CN 3.4 Abyss usage rates and comps (sample size 92k+)

Yeah, the numbers are skewed for all sorts of biases to include enemy selection dictating which elements have an edge for a given abyss, but when some units have been at or near the top since they were introduced for more than a year or since release, I think you can infer some units have fallen out of favor - or can you assume there are just a lot of new players who just happen to want to use Kazuha, Zhongli, Bennett, and other units that have enjoyed generally high levels of use?

Mugendramon posted...

However, exactly as I described in my previous post and as demonstrated by your data, the lower owned and more recent 5-stars are, in fact, quite on the opposite instead getting majorly favoured/skewed in the topic creator's and my post #26 images over the older and/or characters with higher owned percentages (and not necessarily due to being "stronger").

Examples of lower percentage characters who were artificially favoured over Ayaka, Ganyu, Tighnari, etc. in post #1 and post #26 images due to being owned by smaller proportions of users and/or due to being more recently pulled:
* Alhaitham (30,3% own%), Yae Miko (36,8%), Wanderer (31,7%), Itto (22,9%) or Cyno (20,4%), spread over the previous three pictures;
* Nilou (28,1 own%) and Shenhe (15,4%).

Higher percentage characters who were artificially biased against in post #1 and post #26 images due to being owned by larger proportions of users:
* Xiangling (100 own%), Raiden Shogun (86,9%), Ayaka (60,2%), Ganyu (54,2%), Tighnari (45,1%), who were for example artificially "surpassed" by Yoimiya (41,7% in post #26's second image as reposted below);
* Non-Tighnari Standard 5-stars in general like Keqing (+80 own%).

Very well. Let's take a look at the data systematically. It doe snot show what you claim it shows.

Here is the tabulated data for the limited 5-star characters based on this sample. It only shows limited 5-star characters so we have an apples to apples comparison. We avoid the problem that 4-stars are generally weaker than 5-stars, and standard 5-stars are generally weaker than limited 5-stars.

Data: https://bit.ly/3HyMtKF
R code: https://bit.ly/3Y8236T

If there is a systematic penalty for higher ownership rate, we would expect to see some significant relationship between certain variables. Let's check if any of these is true.

The most obvious: If there is a penalty for higher ownership rate, we might see a negative linear relationship between ownership rate and usage rate. Actually, the coefficient of regression is 0.81, so the relationship is positive--the higher the ownership rate, the higher the usage rate--exactly the opposite of your claim. The p-value is 0.0002, so the relationship is highly significant. The adjusted R-squared is 0.446, so ownership rate is not particularly good at predicting usage rate. But they still have a positive relationship.

Moving onto another test: How well correlated are usage rate and deployment rate? If the correlation is high, then the two variables are reasonable substitutes for each other. The correlation is 0.898, which is very high for a correlation outside of the physical sciences. Usage rate and deployment rates are good matches for each other. Incidentally, the coefficient of regression is 1.03, the p value is 2.6e-9, and the adjusted R-squared value is 0.798, all indicating that usage rate and deployment rates are good substitutes for each other.

Another measure is the usage rate premium. This is the usage rate minus the deployment rate. If there is an ownership penalty, then we would expect the usage rate premium to have a significant negative relationship with ownership rate. The regression coefficient has a p-value of 0.36, meaning there is no significant relationship between the usage rate premium and the ownership rate.

The data indicates that there is no systematic penalty for higher ownership rate. What you and the content creators have done are to cherry pick the data to support a preconceived notion.

Sampling bias being the major problem at hand thanks to the lack of more randomized sampling (in the grand scheme of things) and apparently due to being heavily mobile platform-biased, which hugely inflates things in favour of more recent mobile/casual-friendly and rarer units/banners at the detriment of previous units who are much more owned in the grand scheme of things like Xiangling, Raiden, Ayaka or Ganyu for example.

WeChat is a nearly universal, cross platform app used for communications, payments, government applications, banking, social media, and many things. Lacking WeChat is like having no device that runs iOS, Android, or Windows in the west--not impossible to function but crippling. The fact that the device is collected through a WeChat mini-app simply means it is collected from Chinese users.

* Hu Tao sold dramatically more than Yoimiya in practice (data available for Nahida's banner shown practically everyone pulled for Nahida in that character banner);
* Therefore, if there's only an absurdly unrealistically low 35% own rate with Yoi even going as far as having a bit more owned%, what this actually means is simply that the data in itself is simply massively skewed and missing major parts of the picture, hence being an awful indicator to impartially display what is actually taking place, for example there are simply a lot of Yoi mobile fans using the app (but even then it's still heavily skewed in favour of the in real life way rarer Yoi vs the way more prolific Hu Tao, considering the latter's banner big sales are also from mobile platforms), as opposed to a representation of a sample from a grander scheme of things, creating a huge false image.

You mistake revenue and ownership. Hu Tao gains significantly from C1. Yoimiya gains almost nothing from constellations. There is a large gap between Homa and Hu Tao's next tier of weapons (all BP or gacha 4-stars), which is much better than her best 3-star weapon (Liyue chest limited White Tassel). Yomiya does about the same damage with a 5-star weapon, Rust, or Slingshot.

It is the same reason Zhongli has higher ownership rate than Raiden, despite having about half the revenue from his banners: Zhongli does not benefit much from constellations, and Vortex Vanquisher is terrible. Raiden gains a lot from C2, and Engulfing Lightning is excellent.

ryossalivar posted...

Yeah, the numbers are skewed for all sorts of biases to include enemy selection dictating which elements have an edge for a given abyss, but when some units have been at or near the top since they were introduced for more than a year or since release, I think you can infer some units have fallen out of favor - or can you assume there are just a lot of new players who just happen to want to use Kazuha, Zhongli, Bennett, and other units that have enjoyed generally high levels of use?

Supports are unique and can be used in multiple teams, not so much dps. They may be more boring to pull but supports are what make a dps, not the actual dps unit

Supports will always appear more often at the top. We just didn't have as much variety for dps before and we have gotten main dps left and right as of late

ryossalivar (Post #41)
Yeah, the numbers are skewed for all sorts of biases to include enemy selection dictating which elements have an edge for a given abyss, but when some units have been at or near the top since they were introduced for more than a year or since release, I think you can infer some units have fallen out of favor

How are the much more highly owned Raiden and Xiangling "falling out of favour" compared to the much rarer owned Yoimiya for example, is the real massively skewed data problem at hand which is painting a heavily misleading picture to the vast majority of viewers who'll have no idea that the true reason some of the characters are showing much higher than the others is precisely because they are being heavily favoured due to recency and/or low owned rate percentage technicality "loopholes".

This means, if anything, that super rare characters like Shenhe for example are in reality vastly ignored/expendable and weren't pulled by the extreme majority of players (contrast with the single banner Yelan for example) yet the images give a heavily distorted idea of what is actually taking place.

- or can you assume there are just a lot of new players who just happen to want to use Kazuha, Zhongli, Bennett, and other units that have enjoyed generally high levels of use?

This is a false equivalence because support characters have a much greater probability to be slotted in teams vs on-field DPS characters, as I specifically addressed in my previous post; this was also the reason in my post #40 I have put the rarer and more support-based Nilou and Shenhe on a different paragraph than the DPS 5-stars which are mostly going after different (much fewer) team slots than supports.

‎

Ihelp (Post #35)
I have to ask what kind of player are you, speedrunner, or just meta for fun?

With how passionate you are I imagine you must have c6 r5 hu tao with c2-4 yelan/r5 elegy.

Well yes, but actually no: as you can see I am by all means a gigantic metachurl although I'm not a whale and my posts are generally assuming realistically accessible C0 5-star characters primarily equipped with 4-star weapon levels and assuming normal gameplay usage to be relatable to players in general, as opposed to speedrunning and the like (which heavily prefers certain characters and teams than others and has vastly different targets).

‎

reignofkain (Post #36)
Falling off would imply Hu Tao has started on a rather irreversible downward trend, she has not. Her usage is down one cycle, then up the next as Abyss blessings change to favor her rerun.

My point exactly; I was one of the posters disputing the "falling off" part (which in turn would have some inherently pretty bad implications to every other DPS character in the game (especially on-field ones) at realistic levels of investment, since they would be generally affected relatively a lot more than Hu Tao personally).

Hu Tao's damage hasn't changed, as evidenced by people literally brute-forcing ASIMON instead of bringing Electro against it. She doesn't magically bring other units' damage down, the enemy composition and blessings are simply giving people a little leeway into using other teams that function very well.

In this topic's particular case that is only part of the problem associated to Usage Rates, as I've been discussing in my other points and posts.

‎

ManChooses (Post #37)
Yes, that's what's happening. Most Main-DPS's value is falling down at low invest bc supports and reactions are becoming too strong.

Low Invest MainDPS is becoming the least relevant role in the game. Basically the top of "roll if you like 'em".

While it is true that nowadays there is greater interchangeability to the on-field damage dealers than before (some bringing relatively more powerful capabilities to the table than others), which in turn makes easily slotted characters like Kazuha or Yelan by default more meta for long term players than the more interchangeable on-field DPS characters, at the end of the day the on-field DPS characters are still arguably even more critical, especially at first to newer accounts with a much lower overall selection of characters (in which case picking up a new 5-star DPS character is still a major account power-up, considering C0 5-star characters are more or less like a "shortcut" that is like instantly pulling higher constellation 4-stars of comparable roles, especially the stronger top meta 5-stars), and considering 4-stars in general are gradually getting increasingly harder to constellate for players at large as the 4-star pool widens over time, with 4-stars increasingly being actively used to bait players into pulling for expensive 5-star constellations.

‎
There is also the situation with players being attached to the on-field characters more, which is pretty important for HoYo to keep selling subsequent supportive 5-star characters (like Shenhe for Ayaka and the other cryo waifus) and their own separate armament and constellations, so they'll keep cycling associated content/Abyss stages to intentionally incentivize players to roll for these characters (same for the closely associated Ventiable content for example). After all, if we pull Kazuha or Zhongli first but then don't have a variety of strong characters to actually buff/shield then it's basically putting the cart before the horse by pulling the support characters first before the DPS ones to actually be supported.

Genshin Impact is a game meant to be played for the next decade and at the end of the day the long-term value of 5-star on-field DPS characters still remains pretty important (especially to forward-thinking players more calculistic about damage gains vs Primogems spent and the long term potential of given units/teams) even at lower levels of investment, since pulling 5-star on-field DPS characters is the best way to go to start building a new account around them, starting artifact farming and the like.

‎

‎Dorami (Post #42)
There is a large gap between Homa and Hu Tao's next tier of weapons (all BP or gacha 4-stars), which is much better than her best 3-star weapon (Liyue chest limited White Tassel). Yomiya does about the same damage with a 5-star weapon, Rust, or Slingshot.

The above paragraph is false and the opposite of reality:

• Yoimiya (three relevant weapons) is the one lacking access to substantive accessible weapon selection vs Hu Tao (eight~nine relevant weapons);
• In other words, Yoimiya players wanting better weapons need to specifically pull her Limited 5-star bow with it's more niche passive in order to substantially surpass her 3-star Slingshot and the 4-star Rust bows, her only relevant alternative weapons boiling down to a minuscule three relevant weapons, two of which gacha;

• What the above means is the exact opposite of what you mentioned: in truth, Yoi is far more reliant on the weapon banner than Hu Tao, since the latter has a much wider selection of weapons (four~five 4-star, one 3-star and three 5-star relevant polearms, eight~nine total polearms or about three times more as Yoi);
• Hu Tao's F2P best in slot is the Missive Windspear, which is like a free R1 Deathmatch battle pass polearm or R1 Dragon's Bane gacha 4-stars;
• C0 Hu Tao with above F2P weapon can still substantially personally and in a team level largely out-DPS C0 R1 Thundering Pulse Yoi in a realistic team with both at high and equal levels of artifact investment, meaning it is ironically a lot cheaper to pull C0 Hu Tao and then learn how to play her than it is to pull the Thundering Pulse to begin with.

https://gamefaqs.gamespot.com/a/user_image/0/7/8/AAdJrwAAEKIe.png

Starting as of this post and to make the enormous data skewing problem more easier to understand, not only in this topic alone but with the feel-based absolute junk metrics that are Usage Rates and which should be blatantly self-evident to anyone not under the effects of Naku Weed...

‎

• ...I am now going to rephrase the topic creator's incredibly misleading points and then I am going to start to ask for mathematical evidence on why Yoimiya is actually "stronger" or "more meta" than the Raiden Shogun and Xiangling, not to mention Childe, Ayaka or Ganyu:

https://gamefaqs.gamespot.com/a/user_image/2/8/7/AAdJrwAAEJsf.jpg

• "The Raiden Shogun goes down a tier from the previous cycle, presumably with Yoimiya's strong rerun heavily eating into Raiden's usage for Bennett/Dendro teams.
• Yoimiya with a massive 52,4% usage rate.
• Raiden and Xiangling appear to be struggling to hold on to relevancy despite their reputation as “T0” characters, with their usage rates more aligned now with the likes of Childe and Ayaka than the top tiers.

Definitely a bad cycle for Raiden, but the fact that it’s this easy to grief her playstyle with unfavorable match-ups should be held against her in character evaluations IMO. Same with Xiangling."

Hopefully this makes the blatant problem at hand easier to understand.

‎

‎Dorami (Post #42)
You mistake revenue and ownership. Hu Tao gains significantly from C1. Yoimiya gains almost nothing from constellations.

It is the same reason Zhongli has higher ownership rate than Raiden, despite having about half the revenue from his banners: Zhongli does not benefit much from constellations, and Vortex Vanquisher is terrible. Raiden gains a lot from C2, and Engulfing Lightning is excellent.

WeChat is a nearly universal, cross platform app used for communications, payments, government applications, banking, social media, and many things. Lacking WeChat is like having no device that runs iOS, Android, or Windows in the west--not impossible to function but crippling. The fact that the device is collected through a WeChat mini-app simply means it is collected from Chinese users.

Raiden example is nonsensical given her still massive own% rate that is about two and a half times greater than Hu Tao's in the heavily skewed data you posted despite Hu Tao having the second highest-selling debut once we account for missing Homa sales and her top selling solo rerun.

‎
There was no "mistaking of revenue and ownership":

• What actually happened is that the data you presented is simply massively skewed against earlier day units/banners and/or more highly owned units and non-mobile platform players (for whom button-destroying gameplay gets devalued a lot, much less when there's lots of damage to be gained elsewhere with the same Primogems);
• While heavily skewing trends in favour of more recent players and/or lower owned units and by not including much earlier day player sampling (massive sampling bias problem), for example numerous Yoimiya mobile fans simply flocked to that app.

As opposed to your heavily skewed data, the following is the closest to true "nearly universal, cross platform" randomized data we actually got globally before HoYo shut it down, once more as also shown in the previous TenTen video, note the comically gargantuan discrepancy reaching at least as high as a 72 own rate% Hu Tao vs your data's outrageously and unrealistically low 34,8 own rate% Hu Tao which is more than two times less vs the following most randomized/universalistic sampling:

https://gamefaqs.gamespot.com/a/user_image/0/9/1/AAdJrwAAEKIr.jpg

https://gamefaqs.gamespot.com/a/user_image/0/9/2/AAdJrwAAEKIs.png

Many characters like Yoi didn't get reruns by then, which is how they're so low compared to today, but the point is that Hu Tao's 72% owned (vs your absurdly low 34,8%) and Klee's 37% (vs your significantly lower 24,9%) older players samples were much more well sampled and encompassed in the above greater picture due to their greater sample randomization and it's own recency at the time, as drastically opposed by the heavily mobile-preferential data you posted (itself still heavily slanted against the older Hu Tao since in practice she again vastly outsold Yoi even on mobile).

• Even so, Raiden notably already had the exact same 86 own% and Xiao has an almost identical 39% (vs your 38,1%);
• While Ayaka's 55% (vs your 60,2%), Ganyu's 48% (vs your 54,2%), Venti's 61% (vs your 56,8%) and the ~80% Standard 5-stars are all already pretty close to the app data you posted.

Very well. Let's take a look at the data systematically. It doe snot show what you claim it shows.
...
The data indicates that there is no systematic penalty for higher ownership rate. What you and the content creators have done are to cherry pick the data to support a preconceived notion.

You are projecting, since:

1. You yourself have completely failed to demonstrate how Usage Rates aren't extremely favourable to specific units (especially lower owned and/or more recent ones) at the great detriment of older and/or higher owned characters (which get greatly devalued in these deeply misleading and dishonest Usage Rate tables that in truth aren't in reality actually demonstrating any effective unit strength vs others, since correlation does not imply causation), as exemplified above and in prior posts of mine;
2. You have also failed to demonstrate how exactly is Yoimiya actually stronger than Raiden, Xiangling or the other top meta characters and you're now attacking a tangential straw man that isn't actually addressing these arguments at hand;
3. All the while you're equivocating about and tip-toeing around the simple fact that the other favoured Limited 5-stars were specifically pulled on-demand by said subsets of players, i.e. according to such players' personal preferences/tastes (which are very far from necessarily translating to greater power)...

You are defending deeply flawed and misleading data with heavy sampling bias that is severely biased in favour of:

• Player favouritism as opposed to actual unit gameplay strength and meta relevance;
• Mobile and/or casual-oriented audience;
• Newer banners/characters including much lower owned ones (= huge false equivalence vs higher owned characters, who will inherently be at a major unfair disadvantage while simultaneously not actually showing the new characters as being stronger than the older top meta ones to begin with, and actually sending the exact opposite message to most people like in this topic).

Effective immediately, I'm going to start asking for evidence on how Yoimiya is allegedly "more meta" than Raiden, Xiangling and the other top meta regulars.

Mugendramon posted...

As opposed to your heavily skewed data, the following is the closest to true "nearly universal, cross platform" randomized data we actually got globally before HoYo shut it down, once more as also shown in the previous TenTen video, note the comically gargantuan discrepancy reaching at least as high as a 72 own rate% Hu Tao vs your data's outrageously and unrealistically low 34,8 own rate% Hu Tao which is more than two times less vs the following most randomized/universalistic sampling:

https://gamefaqs.gamespot.com/a/user_image/0/9/1/AAdJrwAAEKIr.jpg

https://gamefaqs.gamespot.com/a/user_image/0/9/2/AAdJrwAAEKIs.png

Many characters like Yoi didn't get reruns by then, which is how they're so low compared to today, but the point is that Hu Tao's 72% owned (vs your absurdly low 34,8%) and Klee's 37% (vs your significantly lower 24,9%) older players samples were much more well sampled and encompassed in the above greater picture due to their greater sample randomization and it's own recency at the time, as drastically opposed by the heavily mobile-preferential data you posted (itself still heavily slanted against the older Hu Tao since in practice she again vastly outsold Yoi even on mobile).

Nice try. I notice that you did not cite your source, but I, as well as many users on this board, recognize that website. It is spiralabyss.org. This is the "submit your data" portion of the website back then: https://web.archive.org/web/20211031080545/https://spiralabyss.org/submit-your-data

Notice the instructions ask the player to make their UID public, then enter and submit the UID. That is a voluntary submission process, not a random sampling. Your source has a selection bias from voluntary submission, precisely as the Teyvat mini helper app does. You have failed to demonstrate that one source is more biased than the other.

You also failed to reply to the explanation of why the CN data is not merely mobile based.

You are projecting, since:
1. You yourself have completely failed to demonstrate how Usage Rates aren't extremely favourable to specific units (especially lower owned and/or more recent ones) at the great detriment of older and/or higher owned characters (which get greatly devalued in these deeply misleading and dishonest Usage Rate tables that in truth aren't in reality actually demonstrating any effective unit strength vs others, since correlation does not imply causation), as exemplified above and in prior posts of mine;

Actually, I have demonstrated that there is no significant negative relationship between ownership rate and usage rate for comparable units, e.g. among limited 5-star units. See my previous post #42. If you do not understand statistical analysis, say so.

And a reminder of the burden of proof: The null hypothesis is that there is no significant relationship between ownership rate and usage rate. If you want to claim otherwise, demonstrate it, not through handwaving of Youtube videos or cherry picking of specific percentages. Demonstrate it through statistical analysis and show your work.

Dorami (Post #46)
1. Nice try. I notice that you did not cite your source, but I, as well as many users on this board, recognize that website. It is spiralabyss.org. This is the "submit your data" portion of the website back then: https://web.archive.org/web/20211031080545/https://spiralabyss.org/submit-your-data

2. Notice the instructions ask the player to make their UID public, then enter and submit the UID. That is a voluntary submission process, not a random sampling. Your source has a selection bias from voluntary submission, precisely as the Teyvat mini helper app does. You have failed to demonstrate that one source is more biased than the other.

3. You also failed to reply to the explanation of why the CN data is not merely mobile based.

It is pretty sad how you were now reduced to explicit lying to try to attack my credibility after I specifically falsified your previous incorrect assertion that the useless Usage Rate trash data you provided was in any way shape or form "nearly universal" when it is literally missing gigantic chunks of users of top selling characters of all things, thereby showing just how skewed and comically flawed the whole Usage Rate picture and associated conclusions are, with how detached from reality with it's demonstrated massive sampling bias problem they are.

‎
1. This is a complete lie: Dorami is deliberately playing dumb since the source in question is implicit and a central part of the debate and the literal specifics of how the data is acquired were even addressed in-depth in the TenTen video that is central to and has been continually brought up in this discussion, what I did was simply transplant the Own% information as shown in that video over here directly to contrast with the comically low Own% from the other source; literally one of the first things shown in the video is a +70 Own% Hu Tao, as corroborated by sales data and basic common sense.

https://www.youtube.com/watch?v=LsLOAbzDuPQ

https://gamefaqs.gamespot.com/a/user_image/0/3/9/AAdJrwAAEKnH.jpg

2. False:

• First you haven't shown how heavily skewed Chinese mobile-centric data is in any way "more universalistic" than literal global data I brought up, unless the universe is only comprised of China;
• Secondly, with the exception of missing subsequent reruns which hadn't yet taken place at the time as I detailed, my data is literally even far more complete than the one you displayed for relevant characters at hand like the Raiden Shogun (exact same Own% in my samples), in other words the data I provided successfully sampled many players who weren't encompassed by and were excluded from your heavily skewed samples that are far more biased towards more recent players/characters/banners.

Thus my source literally offered far more randomized and way better sampling than your heavily skewed one which is missing major swathes of the game's population which were filled in by my sample, and thus my source is objectively far less skewed than your one, providing major missing pieces of the puzzle that didn't even exist/went under the radar of your source to begin with.

3. Completely false, I never said the CN data "is merely mobile based", this is your invention/straw man, so please refrain from putting words in my mouth.

‎

Actually, I have demonstrated that there is no significant negative relationship between ownership rate and usage rate for comparable units, e.g. among limited 5-star units. See my previous post #42. If you do not understand statistical analysis, say so.

And a reminder of the burden of proof: The null hypothesis is that there is no significant relationship between ownership rate and usage rate. If you want to claim otherwise, demonstrate it, not through handwaving of Youtube videos or cherry picking of specific percentages. Demonstrate it through statistical analysis and show your work.

If you do not understand what a straw man fallacy and how the burden of proof is on you, then say so.

• You haven't demonstrated anything of relevance to the actual argument at hand since you haven't even shown how has the mighty, prolific meta giant Yoimiya griefed and bested the weakly, poor little Raiden Shogun, Xiangling and other top meta "underdogs";
• In other words, you haven't even demonstrated the merit to using trashy Usage Rates in the first place, an inherently fatally flawed, feel-based popularity contest indicator vastly twisted by player favouritism and banner of the day as opposed to necessarily actual displays of characters' gameplay strengths and objective meta advantages to given characters.

As I specified in my previous post:

• Please provide the mathematical evidence to demonstrate how Yoimiya is stronger than the likes of the Raiden Shogun, Xiangling and the other top meta regulars, which is the kind of deeply misleading messages sent by these junk usage rate metrics heavily driven by simple favouritism of much smaller subsets players in this heavily skewed, useless data.

As I said before:

• You completely failed to demonstrate how Usage Rates aren't extremely favourable to specific units (especially lower owned and/or more recent ones) at the great detriment of older and/or higher owned characters (which get greatly devalued in these deeply misleading and dishonest Usage Rate tables that in truth aren't in reality actually demonstrating any effective unit strength vs others, since correlation does not imply causation).

https://gamefaqs.gamespot.com/a/user_image/0/4/0/AAdJrwAAEKnI.jpg

If you were telling the truth and if certain characters weren't being dramatically more disadvantaged than others who in turn are dramatically more advantaged such as simply due to being owned by dramatically less players and/or being newer releases, then posts like these from misled players who were unfortunately memed by the trash metrics at hand and my own counterexample from the image above wouldn't even exist.

Mugendramon posted...

It is pretty sad how you were now reduced to explicit lying to try to attack my credibility after I specifically falsified your previous incorrect assertion that the useless Usage Rate trash data you provided was in any way shape or form "nearly universal" when it is literally missing gigantic chunks of users of top selling characters of all things, thereby showing just how skewed and comically flawed the whole Usage Rate picture and associated conclusions are, with how detached from reality with it's demonstrated massive sampling bias problem they are.

‎
1. This is a complete lie: Dorami is deliberately playing dumb since the source in question is implicit and a central part of the debate and the literal specifics of how the data is acquired were even addressed in-depth in the TenTen video that is central to and has been continually brought up in this discussion, what I did was simply transplant the Own% information as shown in that video over here directly to contrast with the comically low Own% from the other source; literally one of the first things shown in the video is a +70 Own% Hu Tao, as corroborated by sales data and basic common sense.

https://www.youtube.com/watch?v=LsLOAbzDuPQ

https://gamefaqs.gamespot.com/a/user_image/0/3/9/AAdJrwAAEKnH.jpg

2. False:
* First you haven't shown how heavily skewed Chinese mobile-centric data is in any way "more universalistic" than literal global data I brought up, unless the universe is only comprised of China;
* Secondly, with the exception of missing subsequent reruns which hadn't yet taken place at the time as I detailed, my data is literally even far more complete than the one you displayed for relevant characters at hand like the Raiden Shogun (exact same Own% in my samples), in other words the data I provided successfully sampled many players who weren't encompassed by and were excluded from your heavily skewed samples that are far more biased towards more recent players/characters/banners.
Thus my source literally offered far more randomized and way better sampling than your heavily skewed one which is missing major swathes of the game's population which were filled in by my sample, and thus my source is objectively far less skewed than your one, providing major missing pieces of the puzzle that didn't even exist/went under the radar of your source to begin with.

3. Completely false, I never said the CN data "is merely mobile based", this is your invention/straw man, so please refrain from putting words in my mouth.

‎

If you do not understand what a straw man fallacy and how the burden of proof is on you, then say so.

* You haven't demonstrated anything of relevance to the actual argument at hand since you haven't even shown how has the mighty, prolific meta giant Yoimiya griefed and bested the weakly, poor little Raiden Shogun, Xiangling and other top meta "underdogs";
* In other words, you haven't even demonstrated the merit to using trashy Usage Rates in the first place, an inherently fatally flawed, feel-based popularity contest indicator vastly twisted by player favouritism and banner of the day as opposed to necessarily actual displays of characters' gameplay strengths and objective meta advantages to given characters.

As I specified in my previous post:
* Please provide the mathematical evidence to demonstrate how Yoimiya is stronger than the likes of the Raiden Shogun, Xiangling and the other top meta regulars, which is the kind of deeply misleading messages sent by these junk usage rate metrics heavily driven by simple favouritism of much smaller subsets players in this heavily skewed, useless data.

As I said before:
* You completely failed to demonstrate how Usage Rates aren't extremely favourable to specific units (especially lower owned and/or more recent ones) at the great detriment of older and/or higher owned characters (which get greatly devalued in these deeply misleading and dishonest Usage Rate tables that in truth aren't in reality actually demonstrating any effective unit strength vs others, since correlation does not imply causation).

https://gamefaqs.gamespot.com/a/user_image/0/4/0/AAdJrwAAEKnI.jpg

If you were telling the truth and if certain characters weren't being dramatically more disadvantaged than others who in turn are dramatically more advantaged such as simply due to being owned by dramatically less players and/or being newer releases, then posts like these from misled players who were unfortunately memed by the trash metrics at hand and my own counterexample from the image above wouldn't even exist.

What the f*** is this

ManChooses posted...

Well Yoimiya is better than Raiden at fighting Thunder Manifestation, let me tell you that x'D

Anyway, i wonder how Hoyoverse will Target bloom stuff in Abyss later. I was thinking of enemies that pulsate element and mess ownership of some vegetables.

Just put dendro hypostasis in both side of abyss