满命散兵余响套分析贴有误，余响触发次数低于预期

贝露赛布布

我一直在思考的是前提条件是否准确，因为原贴并没有给出证据证明主A与副A时间间隔，而这个条件会极大的影响余响触发次数，即总伤害。今天在discord上与Zel讨论时，他说他和KQM(theorycrafting community)之前已经在beta和正式服做过测试。在测试服里，Zel测到6命附加攻击时间间隔为6-7帧，而KQM在正式服测为8帧即0.13秒。由于余响套的效果在0.2秒内仅触发一次，每次持续0.05秒。那么如果第1次攻击触发效果，第2次攻击由于时间在0.05-0.2秒之间，则不会触发效果，因此前一贴关于余响的结论有误。

值得注意可能经常被忽略掉的一点是，基于余响的底层机制，余响的触发受延迟影响。当延迟较高(>100)时，触发次数会明显减少，比如在某段攻击击中怪物到服务器判定施加加伤模块，受延迟影响会出现在一至俩段伤害之后，甚至存在可能有些攻击不被服务器注册(即表现为触发概率下降)。这种现象的出现是受制于余响的底层机制堆栈注册在服务器端而不是用户端。


上图表示，散兵第一次攻击和第一次副攻击出现的间隔恰为7帧。视频是在60帧下录制的，因此可以直接比较帧数。

讨论：为什么存在视频显示某次攻击后某段与附加攻击均享受了余响的加成？这里我个人推测是由于延迟导致附加攻击被推迟了，服务器独立的给该次攻击施加了判定模块，亦或者视频中所展现的为前一次享受余响加成的附加攻击，亦或者其他原因(实际攻击帧数比b站视频更快，或者MHY犯低级错误)。我逐帧对比了楼里发的视频，可以看到第一段普攻21490(含余响加成)与随后的附加攻击12156(含余响加成)的间隔为8帧(假设原视频是在60帧下录制的)即0.13秒。根据余响的底层机制，理论上附加攻击由于在0.05s-0.2s内不能享受前一次攻击触发的余响，同时它本身也无法再触发余响，也许只有MHY的程序员知道原因(也许他们也不希望知道存在bug)。

关于余响的问题：我在KQM的文档里找到一些以往关于余响的记录([KQM文档])，感兴趣的可以看一下，大部分都附有相关的youtube或者文档佐证发现。里面有很多关于余响的一下发现，这里仅提几点：

一、距离的变化可能会导致俩次不同的攻击处于链接(link)状态，使得某些攻击可以同时触发余响(<0.05)或者独立判定。Yoimiya's N1 Procing Echoes Depends on Range By: The Golden Man#7947 Added: 2022-11-09 (v. 3.2) Last tested: 2022-04-04 (v. 2.6) Finding: How Yoimiya's N1 hits proc Echoes differs depending on range. Cause unknown, but the most obvious explanation is that getting closer causes the second hit to happen within the 0.05 buff procced by the first. However, I could not get the second N1 hit to proc without the first proccing, which should theoretically be possible if the first N1 does not proc the buff given the above explanation. Evidence: Two videos show my Yoimiya doing N1 versus a Ruin Guard, one at range and one as close as possible. When at range, each N1 hits proc independently: they can both proc, neither proc, or either one can proc. When as close as possible to the Ruin Guard, the N1 hits become linked: they either proc together or not at all.
宵宫近距离(贴身)攻击小宝时，第一段攻击处于链接状态，能同时触发或者不触发余响。中等距离攻击时，第一段攻击的俩次攻击独立判定。

二、一个角色的部分攻击处于链接状态Kazuha NA Interaction with Echoes By: WeirdFool#5729 Added: 2022-11-09 (v. 3.2) Last tested: 2022-04-04 (v. 2.6) Finding: Kazuha's 3-hit N5 is "linked" Kaz's 2-hit N3 is not "linked"
万叶第三段攻击的俩次攻击没有link，但是第五段攻击的前俩次攻击link。

三、余响触发受Ping和FPS影响Echoes Trigger Affected by Ping and FPS By: Mystathi#9705 and valorha#6262 Added: 2022-11-09 (v. 3.2) Last tested: 2022-08-14 (v. 2.8) Discussion Finding: Further data that could point to the set working as intended but be subject to ping/possibly low FPS/character multihits quick enough to take advantage of the lag between server and client (like Zhongli, or indicates that his spearkick is simply working differently compared to other multihits).

Evidence:

A typical N5D of Zhongli results in a higher proc rate than a full Ayato E sequence at 200+ ms, because of the spearkick all proccing together, while Ayato has no multihits (1500 hits Zhongli has 47% proc rate vs 1573 hits Ayato has ~34% proc rate)
Spacing hits seems to result in higher proc rate (around the theoretical one): ~50% by doing N1J vs 47% for N5D with Zhongli (777 hits)
Less data so just to observe a pattern: doing full combos with Ayato NAs results in a possibly higher proc rate than with E thanks to 2-hits on N4 --> need more data and to compare it with other characters to see if multihits with higher ping usually link and therefore results in a higher proc rate
More data from different testers on Zhongli that support the theory that the lower the ping, the lower the proc rate and vice versa (more ping = more spearkick hits linking, though we haven't found the higher cap where it becomes less)
余响的触发受网络延迟、帧数、多段攻击动作速度的影响(如果足够快可以减少服务器和用户间延迟的影响)。

四、在延迟超过100ms之后，余响触发几率降低。Echoes Proc Rate Falls Off After 100+ Ping By: mementomori#0001 Added: 2022-11-09 (v. 3.2) Last tested: 2022-04-17 (v. 2.6) Discussion Finding: EoO proc rate appears to fall off a little bit after 100-110 ping.
以绫人为例。

五、余响加伤害的buff与帧数关联。Echoes Buff Tied to Frame Data By: ItzSomebody#0029 Added: 2022-11-09 (v. 3.2) Last tested: 2022-04-04 (v. 2.6) Discussion Finding: Echoes effect can be shared with subsequent attacks as long as they fit in the 0.05 second interval.
余响的加上可以与接下来的攻击共享只要在0.05秒的区间内。

对使用余响玩家的建议：我的建议是尽可能的近距离贴身攻击同时确保网络延迟低，避免丢伤害。这似乎可以使散兵第一段攻击和附加攻击处于链接状态(基于楼里b站视频)。至于第二段和其附加攻击，第三段和附加攻击，第三段的俩次攻击是否互斥，仍然需要实验去验证。

圣遗物的建议：如果不介意楼阁重击起手，建议选择楼阁套。余响套的触发受多种因素的影响，目前仍然不确定为什么存在散兵第一段攻击和满命附加攻击在间隔8帧的情况下能同时触发余响特效的情况，同时不确定是否是每段攻击的附加攻击都能共享，也不清楚第三段攻击的俩此攻击是否在0.05秒与0.2秒的区间，如果在此区间，意味着第三段攻击的俩次攻击只能触发单次。同时，攻击距离、网络延迟和帧数是目前已经被确认一定会对余响的触发造成影响，也即在一些情况下余响会丢伤害。

只谈正事

漂亮，不错错。