用對(duì)能在短期內(nèi)幫助它們的任務(wù)（在有陰謀進(jìn)行時(shí)選用支持計(jì)謀，在沒(méi)錢(qián)時(shí)選用組織軍隊(duì), 諸如此類）。

There's a whole lot more that we've taught the AI to do, most of them small - but small things add up. They will send non-primary children to be educated by vassals they want to appease (it's 15 opinion per child after all!), make their children learn the languages of their subjects to prepare them for a smooth succession, hire seneschals if they have more than one county with very low control, etc, etc. It really does add up.

我們還教會(huì)了AI做一大堆其他的事情，當(dāng)中大部分都微不足道，但積少成多。AI會(huì)將非主要子女送到想討好的封臣那去接受教育（畢竟一個(gè)小孩可以換整整15好感?。?，讓它們的子女學(xué)習(xí)封臣的語(yǔ)言以讓他們繼位時(shí)更加順利，在多于一個(gè)伯爵領(lǐng)控制力極低時(shí)聘請(qǐng)總管等等等等。真的是積少成多。

All of these changes lead to AI realms being more stable and better at recovering from bad wars or messy successions. Of course, there's still a lot of the hallmark CK3 chaos that we all know and love - it's just not all there is, now!

上述的所有改動(dòng)都讓AI的領(lǐng)國(guó)變得更加穩(wěn)定，并可以更快地從戰(zhàn)爭(zhēng)泥潭或者混亂繼承中恢復(fù)過(guò)來(lái)。當(dāng)然，游戲里仍然有不少我們熟知且熱愛(ài)的CK3標(biāo)志性混亂——現(xiàn)在的還不是全部！

Let's compare two areas, one from the Live version, and one from the upcoming update. They are 200 years from game start (in fact, these are the same saves used for comparison stats earlier in the DD).

讓我們比較下兩個(gè)地區(qū)吧，一個(gè)是來(lái)自現(xiàn)行版本的，而另一個(gè)則是來(lái)自即將到來(lái)的更新版本的。兩者都是游戲開(kāi)局之后200年的情況（事實(shí)上，它們都是上文中用于比較數(shù)據(jù)的同一局游戲）。

[Picture - Comparison between an area in Live and the Update]
[圖例 - 在現(xiàn)行版本與更新版本中對(duì)一個(gè)地區(qū)的比較]

Do note that these changes do not mean you'll see the same realms in every game. It's still as it's always been, where you might have a powerful Burgundy in the south or a Wales controlling the British Isles; it's just that, now, the emerging Burgundian and Welsh realms will be much stronger and able to thrive! Additionally, given the AI is now consolidating a larger, more centralized domain, partition splits are nicer and border gore is reduced.

請(qǐng)注意，這些修改并不表示你每局都會(huì)看到一模一樣的版圖。一切還是一如往常，你仍可能會(huì)在南方看到一個(gè)強(qiáng)大的勃艮地或者一個(gè)控制著英倫三島的威爾士；只不過(guò)現(xiàn)在，新生的勃艮地和威爾士王國(guó)將會(huì)變得更強(qiáng)并且繁榮昌盛！除此以外，由于AI現(xiàn)在可以牢固地控制更大更加集中的直轄領(lǐng)，因?yàn)榕上刀至训那闆r變得更為理想了，同時(shí)出現(xiàn)陰間版圖的情況也減少了。

Ikh Mongol Uls

大蒙古國(guó)

No matter what changes we've made to the Mongols in the past, they always fall on one single point: when they grow large enough, it takes them too long to march their troops to their target, and thus lose too many troops to attrition and too much warscore to time.

不論過(guò)去我們對(duì)蒙古人做了什么改動(dòng)，他們總會(huì)在到達(dá)一個(gè)點(diǎn)上衰落：當(dāng)他們擴(kuò)大到一定程度時(shí)，他們把太多的時(shí)間花在了行軍到目標(biāo)上，并因此損耗了太多部隊(duì)，以及因時(shí)間而失去戰(zhàn)爭(zhēng)分?jǐn)?shù)。

We've taught the AI to raise their troops closer to the target of their wars. They will seek to raise in a safe county within their realm that is close, but not bordering, to the wargoal. While this is the most impactful for the Mongols, all Duke or above rulers will now do this. What this means for the Mongols: they will no longer march all the way from far eastern Mongolia after every conquest. This change alone magnifies their potential for conquest tenfold (at least). They now fulfill their purpose; they shake up the world and reshape the lands they conquer. More than ever, you must now seriously consider if you wish to submit to the Great Khan…

我們教會(huì)了AI在離他們的戰(zhàn)爭(zhēng)目標(biāo)比較近的地方集結(jié)部隊(duì)。他們將會(huì)在自己的領(lǐng)地里找一個(gè)既相近又不與戰(zhàn)爭(zhēng)目標(biāo)與接壤的安全伯爵領(lǐng)集結(jié)部隊(duì)。雖然此改動(dòng)對(duì)蒙古人的影響最大，不過(guò)現(xiàn)在所有公爵級(jí)以上的統(tǒng)治者也會(huì)這么做了。這對(duì)蒙古人意味著：他們不再需要在每次征戰(zhàn)以后都從蒙古東部行軍過(guò)來(lái)了。而僅僅這一單項(xiàng)改動(dòng)就把蒙古人的征服潛力放大了十倍（至少）。他們現(xiàn)在可以實(shí)現(xiàn)他們目的了；撼動(dòng)世界并重塑他們所征服的土地。而現(xiàn)在你必須比以往任何時(shí)候都更加認(rèn)真地考慮，你是否愿意臣服于大汗了……

In addition to the new raising behavior, we have improved the way the Mongols pick their targets (making them less likely to gallop into the mountains of Tibet or immediately crash against a massively strong Byzantine empire), and we've changed how successor khanates work. The Golden Horde, Ilkhanate, Chagatai, etc, were never long for this world after they spawned. There were many reasons for this: they didn't get a proper domain, no de jure lands, no gold, etc. While they still dissolve with some regularity (as they should) they now sometimes persist, and even thrive, as we've improved their starting positions by assigning them some starting resources, domain, and De Jure land depending on what they’ve managed to conquer.

除了新的集結(jié)行為，我們也改善了蒙古人選擇目標(biāo)的方式（讓他們不要經(jīng)常再在西藏的山里騎馬飛馳或者光速與無(wú)比強(qiáng)大的拜占庭帝國(guó)開(kāi)戰(zhàn)），同時(shí)我們也修改了繼承者汗國(guó)的運(yùn)作方式。金金金金帳汁國(guó)，伊兒汗國(guó)，察合臺(tái)汗國(guó)等等在它們生成后都不曾在這世界里長(zhǎng)存。當(dāng)中有許多原因：沒(méi)有理想的領(lǐng)土，沒(méi)有法理領(lǐng)土，沒(méi)錢(qián)等等。雖然他們有時(shí)仍會(huì)定期滅亡（就如同他們應(yīng)該那樣），現(xiàn)在他們有時(shí)會(huì)繼續(xù)存在，甚至變得繁榮昌盛，因?yàn)槲覀兏鶕?jù)他們成功征服的東西為它們分配起始資源、領(lǐng)地及法理領(lǐng)土以改善他們的開(kāi)局。

While it isn’t overly common that the Mongols exceed their historical conquests, here's an example of how far they can spread under somewhat-ideal circumstances (long-lived Khans).

雖然蒙古人超出其歷史征服范圍的情況并不常見(jiàn)，但有一個(gè)他們可以在理想的情況下（長(zhǎng)壽的大汗）可以擴(kuò)張多遠(yuǎn)的例子。

[Image - Initial state of the world]
[圖例 – 世界開(kāi)局狀態(tài)]

[Image - Some years in: note that Tibet was taken because ü had land in mongolia]
[圖例 - 幾年后：注意，吐蕃已經(jīng)被拿下了。因?yàn)樾l(wèi)在蒙古里擁有領(lǐng)地。]

[Image - The newly-shattered Persia gets absorbed]
[圖例 – 剛剛碎成一地的波斯被吞并了。]

[Image - Strategically avoiding a Byzantium that is almost their strength, the Mongols go towards eastern europe]
[圖例 – 戰(zhàn)術(shù)性地避開(kāi)一個(gè)力量與自身不相伯仲的拜占庭，蒙古人們正向東歐前進(jìn)進(jìn)進(jìn)進(jìn)進(jìn)。]

[Image - After taking all of eastern Europe, they turn to Arabia and the Fatimids]
[圖例 – 在拿下整個(gè)東歐后，他們轉(zhuǎn)向了阿拉伯及法蒂瑪。]

[Image - Eventually, Hungary falls after an intense war]
[圖例 - 最終，匈牙利在一場(chǎng)激戰(zhàn)后淪陷了。]

[Image - The Khan dies and successor khanates spawn in. Who knows how long they’ll last?]
[圖例 – 大汗去世了，而繼承者汗國(guó)登場(chǎng)了。不知道他們能存續(xù)多久？]

That's it for now. The full update notes with all the details will come soon, so stay tuned!

這就是現(xiàn)在有的內(nèi)容了。完整的補(bǔ)丁附記和所有細(xì)節(jié)都會(huì)在不久后發(fā)布，所以保持關(guān)注啦！

翻譯：一個(gè)幽靈 雷敏Duke 倫敦 Arcanunu 視力不好的胡德

校對(duì)：三等文官猹中堂

歡迎關(guān)注UP主和主播小牧Phenix！

歡迎關(guān)注牧游社微信公眾號(hào)和知乎專欄！微信公眾號(hào)改版為信息流，歡迎【置頂訂閱】不迷路，即時(shí)獲得推送消息！

B站在關(guān)注分組中設(shè)置為【特別關(guān)注】，將會(huì)在私信內(nèi)及時(shí)收到視頻和專欄投稿的推送！

歡迎加入牧有漢化，致力于為玩家社群提供優(yōu)質(zhì)內(nèi)容！組員急切募集中！測(cè)試群組822400145!?

本作品英文原文著作權(quán)屬Paradox interactive AB所有，中文譯文著作權(quán)屬牧有漢化所有。?