欧陆风云4开发日志 | 9/28 平衡性改动与AI改进

牧游社 牧有汉化翻译

Europa Universalis IV - Development Diary 28th September 2021

Johan, Studio Manager Paradox Tinto

Hello everyone and welcome to a rather meaty development diary for Europa Universalis IV. This week we'll release 1.31.6, which is a compatibility patch for the launch on Epic Store, where EU4 will be freely available for a short period of time.

大家吼，欢迎来到这篇非常“丰满”的欧陆风云4日志。这周我们会为在Epic商店发行而发布1.31.6版本号兼容补丁。玩家短时间内可以在该平台免费领取EU4。

This patch does not contain any bug fixes, improvements or balance changes, as we had to make sure we had a stable version for this release. Some of our programmers and qa have been working diligently to create this new version for a long while now.

这个补丁不包括任何bug修复、优化或者平衡性改变，因为我们必须确保为本次发布提供一个稳定的版本。一些我们的程序员和QA已经为这个新版本勤奋工作了很久。

Meanwhile the core-team have been working on the major 1.32 patch which we aim to have out later this year. While the other development diaries will focus mostly on what content we will add to the immersion pack, today we will talk about a lot of balance changes!

与此同时，核心团队还在为我们计划在今年晚些时候发布的关键补丁1.32版本号而奋斗。其它开发日志主要关注我们即将在沉浸包中加入的内容，而我们今天则要来谈谈平衡性方面的改动！

Concentrate Development

集中发展度

As we all know, the Concentrate Development feature, while technically working as designed, has a few drawbacks, as it can become very unbalanced and immersion breaking. For 1.32 we have been reworking this.

如我们所知，集中发展度的特性即便严格按照设计时想法执行，也还是存在一些问题的，因为它会变得非常不平衡并会破坏代入感。我们在1.32版本号中重置了该功能。

Now when you pillage or concentrate a set of provinces, the horde "raze" function is called, but with 20% reduction of the development. This is then converted into monarch points in the source province(s), using the cost it would improve development in each of these provinces.

现在当你在一些省份使用劫掠或者集中发展度功能的时候，会出现游牧部落的“焚掠”作用，但是减少20%的发展度。然后被劫掠省份失去的发展度会按照此地提升发展度的花费来换算为君主点数。

These "monarch points" are then taken to the capital of your country, with a loss of 20% of points when applicable, as some countries have free concentrate development. What is left is then used to "buy" development in your capital province. Any "monarch points" left over are then lost.

这些“君主点数”接下来被带到本国首都，且会在符合条件的情况下损失20%的点数，因为有一些国家有免费的集中发展度。随后剩下的点数可以用来在你的首都省份“买”发展度。任何没有用的“君主点数”就会消失。

A shout out to the post who inspired this new design!

分享一下启发了这个新设计的帖子！

https://forum.paradoxplaza.com/forum/threads/concentrate-development-is-not-for-a-tall-play-addressing-elephant-in-the-room.1473575/#post-27522944

We also fixed the way you could concentrate development twice, by first owning it, and then from the released vassal. Liberty desire now also increases for the subjects when you concentrate their development.

我们还修复了你可以两次集中发展度的方法。该方法是先拥有省份再从放出的傀儡那里集中发展度。现在当你从附庸那里集中发展度的时候，独立倾向也增加了。

Misc Balance Changes

杂项平衡改变

The catholic curia powers are gaining some improvements in 1.32 as well, with all getting another bonus and here are some examples.

天主教的教廷力量在1.32版本号也得到了一些提升，所有能力都获得了额外加成。这里是一些例子：

Bless Ruler - Now also gives +10% Land Morale
为统治者赐福 – 现在另外获得10%陆军士气。

Indulgence for Sin - Now gives +10% Improve Relations
赎罪券 – 现在获得10%改善关系。

Send Papal Legate- Also adds 10% cheaper annexation costs.
派遣教廷使节 – 额外减少10%的吞并花费。

We made some tweaks to curry favors, so that it now relies far more on relative military power, so it's harder to gain for a smaller, weaker nation.

我们对获取人情做了点调整，因此该功能现在对相对军事力量更看重了，所以小国弱国更难获得人情了。

When it comes to natives, we added a new reform to slow down the path to becoming a duchy for settled tribes, as well as adding feudalism as a requirement.

在土著方面，我们加入了新的改革，让土著达到可以成为定居部落的公国等级的过程更慢了。拥有封建主义思潮也成为了其中的要求。

We also added in cooldowns for migration into non-owned lands, made costs for adding tribal land that scales with owned provinces and development. There are also penalties for migration into winter territory. Also added a relation penalty that increases each month that a tribe is inside tribal land of another country.

我们还延长了向无主之地迁徙的冷却期，并根据拥有省份和发展度衡量添加部落土地的花费。迁徙至入冬的领土也会有惩罚。我们另外为迁入他国部落土地的部落增设了每个月增加的关系惩罚。

And of course we are rebalancing, i.e. nerfing, the native development growth for 1.32, as it is completely unrealistic and does not work with the rest of the game.

当然我们也在1.32版本号平(xue)衡(ruo)了土著的发展度增长，因为这完全不现实而且与游戏其他内容不搭配。

We also changed when the AI released vassals for being over the governing capacity to be a much higher threshold, and also giving lucky nations a nice little boost to their governing capacity.

我们还提高了AI在超过行政容量时放出附庸的门槛，幸运国也得到了一点不错的行政容量加成。

Now I hand over the keys to @Gnivom who will talk about some changes to Institutions and the AI for 1.32

我现在把话筒交给讲解1.32版本号中思潮与AI改动的@Gnivom。

Institution Changes

思潮改动

We've decided to change how Institutions affect tech cost.

我们打算改变思潮影响科技花费的方式。

1. So what was the problem?

1、所以问题是什么？

We all know how Institutions work:

我们都知道思潮是怎么运作的：

文艺复兴！
旅行者提到了在遥远的土地上正在重生的古代思想，我们很难严肃对待这些传闻，但是，根据报告，所谓的「文艺复兴」思想家们的确发掘出了一些巨大的秘密。
- 这改变了一切……
没有完全接纳文艺复兴思潮的国家将会每年获得1%的科技花费惩罚，至多50%。

Hard to take seriously indeed

的确很难把这个当真。

Countries outside of Europe, and especially in the Americas and sub-Saharan Africa, have to pay more for a technology the longer they wait. This means they’re in a hectic struggle against time, as those who fall behind their neighbors will struggle to catch up.

欧洲之外的国家，尤其是美洲国家和撒哈拉以南非洲的国家，等得越久就要在科技上花费越多。这意味着它们要忙于与时间赛跑，而那些落伍于邻国的国家要艰难地追赶。

2. The New System

2、新系统

From 1.32, every technology has a fixed set of required institutions. If you haven’t yet embraced one or more required institutions, that technology costs you 50% more for each.

自1.32版本号起，每项科技都存在固定的所需思潮。若你未接纳一项及以上的所需思潮，每项未接纳思潮将额外增加科技花费50%。

This updated Technology View should hopefully be self explanatory:

更新后的科技界面有望解释该项变动。

当前的思潮、邻国加成、领先于时代对该项科技花费的惩罚或加成。

未接纳文艺复兴：+50%
领先于时代：+10%（+1年）

These are some nice properties of the new system over the old:

相较于旧系统，新系统有以下优势：

Institutions now cause a rubber-band effect on tech instead of a snowball effect
思潮现在对科技产生的是皮筋效应，而不是雪球效应。（突然暴毙，而不是慢性死亡）。

Some movement in Florence doesn’t magically affect the Aztecs (until it actually spreads there)
佛罗伦萨的运动不会神奇地影响阿兹特克人（直到真的传播到了那里）。

An institution spawning is now a good thing
思潮的生成现在是件好事。

Asian nations will find themselves getting hit by a +50% tech cost from Renaissance earlier than they would in 1.31, but as they get further behind, their tech cost will stabilize even if they don't develop for institutions.

亚洲国家会发现自己比1.31版本号更早受到文艺复兴+50%科技成本的冲击，但随着他们越来越落后，即使他们不发展思潮，他们的科技成本也会稳定下来。

European nations won't be affected as much, but those on the forefront of technology will notice the increased tech cost, making institutions more relevant in this region than previously.

欧洲国家受到的影响较小，但处于科技前沿的国家会注意到增长的科技成本，使得思潮对该地区比以前更有作用。

When changing the code for this, it turned out we based a lot of seemingly unrelated things on countries' "tech cost from institutions", most of which were simply translated from the pre-Rights of Man "technology group" system. For example, joining a Trade League requires having at most +200% tech cost from institutions. Who would have known? You’ll find a lot of these addressed in 1.32's changelog.

当为此修改代码时，我们发现很多看起来无关的东西是基于“思潮导致的科技成本”，其中绝大部分都是由“人权”DLC之前的“科技组”系统中简单移植来的。例如，加入贸易联盟要求最多有+200%“思潮导致的科技成本”。谁会想到呢？你们会在1.32版本号的改动日志中看到很多这样的东西。

AI Improvements

AI提升

Every version of EU4 has had its own AI issues, and it's no secret that it has been getting worse for a while. For 1.32 some of the most debilitating issues have been fixed, and hopefully without causing too many new ones.

每个版本的EU4都有过自己的AI问题，而且众所周知这个问题最近正越变越糟。1.32版本号解决了一些最令人窒息的问题，同时我们希望不会导致太多的新问题。

EU4's AI is the product of years of incremental development, by a large number of developers. Most of the code is written after EU4's release, but a few lines date back as far as the late 90's. Writing AI is hard, so many of the systems are complex. And given the number of different developers, they don't all work in harmony. Essentially, for any given choice the AI has to make in army, navy or budgeting; it often has a dozen or so voices in its head telling it what to do. It's supposed to take all of them into account, but often the loudest voice drowns out all the rest.

EU4的AI是大量开发者经过多年开发堆积形成的。大部分代码是EU4发售后完成的，但是一些可以追溯到90年代。编写AI很难，有那么多复杂的系统。同时考虑到开发者的人数，他们的工作并不总是很和谐。通常情况下，AI在进行陆军、海军或预算决策时，会有一堆声音告诉他们要怎么做。按理来说应该考虑所有声音，但往往吼得最大声的那个会压过其他声音。

评估权重分数：-131244.0

权重分数检查项目：
基础：+110000
地形：-0.000
超过补给限制：+20000.000
敌方边境：
基础：-3000
距离：+1756.020
“杀光他们”：-260000
权重分数检查项目返回值：-131243.969

(yes, lower numbers are better)
（是的，数字小点会更好）

Here, the Ottoman army has made good progress on the Mamluk fort of Tabuk. But a voice intended for a completely different situation screams one order of magnitude louder than all other voices, that it should kill the rebels in Qahirah.

上图中，奥斯曼军队在马穆鲁克的塔布克要塞攻势喜人。但是一个比其他声音大得多的尖叫想要你掉头去做其他事，去消灭开罗的叛军。

Issues like this hinder the systems that *are* made for the given situation from doing their job. A lot of what has been done for 1.32's AI is to find and fix cases where systems interfere with each other.

这种问题阻碍了那些为给定情况设计的系统发挥作用。1.32版本号中AI解决的很多问题都是为了修复系统间的干扰问题。

Major Fixes

主要修正

1. Army

1、陆军

This is really the result of a bunch of small fixes, but essentially the AI will generally be more competent at achieving stuff with its armies, although individual moves can still be erratic. They will also keep a more significant army in peacetime

这其实是一堆小修正的合集，但最终AI用陆军达成目标的能力将更强，虽然个别操作仍将左右横跳。另外他们会在和平时期保持更多的陆军。

2. Naval Invasions

2、海军入侵

The AI can now decently perform naval invasions again, without too much shenanigans. This has a huge impact on the European AIs’ colonization efforts.

AI又可以继续执行像样的海军入侵，不太那么胡闹了。这对欧洲AI的殖民行为有巨大影响。

An almost fully colonized New World from one of our observer runs
在一局观察者模式运行中，新世界几乎被完全殖民了。

3. Economy

3、经济

Lots of small fixes mean the AI's economy should now be more solid. Budgeting has been improved and the choice of buildings to construct puts more emphasis on expected financial return.

一堆小修正，AI的经济更稳定了。预算管理得到改进，AI的建筑选择更强调经济回报。

A special shoutout to @Tempscire, whose reverse engineering of AI army behavior is somehow easier to understand than the code itself. We implemented his suggestion for terrain evaluation as-is, and changed how combat width is considered, albeit not as he suggested for technical reasons. Unfortunately he has many other good suggestions that will not be implemented at this point.

特别鸣谢@Tempscire，他对AI军队行为的逆向工程不知为何比代码本身更容易理解（屎山石锤了）。我们按原样实施了他关于地形评估的建议，并改变了战斗宽度的权衡方式，尽管由于技术原因没有按照他的建议。不幸的是，虽然他还有许多其他的好建议，但现在还不能实施。

翻译：Strategemata 毛里求斯大酋长

校对：113322qwe1 三等文官猹中堂

欢迎关注UP主和主播小牧Phenix！

欢迎关注牧游社微信公众号和知乎专栏！微信公众号改版为信息流，欢迎【置顶订阅】不迷路，即时获得推送消息！

B站在关注分组中设置为【特别关注】，将会在私信内及时收到视频和专栏投稿的推送！

欢迎加入牧有汉化，致力于为玩家社群提供优质内容！组员急切募集中！测试群组822400145!