








Thrall starting attributes (strength, agility, vitality, accuracy, and survival) are fixed based on the specific type of thrall: a Cimmerian Berserker will always start with 15 strength and 30 vitality. Total growth chance, summed across all stats, is also fixed for each type of thrall: a Cimmerian Berserker's growth chances will always add to 295%, while Beastmaster Teimos will always add to 356% or 357% (depending on rounding). The distribution of growth chances across each attribute is randomized within a range that's again unique to each source thrall type.
At each level gain, thralls will increase their attributes according to their corresponding growth chances: the listed chance (with any bonuses from food) applies to a +1 gain of a stat point, after which a second check at half the listed value is made for a second +1 gain. A thrall with a 100% growth chance in strength, then, has a 50% chance to gain +1 and 50% chance to gain +2 per level, with a total of 30 points expected over 20 levels. It's theoretically possible for a thrall to gain 40 points in one or even all stats over the course of 20 levels, though the chances of such "perfect rolls" are astronomically low (~1.3 per 100,000 for +40 in a single stat at 100+14% growth).
Additionally, at levels 10, 15 and 20, the thrall is randomly assigned a selection from the pool of Follower Perks that add to (and sometimes subtract from) attributes.
The randomization of initial growth chances, realized per-level attribute growth, and selected follower perks makes for an enormous range of possible outcomes and renders discussion of the "best" thrall somewhat academic. A "worse" thrall with even marginally better luck with attribute growth and perks can easily be significantly better than a "better" thrall that didn't get as lucky with its random chances.
在此我想问个问题,请问你会在贴吧回复层中楼吗?既你有什么想说的,想问的在我的这个帖子下面发,而不是重新开一层。不然我都不知道你对谁说的。我已经把我之前存档的数据,有测的,有计算的,测的误差与理论误差范围是多少你都可以自己去看。有些问题我就不讨论了,显然你没啥认知。比如去新图带什么,一次能带多少这个问题都是有错误的,比如一次能带15个奴隶左右而不是8个,带一次需要2天。以及去新图并不要这些奴隶,而是反过来从新图带回老图。并不是因为新图工匠比老图更厉害,而是因为新图本质上抓传奇工匠比老图简单,其次是新图工匠穿好看的衣服,所以基本上都是从新图带回老图的。其次是做一套力量套只需要4*4=16个古金,而新图废料的重量是10=100废料=500古金。只需要学习了技能就可以轻松在老图做。所以其实力量套并非什么很难获取的东西。这些都是后话,随着游戏的深入理解,你就会明白其中的原理。其实我也不是很想给斗士的研究打下什么比较无趣的刻板印象。比如其实游戏里的斗士在有一点理解之后你会发现其实输出还远不如人。就像很多人只知道打BOSS的时候在后面举盾让奴隶上,却不知道对于1W5K血的血月BOSS来说,不靠奴隶让人去单挑30秒就能杀了血月。当然说了这些都没什么用。
接下来是正题,首先你的研究方法,你自认为的数据。说实话在我眼里,当6个月前我还是个萌新的时候我和别人吵架哪个奴隶厉害的时候,我也是用你这种土办法。拿个奴隶,穿套衣服,给个武器,刷个怪然后测时间。当时我测的比你还多还请淡定。但是后来随着游戏理解的深入我并不屑于这套研究方法。怎么说呢,但凡你算个方差算个平均值,而不是连初中教的统计学方法给点数据,也不至于一点都不处理。我就说下结果,当初测的与理论计算其实很贴近,别看时间方差大,实际上平均值与理论其实是切合的。
再后来,就是我上面的这些数据就是对你的问题的解答。那段英文是对你关于练废率的解答。其中有测的,有推算的,甚至参考了国外RNG1000个特权给出的数据。其中还包括护甲计算之类的内容,希望对你建立标准模型的时候有所帮助。如果你想看详细的推导过程,你可以看精品贴里我关于奴隶站桩测试的帖子。

怎么说呢,就好比我说其实探宝客除了匕首之外其实长矛单手剑没啥区别一样,依旧很多人对长矛探宝客存在刻板印象说长矛AI更好一样。当我用数据指出他们的认知局限的时候他们都会认同。而你发表的意见却得到大多数人的反对,我觉得你更应该反思一下你的模型设想到底是缺少什么令人信服的东西。
别的不说,现在有个难点就是100+14%和64+14%这两种成长的在基于属性成长机制下他们概率的二项分布;初始力量与奴隶成长总属性与其成长概率的关系式;奴隶特权概率与成长概率的关联依旧是奴隶研究悬而未解的课题。我希望你有兴趣可以去研究下这三个课题,解答了这些问题反而能让你的话语更有说服力。而不是一大堆情绪表达