DPS discrepancy between classes (courtesy of ESO Logs)

MartiniDaniels

kargen27 wrote: »

The data means nothing if each class isn't being played by the same player. What you need is at least 100 parses of each class by the same person with the same group providing the same buffs and debuffs.
Then you need to do that several more times with different people also doing all classes with the same group.

And to get fair numbers the person doing the testing would need to be proficient with each class. You can't expect someone who spent hours and hours on a practice dummy perfecting his rotation on a DK to give you fair numbers on a nightblade they have barely played.

Well, you understand that it's not like everybody run on random toons to vSS and then this statistics appear? On the PTS those persons proficient with every class did all the measurements and posted results. Thereafter every monkey mimic'd ready-to-go builds and so in original post we can see thousands of stamcros, stamblades, magsorcs and magplars which is exact match of Liko's top classes for stamina and magicka. There are some magblades who forgot that they are together with wardens now, some magcros whose abilities are bugged and some persistent stamDK and stamplars which are in good positions but most probably carried by lokki and without it they will be weaker then 2-pet sorc with 5 buttons to press instead of 10.
Oh ok, there are some top-notch stamsorcs too, but as far as i know only hardcore fans were playing on stamsorcs and all other switched to magsorcs long ago.

T3hasiangod

MLGProPlayer wrote: »

SidraWillowsky wrote: »

MLGProPlayer wrote: »

@ at all the folks saying this data is misleading because it doesn't fit their narrative that the game is perfectly balanced.

Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have. This is the best we're going to have, and it's good enough to draw conclusions from. It's almost as if there is a different standard for collecting data for a dissertation and publication and analyzing video game balance.

Have you ever in your life taken a statistics course? I'm guessing no, because if you had, you'd know about confounding variables. Let's play a game:

What's wrong with drawing definitive conclusions from the following two graphs?





No, higher ice cream sales don't cause higher murder rates or vice versa- there is a variable that accounts for both increases: season. Murder rates are higher in the summer, as are ice cream sales.

The ironic thing is that YOU are the one interpreting the data to fit your narrative by choosing to ignore any factors aside from class that may be affecting those numbers.

Looking at your chart, sure, some classes are more popular and tend to be considered stronger. No one is refuting that. The issue, however, is that you're not thinking about those factors- the popular, "stronger" classes are probably going to be played more frequently by the "pros". The "pros" are probably going to be running Sunspire more frequently for the scores. What I'm saying is that the percentage of top players playing on stamDKs or stamSorcs is going to be higher than the percentage of top players using some of the lower-DPS classes. And since these guys pull insanely high numbers, those are going to inflate the mean, and inflate it more for the classes that they're playing more frequently. That's one factor that I think may be at play, but in the end we don't know enough about the data to conclude that, which applies to your OP as well.

Also, LOL:
Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have.

I'm a statistician and I work in public health. One of my projects is working with data from children who have come into contact with the child welfare system who have been tracked over time to examine the effect of the CW system on social/emotional/behavioral outcomes. The data can be really messy sometimes, but if we sat back and said "oh, we'll work with what we have without any critical thinking whatsoever", the impact of any analyses done haphazardly could very well make a negative impact on the lives of children later on. My example is much more serious than yours but jesus, everything you've stated as a fact about data and analysis has been wildly incorrect.

I'm a PhD student. I'm well aware of how to interpret data.

But comparing public health data to *** video game damage stats is laughable.

You can often improve the quality of data in the real world (especially if you're the one collecting it). In a game where you are at the mercy of the developer and the limited tools they provide you, there is only so much you can do with it, and only so much effort you're going to put into it. Nobody is publishing a *** peer reviewed paper with this data.

Yes, you should work with what you have, but you do not simply blindly interpret the data as-is. If you know your data has issues like selection bias or some other issue that prevents good interpretation, you note that in your discussion and take steps to try to correct for these issues. What you are advocating for would absolutely get thrown out by any self-respecting scientist.

Bad data does not always means bad results, but bad data combined with bad practices will result in absolutely terrible results and interpretation. Part of being a good scientist is knowing how to interpret the data, what weaknesses your data has, how to address those weaknesses, and know where to draw the line when it comes to using this data to draw further conclusions.

I would not use results from a 95 percent Caucasian population to recommend actions for a 95 percent African-American community because the Caucasian results are likely not applicable to the African-American community. Even if the data we collected from the Caucasion population is free from any sort of bias or flaws, we still cannot use these results to apply them to other dissimilar populations. That is basic statistics 101.

This is the same concept: the results are skewed and suffer from selection bias, so while we can say there is likely good internal validity (i.e. we can draw conclusions within this specific population), we may not have good external validity (i.e. we cannot take the conclusions and apply it to other populations).

idk

.

MLGProPlayer wrote: »

SidraWillowsky wrote: »

MLGProPlayer wrote: »

@ at all the folks saying this data is misleading because it doesn't fit their narrative that the game is perfectly balanced.

Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have. This is the best we're going to have, and it's good enough to draw conclusions from. It's almost as if there is a different standard for collecting data for a dissertation and publication and analyzing video game balance.

Have you ever in your life taken a statistics course? I'm guessing no, because if you had, you'd know about confounding variables. Let's play a game:

What's wrong with drawing definitive conclusions from the following two graphs?





No, higher ice cream sales don't cause higher murder rates or vice versa- there is a variable that accounts for both increases: season. Murder rates are higher in the summer, as are ice cream sales.

The ironic thing is that YOU are the one interpreting the data to fit your narrative by choosing to ignore any factors aside from class that may be affecting those numbers.

Looking at your chart, sure, some classes are more popular and tend to be considered stronger. No one is refuting that. The issue, however, is that you're not thinking about those factors- the popular, "stronger" classes are probably going to be played more frequently by the "pros". The "pros" are probably going to be running Sunspire more frequently for the scores. What I'm saying is that the percentage of top players playing on stamDKs or stamSorcs is going to be higher than the percentage of top players using some of the lower-DPS classes. And since these guys pull insanely high numbers, those are going to inflate the mean, and inflate it more for the classes that they're playing more frequently. That's one factor that I think may be at play, but in the end we don't know enough about the data to conclude that, which applies to your OP as well.

Also, LOL:
Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have.

I'm a statistician and I work in public health. One of my projects is working with data from children who have come into contact with the child welfare system who have been tracked over time to examine the effect of the CW system on social/emotional/behavioral outcomes. The data can be really messy sometimes, but if we sat back and said "oh, we'll work with what we have without any critical thinking whatsoever", the impact of any analyses done haphazardly could very well make a negative impact on the lives of children later on. My example is much more serious than yours but jesus, everything you've stated as a fact about data and analysis has been wildly incorrect.

I'm a PhD student. I'm well aware of how to interpret data.

Good for you. However, some of your comments relating to how you have interpreted the data are called into question. The most obvious is the quote I provided below.

.

MLGProPlayer wrote: »

We have a DPS gap between the top and bottom class that is ~50%. Balance has never been worse in this game.

Not sure what firing Wrobel accomplished.

The SS you provided in the OP do not show any class getting only 50% of the damage of the top class. You would have to use the WW parses to back up your comment here and WWs are clearly not a class.

Granted, I do agree those SS do show to much difference between the classes that hopefully Zos will learn from, but your overall analysis has flaws as has been pointed out in this thread. As a MLG Pro player I expect you can see some of these points.

Runefang

MartiniDaniels wrote: »

Runefang wrote: »

cheifsoap wrote: »

T3hasiangod wrote: »

MLGProPlayer wrote: »

@ at all the folks saying this data is misleading because it doesn't fit their narrative that the game is perfectly balanced.

Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have. This is the best we're going to have, and it's good enough to draw conclusions from. It's almost as if there is a different standard for collecting data for a dissertation and publication and analyzing video game balance.

No, this is not good enough to draw definitive conclusions. It is enough to say that this is the DPS trend for Sunspire, but not for drawing conclusions on overall DPS balance.

Good video btw, bro. Quick comment; why should anyone care about the "low end" dps for a specific class. No one cares about someone who messes up their rotation or has a laggy connection, as a matter of fact if anything it would be a bad metric to collect. IMO, you want the higher end of the dps scale because they are squeezing out the most dps they can on said trial. If ZOS were to balance with the low end in mind, you would have these uber players pulling crazy dps parses and what does that really do for "balance"? Not a whole lot

Nobody cares what suboptimal set ups (be it group, gear or rotation) pull in terms of dps. You just can’t conclude much from 59 Magden parses other than not many people who are in groups with publicly loaded logs run a Magden.

What you definitely can’t assume is that the top 1% of Magden parses (1 person) is equivalent to the top 1% of Stamblade parses (20 people). That top Magden result is highly unlikely to be representative of the class dps ceiling. Heck the Stamblade is also unlikely to be representative of the ceiling.

Aside from obviously bad data interpretation the other reason to disagree with the conclusion that the game balance is that bad is that Liko pulls numbers on all the classes which are closer to each other. At least I know he is closer to the Magden ceiling on the 21m dummy than the lone Magden parse on esologs.

Again I don’t disagree that class imbalance exists, I’m just disagreeing that it’s as bad as it’s being made out to be.

Sigh, 16k dps difference, "same numbers"

I never once said Liko pulled the same numbers on all classes. Just ones that are closer to each other. Specifically closer to each other than the stupid top 1% of the esologs. A 16k dps spread is much smaller than the 24k spread suggested with that data. It becomes smaller again when you consider the top was 103k with a Necromancer who gets major vulnerability on the dummy.

Consider the top Magden parse on Yolnahkriin at the moment. There is 15% major force uptime in that fight, 0% major vulnerability.

The top Stamblade parse on Yolnahkriin (and top overall) has a 50% major force uptime and 26% uptime on major vulnerability.

You can't tell me that's not comparing apples and oranges given those differences.

MLGProPlayer

idk wrote: »

.

MLGProPlayer wrote: »

SidraWillowsky wrote: »

MLGProPlayer wrote: »

@ at all the folks saying this data is misleading because it doesn't fit their narrative that the game is perfectly balanced.

Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have. This is the best we're going to have, and it's good enough to draw conclusions from. It's almost as if there is a different standard for collecting data for a dissertation and publication and analyzing video game balance.

Have you ever in your life taken a statistics course? I'm guessing no, because if you had, you'd know about confounding variables. Let's play a game:

What's wrong with drawing definitive conclusions from the following two graphs?





No, higher ice cream sales don't cause higher murder rates or vice versa- there is a variable that accounts for both increases: season. Murder rates are higher in the summer, as are ice cream sales.

The ironic thing is that YOU are the one interpreting the data to fit your narrative by choosing to ignore any factors aside from class that may be affecting those numbers.

Looking at your chart, sure, some classes are more popular and tend to be considered stronger. No one is refuting that. The issue, however, is that you're not thinking about those factors- the popular, "stronger" classes are probably going to be played more frequently by the "pros". The "pros" are probably going to be running Sunspire more frequently for the scores. What I'm saying is that the percentage of top players playing on stamDKs or stamSorcs is going to be higher than the percentage of top players using some of the lower-DPS classes. And since these guys pull insanely high numbers, those are going to inflate the mean, and inflate it more for the classes that they're playing more frequently. That's one factor that I think may be at play, but in the end we don't know enough about the data to conclude that, which applies to your OP as well.

Also, LOL:
Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have.

I'm a statistician and I work in public health. One of my projects is working with data from children who have come into contact with the child welfare system who have been tracked over time to examine the effect of the CW system on social/emotional/behavioral outcomes. The data can be really messy sometimes, but if we sat back and said "oh, we'll work with what we have without any critical thinking whatsoever", the impact of any analyses done haphazardly could very well make a negative impact on the lives of children later on. My example is much more serious than yours but jesus, everything you've stated as a fact about data and analysis has been wildly incorrect.

I'm a PhD student. I'm well aware of how to interpret data.

Good for you. However, some of your comments relating to how you have interpreted the data are called into question. The most obvious is the quote I provided below.

.

MLGProPlayer wrote: »

We have a DPS gap between the top and bottom class that is ~50%. Balance has never been worse in this game.

Not sure what firing Wrobel accomplished.

The SS you provided in the OP do not show any class getting only 50% of the damage of the top class. You would have to use the WW parses to back up your comment here and WWs are clearly not a class.

Granted, I do agree those SS do show to much difference between the classes that hopefully Zos will learn from, but your overall analysis has flaws as has been pointed out in this thread. As a MLG Pro player I expect you can see some of these points.

72k is the average parse for stamcro. 47k is the average parse for magden/magcro. That's a 53% discrepancy.

50% =/= half (if that's what you're thinking).

kargen27

MartiniDaniels wrote: »

kargen27 wrote: »

The data means nothing if each class isn't being played by the same player. What you need is at least 100 parses of each class by the same person with the same group providing the same buffs and debuffs.
Then you need to do that several more times with different people also doing all classes with the same group.

And to get fair numbers the person doing the testing would need to be proficient with each class. You can't expect someone who spent hours and hours on a practice dummy perfecting his rotation on a DK to give you fair numbers on a nightblade they have barely played.

Well, you understand that it's not like everybody run on random toons to vSS and then this statistics appear? On the PTS those persons proficient with every class did all the measurements and posted results. Thereafter every monkey mimic'd ready-to-go builds and so in original post we can see thousands of stamcros, stamblades, magsorcs and magplars which is exact match of Liko's top classes for stamina and magicka. There are some magblades who forgot that they are together with wardens now, some magcros whose abilities are bugged and some persistent stamDK and stamplars which are in good positions but most probably carried by lokki and without it they will be weaker then 2-pet sorc with 5 buttons to press instead of 10.
Oh ok, there are some top-notch stamsorcs too, but as far as i know only hardcore fans were playing on stamsorcs and all other switched to magsorcs long ago.

Sure an exact match of the build, but an exact match of the rotation?

You also have to remember the elite players are looking for every little tweak to get just a bit more DPS out of a build. Something the majority of the games players have no interest in.

For the numbers to truly reflect what the OP claims we need to see results from players of all skill levels in a controlled group. We all know there is a best and worst class. The post is providing data and reaching a conclusion the data doesn't support. He is claiming a high degree of difference. His data though by being incomplete and faulty can't provide any real numbers to how wide the gap is.
He then takes this incomplete data and faulty conclusion and farther concludes the game is broken without defining what broken is and for who.
Sure if you want to be on the leader board and you want to do it playing a bow/bow build then the game may appear broken. Outside of leader board chasing though do these numbers mean anything?

To collect meaningful data you need control groups. This data has no controls at all.

MartiniDaniels

Runefang wrote: »

MartiniDaniels wrote: »

Runefang wrote: »

cheifsoap wrote: »

T3hasiangod wrote: »

MLGProPlayer wrote: »

@ at all the folks saying this data is misleading because it doesn't fit their narrative that the game is perfectly balanced.

Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have. This is the best we're going to have, and it's good enough to draw conclusions from. It's almost as if there is a different standard for collecting data for a dissertation and publication and analyzing video game balance.

No, this is not good enough to draw definitive conclusions. It is enough to say that this is the DPS trend for Sunspire, but not for drawing conclusions on overall DPS balance.

Good video btw, bro. Quick comment; why should anyone care about the "low end" dps for a specific class. No one cares about someone who messes up their rotation or has a laggy connection, as a matter of fact if anything it would be a bad metric to collect. IMO, you want the higher end of the dps scale because they are squeezing out the most dps they can on said trial. If ZOS were to balance with the low end in mind, you would have these uber players pulling crazy dps parses and what does that really do for "balance"? Not a whole lot

Nobody cares what suboptimal set ups (be it group, gear or rotation) pull in terms of dps. You just can’t conclude much from 59 Magden parses other than not many people who are in groups with publicly loaded logs run a Magden.

What you definitely can’t assume is that the top 1% of Magden parses (1 person) is equivalent to the top 1% of Stamblade parses (20 people). That top Magden result is highly unlikely to be representative of the class dps ceiling. Heck the Stamblade is also unlikely to be representative of the ceiling.

Aside from obviously bad data interpretation the other reason to disagree with the conclusion that the game balance is that bad is that Liko pulls numbers on all the classes which are closer to each other. At least I know he is closer to the Magden ceiling on the 21m dummy than the lone Magden parse on esologs.

Again I don’t disagree that class imbalance exists, I’m just disagreeing that it’s as bad as it’s being made out to be.

Sigh, 16k dps difference, "same numbers"

I never once said Liko pulled the same numbers on all classes. Just ones that are closer to each other. Specifically closer to each other than the stupid top 1% of the esologs. A 16k dps spread is much smaller than the 24k spread suggested with that data. It becomes smaller again when you consider the top was 103k with a Necromancer who gets major vulnerability on the dummy.

Consider the top Magden parse on Yolnahkriin at the moment. There is 15% major force uptime in that fight, 0% major vulnerability.

The top Stamblade parse on Yolnahkriin (and top overall) has a 50% major force uptime and 26% uptime on major vulnerability.

You can't tell me that's not comparing apples and oranges given those differences.

You would be right if we have different results in trials and on dummy. But since results are absolutely the same, it is clear that loki is OP, stamnecro is OP, pets are OP. Magplar with braindead rotation outperforming complex magblade is just a free joke from ZOS.

SirAndy

david_m_18b16_ESO wrote: »

Jaimeh wrote: »

You can't make any definitive conclusions for class balance, unless you have the same number of players for each class (and arguably the same players, since individual skill matters greatly) against a standardized test, like parsing on the atro dummy,

This is why you cut the lower 95% of the parse so you only take in count the top5% of players of every class. That way the player skill are about the same.
Basicaly the info provided is what poeple use to see balance in every dam MMOs. The naysayer here are probably just not used to work with such data.

Actually, i work with such data every day and you are making assumptions about the data to make it fit your personal preconceptions.

For instance, it has apparently not occurred to you that a lot of the top players are most likely trying out Necromancer builds (new class and all that) which in turn means that a lot less of the top players are currently playing other classes.

In which case the data does *not* represent a true average of top players across all classes equally, in which case your conclusions are based on a (wrong) personal bias.

MartiniDaniels

SirAndy wrote: »

david_m_18b16_ESO wrote: »

Jaimeh wrote: »

You can't make any definitive conclusions for class balance, unless you have the same number of players for each class (and arguably the same players, since individual skill matters greatly) against a standardized test, like parsing on the atro dummy,

This is why you cut the lower 95% of the parse so you only take in count the top5% of players of every class. That way the player skill are about the same.
Basicaly the info provided is what poeple use to see balance in every dam MMOs. The naysayer here are probably just not used to work with such data.

Actually, i work with such data every day and you are making assumptions about the data to make it fit your personal preconceptions.

For instance, it has apparently not occurred to you that a lot of the top players are most likely trying out Necromancer builds (new class and all that) which in turn means that a lot less of the top players are currently playing other classes.

In which case the data does *not* represent a true average of top players across all classes equally, in which case your conclusions are based on a (wrong) personal bias.

Sir, I worked with statistical data of millions measurements for several years too (mobile networks). And grim truth is that no matter how you will look at things, 16 dB, 16Mbps, 16k dps whatever is difference of huge magnitude when it is within borders of "100" which shows that one thing is notably worse then other and no matter how good you are at making reports and masking truth, report will look very bad.

idk

MLGProPlayer wrote: »

idk wrote: »

.

MLGProPlayer wrote: »

SidraWillowsky wrote: »

MLGProPlayer wrote: »

@ at all the folks saying this data is misleading because it doesn't fit their narrative that the game is perfectly balanced.

Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have. This is the best we're going to have, and it's good enough to draw conclusions from. It's almost as if there is a different standard for collecting data for a dissertation and publication and analyzing video game balance.

Have you ever in your life taken a statistics course? I'm guessing no, because if you had, you'd know about confounding variables. Let's play a game:

What's wrong with drawing definitive conclusions from the following two graphs?





No, higher ice cream sales don't cause higher murder rates or vice versa- there is a variable that accounts for both increases: season. Murder rates are higher in the summer, as are ice cream sales.

The ironic thing is that YOU are the one interpreting the data to fit your narrative by choosing to ignore any factors aside from class that may be affecting those numbers.

Looking at your chart, sure, some classes are more popular and tend to be considered stronger. No one is refuting that. The issue, however, is that you're not thinking about those factors- the popular, "stronger" classes are probably going to be played more frequently by the "pros". The "pros" are probably going to be running Sunspire more frequently for the scores. What I'm saying is that the percentage of top players playing on stamDKs or stamSorcs is going to be higher than the percentage of top players using some of the lower-DPS classes. And since these guys pull insanely high numbers, those are going to inflate the mean, and inflate it more for the classes that they're playing more frequently. That's one factor that I think may be at play, but in the end we don't know enough about the data to conclude that, which applies to your OP as well.

Also, LOL:
Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have.

I'm a statistician and I work in public health. One of my projects is working with data from children who have come into contact with the child welfare system who have been tracked over time to examine the effect of the CW system on social/emotional/behavioral outcomes. The data can be really messy sometimes, but if we sat back and said "oh, we'll work with what we have without any critical thinking whatsoever", the impact of any analyses done haphazardly could very well make a negative impact on the lives of children later on. My example is much more serious than yours but jesus, everything you've stated as a fact about data and analysis has been wildly incorrect.

I'm a PhD student. I'm well aware of how to interpret data.

Good for you. However, some of your comments relating to how you have interpreted the data are called into question. The most obvious is the quote I provided below.

.

MLGProPlayer wrote: »

We have a DPS gap between the top and bottom class that is ~50%. Balance has never been worse in this game.

Not sure what firing Wrobel accomplished.

The SS you provided in the OP do not show any class getting only 50% of the damage of the top class. You would have to use the WW parses to back up your comment here and WWs are clearly not a class.

Granted, I do agree those SS do show to much difference between the classes that hopefully Zos will learn from, but your overall analysis has flaws as has been pointed out in this thread. As a MLG Pro player I expect you can see some of these points.

72k is the average parse for stamcro. 47k is the average parse for magden/magcro. That's a 53% discrepancy.

50% =/= half (if that's what you're thinking).

47k is 65% of 72k based on the old fashion math I learned way back in elementary school. That does not look like it is even close to 53% you are suggesting unless you are massaging the numbers.

50% of 72k would be 36k 36+36=72 It seems pretty simple.

Maybe you can detail that fancy math you are learning in your PhD program so us less learned people have a chance at understanding these funny numbers you are tossing around.

Again, I am not suggesting Zos is doing a great job managing this game. I think Matt has shown a lack of leadership skills. I am just pointing out the most glaring discrepancy of your analysis which is what some are calling into question here in this thread.

DyingIsEasy

WTB SimulationCraft for ESO

LiquidPony

Vildebill wrote: »

MLGProPlayer wrote: »

Vildebill wrote: »

This thread is the perfect demonstration of how people in the world perceive facts. Extremely biased and only in the direction of the person's own interest. Source criticism seems to be the hardest thing to learn.

What's the context of the data? Who are the players? What buffs were available? How good are they at their role (and how do you even measure that?)? Consistency? When I see this graph 100 questions pop up in my mind, questions that needs answers before I can draw an even somewhat precise conclusion. Without that information, the data is worthless.

But yeah, the world is burning, everything is a mess, nothing ever gets better, it was all better before, come and vote on my extremist conservative party and all the world's troubles will be solved. It has worked perfectly fine everywhere, I promise you

It's trials data from the top 1% of performers. It's extremely telling, looking at both raw performance numbers and usage rates, that there is a huge discrepancy between classes.

Yeah probably just as telling as the doomsday Altmer nerf you so happily spread across the forums for over a month, like a prophet in black robes. And look how that turned out.

I'm sorry if I'm sounding harsh, I don't know you in real life, but by the you you always sound here on the forums, talking negatively on a lot of stuff that turns out to be not even remotely as bad, gives me a hard time to take you seriously.

LOL!

Spot on. And guess what? The top mag DPS parses right now are on ... Altmers!

Typical forum warrior nonsense. The same people (who haven't even played the content in question or the new class) overreact to bad tests and misconstrued data every patch because they lack the experience to properly contextualize the data or the wherewithal to do any meaningful testing themselves.

MLGProPlayer

idk wrote: »

MLGProPlayer wrote: »

idk wrote: »

.

MLGProPlayer wrote: »

SidraWillowsky wrote: »

MLGProPlayer wrote: »

@ at all the folks saying this data is misleading because it doesn't fit their narrative that the game is perfectly balanced.

Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have. This is the best we're going to have, and it's good enough to draw conclusions from. It's almost as if there is a different standard for collecting data for a dissertation and publication and analyzing video game balance.

Have you ever in your life taken a statistics course? I'm guessing no, because if you had, you'd know about confounding variables. Let's play a game:

What's wrong with drawing definitive conclusions from the following two graphs?





No, higher ice cream sales don't cause higher murder rates or vice versa- there is a variable that accounts for both increases: season. Murder rates are higher in the summer, as are ice cream sales.

The ironic thing is that YOU are the one interpreting the data to fit your narrative by choosing to ignore any factors aside from class that may be affecting those numbers.

Looking at your chart, sure, some classes are more popular and tend to be considered stronger. No one is refuting that. The issue, however, is that you're not thinking about those factors- the popular, "stronger" classes are probably going to be played more frequently by the "pros". The "pros" are probably going to be running Sunspire more frequently for the scores. What I'm saying is that the percentage of top players playing on stamDKs or stamSorcs is going to be higher than the percentage of top players using some of the lower-DPS classes. And since these guys pull insanely high numbers, those are going to inflate the mean, and inflate it more for the classes that they're playing more frequently. That's one factor that I think may be at play, but in the end we don't know enough about the data to conclude that, which applies to your OP as well.

Also, LOL:
Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have.

I'm a statistician and I work in public health. One of my projects is working with data from children who have come into contact with the child welfare system who have been tracked over time to examine the effect of the CW system on social/emotional/behavioral outcomes. The data can be really messy sometimes, but if we sat back and said "oh, we'll work with what we have without any critical thinking whatsoever", the impact of any analyses done haphazardly could very well make a negative impact on the lives of children later on. My example is much more serious than yours but jesus, everything you've stated as a fact about data and analysis has been wildly incorrect.

I'm a PhD student. I'm well aware of how to interpret data.

Good for you. However, some of your comments relating to how you have interpreted the data are called into question. The most obvious is the quote I provided below.

.

MLGProPlayer wrote: »

We have a DPS gap between the top and bottom class that is ~50%. Balance has never been worse in this game.

Not sure what firing Wrobel accomplished.

The SS you provided in the OP do not show any class getting only 50% of the damage of the top class. You would have to use the WW parses to back up your comment here and WWs are clearly not a class.

Granted, I do agree those SS do show to much difference between the classes that hopefully Zos will learn from, but your overall analysis has flaws as has been pointed out in this thread. As a MLG Pro player I expect you can see some of these points.

72k is the average parse for stamcro. 47k is the average parse for magden/magcro. That's a 53% discrepancy.

50% =/= half (if that's what you're thinking).

47k is 65% of 72k based on the old fashion math I learned way back in elementary school. That does not look like it is even close to 53% you are suggesting unless you are massaging the numbers.

50% of 72k would be 36k 36+36=72 It seems pretty simple.

Maybe you can detail that fancy math you are learning in your PhD program so us less learned people have a chance at understanding these funny numbers you are tossing around.

Again, I am not suggesting Zos is doing a great job managing this game. I think Matt has shown a lack of leadership skills. I am just pointing out the most glaring discrepancy of your analysis which is what some are calling into question here in this thread.

We are talking about the difference between two numbers...

Let's simplify things:

You need to increase magden/magcro DPS by 53% to get it to 72k (47k*1.53=72k). No one is saying that magden/magcro deal 50% of stamcro DPS. What's being said is that there is a 50% difference between them.

What're you're calculating is a 100% difference.

MLGProPlayer

T3hasiangod wrote: »

MartiniDaniels wrote: »

Runefang wrote: »

cheifsoap wrote: »

T3hasiangod wrote: »

MLGProPlayer wrote: »

@ at all the folks saying this data is misleading because it doesn't fit their narrative that the game is perfectly balanced.

Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have. This is the best we're going to have, and it's good enough to draw conclusions from. It's almost as if there is a different standard for collecting data for a dissertation and publication and analyzing video game balance.

No, this is not good enough to draw definitive conclusions. It is enough to say that this is the DPS trend for Sunspire, but not for drawing conclusions on overall DPS balance.

Good video btw, bro. Quick comment; why should anyone care about the "low end" dps for a specific class. No one cares about someone who messes up their rotation or has a laggy connection, as a matter of fact if anything it would be a bad metric to collect. IMO, you want the higher end of the dps scale because they are squeezing out the most dps they can on said trial. If ZOS were to balance with the low end in mind, you would have these uber players pulling crazy dps parses and what does that really do for "balance"? Not a whole lot

Nobody cares what suboptimal set ups (be it group, gear or rotation) pull in terms of dps. You just can’t conclude much from 59 Magden parses other than not many people who are in groups with publicly loaded logs run a Magden.

What you definitely can’t assume is that the top 1% of Magden parses (1 person) is equivalent to the top 1% of Stamblade parses (20 people). That top Magden result is highly unlikely to be representative of the class dps ceiling. Heck the Stamblade is also unlikely to be representative of the ceiling.

Aside from obviously bad data interpretation the other reason to disagree with the conclusion that the game balance is that bad is that Liko pulls numbers on all the classes which are closer to each other. At least I know he is closer to the Magden ceiling on the 21m dummy than the lone Magden parse on esologs.

Again I don’t disagree that class imbalance exists, I’m just disagreeing that it’s as bad as it’s being made out to be.

Sigh, 16k dps difference, "same numbers"

The difference is actually 7k for magicka DPS. For stamina DPS, using Liko's parses, the disparity is is 8k. It is not "16k dps difference" because you cannot compare magicka to stamina DPS. Is that a noticeable difference? Yes. Is it a meaningful difference? Not really, unless you're in one of the maybe half a dozen guilds that compete for world records between both PC/NA and PC/EU.

Let's assume team A goes with 8x magsorcs

Team B goes with 8x magdens

That's a 56k loss to team DPS. You're effectively losing a 9th DD by going with 8x magden. That makes a difference at any level, whether it's score pushing or progression.

Again, there is a reason why in a sample of over 8k, only 59 people played magden. If you have a fully levelled character from any other class, you are always better off playing that class, no matter what content you're doing. Progression guilds aren't in the business of wiping. If you're proficient with a magplar or magsorc, you're expected to play the magplar or magsorc. There is no reason to ever play a substandard class.

This is why Masel's suggestion to give each class unique buffs is the optimal solution of achieving class balance. ZOS is incapable of getting numbers close enough, so there needs to be a shift in focus to making every class useful in endgame content via utility..

lassitershawn

MLGProPlayer wrote: »

T3hasiangod wrote: »

MartiniDaniels wrote: »

Runefang wrote: »

cheifsoap wrote: »

T3hasiangod wrote: »

MLGProPlayer wrote: »

@ at all the folks saying this data is misleading because it doesn't fit their narrative that the game is perfectly balanced.

Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have. This is the best we're going to have, and it's good enough to draw conclusions from. It's almost as if there is a different standard for collecting data for a dissertation and publication and analyzing video game balance.

No, this is not good enough to draw definitive conclusions. It is enough to say that this is the DPS trend for Sunspire, but not for drawing conclusions on overall DPS balance.

Good video btw, bro. Quick comment; why should anyone care about the "low end" dps for a specific class. No one cares about someone who messes up their rotation or has a laggy connection, as a matter of fact if anything it would be a bad metric to collect. IMO, you want the higher end of the dps scale because they are squeezing out the most dps they can on said trial. If ZOS were to balance with the low end in mind, you would have these uber players pulling crazy dps parses and what does that really do for "balance"? Not a whole lot

Nobody cares what suboptimal set ups (be it group, gear or rotation) pull in terms of dps. You just can’t conclude much from 59 Magden parses other than not many people who are in groups with publicly loaded logs run a Magden.

What you definitely can’t assume is that the top 1% of Magden parses (1 person) is equivalent to the top 1% of Stamblade parses (20 people). That top Magden result is highly unlikely to be representative of the class dps ceiling. Heck the Stamblade is also unlikely to be representative of the ceiling.

Aside from obviously bad data interpretation the other reason to disagree with the conclusion that the game balance is that bad is that Liko pulls numbers on all the classes which are closer to each other. At least I know he is closer to the Magden ceiling on the 21m dummy than the lone Magden parse on esologs.

Again I don’t disagree that class imbalance exists, I’m just disagreeing that it’s as bad as it’s being made out to be.

Sigh, 16k dps difference, "same numbers"

The difference is actually 7k for magicka DPS. For stamina DPS, using Liko's parses, the disparity is is 8k. It is not "16k dps difference" because you cannot compare magicka to stamina DPS. Is that a noticeable difference? Yes. Is it a meaningful difference? Not really, unless you're in one of the maybe half a dozen guilds that compete for world records between both PC/NA and PC/EU.

Let's assume team A goes with 8x magsorcs

Team B goes with 8x magdens

That's a 56k loss to team DPS. You're effectively losing a 9th DD by going with 8x magden. That makes a difference at any level, whether it's score pushing or progression.

Again, there is a reason why in a sample of over 8k, only 59 people played magden. If you have a fully levelled character from any other class, you are always better off playing that class, no matter what content you're doing. Progression guilds aren't in the business of wiping. If you're proficient with a magplar or magsorc, you're expected to play the magplar or magsorc. There is no reason to ever play a substandard class.

To be fair, literally every mag spec is substandard compared to stam (esp necro/nb) rn in vSS, vHoF, vMoL, vCrags, and vArenas Also substandard in vAS for top end score pushing and maybe CR soon. :feelsbadman: Def worse for the magdens of the world than the plars and sorcs tho...

MLGProPlayer

lassitershawn wrote: »

MLGProPlayer wrote: »

T3hasiangod wrote: »

MartiniDaniels wrote: »

Runefang wrote: »

cheifsoap wrote: »

T3hasiangod wrote: »

MLGProPlayer wrote: »

@ at all the folks saying this data is misleading because it doesn't fit their narrative that the game is perfectly balanced.

Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have. This is the best we're going to have, and it's good enough to draw conclusions from. It's almost as if there is a different standard for collecting data for a dissertation and publication and analyzing video game balance.

No, this is not good enough to draw definitive conclusions. It is enough to say that this is the DPS trend for Sunspire, but not for drawing conclusions on overall DPS balance.

Good video btw, bro. Quick comment; why should anyone care about the "low end" dps for a specific class. No one cares about someone who messes up their rotation or has a laggy connection, as a matter of fact if anything it would be a bad metric to collect. IMO, you want the higher end of the dps scale because they are squeezing out the most dps they can on said trial. If ZOS were to balance with the low end in mind, you would have these uber players pulling crazy dps parses and what does that really do for "balance"? Not a whole lot

Nobody cares what suboptimal set ups (be it group, gear or rotation) pull in terms of dps. You just can’t conclude much from 59 Magden parses other than not many people who are in groups with publicly loaded logs run a Magden.

What you definitely can’t assume is that the top 1% of Magden parses (1 person) is equivalent to the top 1% of Stamblade parses (20 people). That top Magden result is highly unlikely to be representative of the class dps ceiling. Heck the Stamblade is also unlikely to be representative of the ceiling.

Aside from obviously bad data interpretation the other reason to disagree with the conclusion that the game balance is that bad is that Liko pulls numbers on all the classes which are closer to each other. At least I know he is closer to the Magden ceiling on the 21m dummy than the lone Magden parse on esologs.

Again I don’t disagree that class imbalance exists, I’m just disagreeing that it’s as bad as it’s being made out to be.

Sigh, 16k dps difference, "same numbers"

The difference is actually 7k for magicka DPS. For stamina DPS, using Liko's parses, the disparity is is 8k. It is not "16k dps difference" because you cannot compare magicka to stamina DPS. Is that a noticeable difference? Yes. Is it a meaningful difference? Not really, unless you're in one of the maybe half a dozen guilds that compete for world records between both PC/NA and PC/EU.

Let's assume team A goes with 8x magsorcs

Team B goes with 8x magdens

That's a 56k loss to team DPS. You're effectively losing a 9th DD by going with 8x magden. That makes a difference at any level, whether it's score pushing or progression.

Again, there is a reason why in a sample of over 8k, only 59 people played magden. If you have a fully levelled character from any other class, you are always better off playing that class, no matter what content you're doing. Progression guilds aren't in the business of wiping. If you're proficient with a magplar or magsorc, you're expected to play the magplar or magsorc. There is no reason to ever play a substandard class.

To be fair, literally every mag spec is substandard compared to stam(esp necro/nb) rn in vSS, vHoF, vMoL, vCrags, and vArenas rn Also substandard in vAS for top end score pushing and maybe CR soon. :feelsbadman:

I know. That was just a simplified example.

Kolzki

I can only hope that the combat team consider some of the discussion here and look at the data available to them more closely before using the log graphs for balancing.

MLGProPlayer

Kolzki wrote: »

I can only hope that the combat team consider some of the discussion here and look at the data available to them more closely before using the log graphs for balancing.

The discussion here is nothing but conjecture. Nobody has shared any competing evidence.

What we have are:

- Significant differences between the classes in solo parses
- Significant differences between the classes in group content
- Significant differences in usage rates between classes

All should be raising alarm bells in Wheeler's office (if they are doing their jobs).

lassitershawn

MLGProPlayer wrote: »

lassitershawn wrote: »

MLGProPlayer wrote: »

T3hasiangod wrote: »

MartiniDaniels wrote: »

Runefang wrote: »

cheifsoap wrote: »

T3hasiangod wrote: »

MLGProPlayer wrote: »

@ at all the folks saying this data is misleading because it doesn't fit their narrative that the game is perfectly balanced.

Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have. This is the best we're going to have, and it's good enough to draw conclusions from. It's almost as if there is a different standard for collecting data for a dissertation and publication and analyzing video game balance.

No, this is not good enough to draw definitive conclusions. It is enough to say that this is the DPS trend for Sunspire, but not for drawing conclusions on overall DPS balance.

Good video btw, bro. Quick comment; why should anyone care about the "low end" dps for a specific class. No one cares about someone who messes up their rotation or has a laggy connection, as a matter of fact if anything it would be a bad metric to collect. IMO, you want the higher end of the dps scale because they are squeezing out the most dps they can on said trial. If ZOS were to balance with the low end in mind, you would have these uber players pulling crazy dps parses and what does that really do for "balance"? Not a whole lot

Nobody cares what suboptimal set ups (be it group, gear or rotation) pull in terms of dps. You just can’t conclude much from 59 Magden parses other than not many people who are in groups with publicly loaded logs run a Magden.

What you definitely can’t assume is that the top 1% of Magden parses (1 person) is equivalent to the top 1% of Stamblade parses (20 people). That top Magden result is highly unlikely to be representative of the class dps ceiling. Heck the Stamblade is also unlikely to be representative of the ceiling.

Aside from obviously bad data interpretation the other reason to disagree with the conclusion that the game balance is that bad is that Liko pulls numbers on all the classes which are closer to each other. At least I know he is closer to the Magden ceiling on the 21m dummy than the lone Magden parse on esologs.

Again I don’t disagree that class imbalance exists, I’m just disagreeing that it’s as bad as it’s being made out to be.

Sigh, 16k dps difference, "same numbers"

The difference is actually 7k for magicka DPS. For stamina DPS, using Liko's parses, the disparity is is 8k. It is not "16k dps difference" because you cannot compare magicka to stamina DPS. Is that a noticeable difference? Yes. Is it a meaningful difference? Not really, unless you're in one of the maybe half a dozen guilds that compete for world records between both PC/NA and PC/EU.

Let's assume team A goes with 8x magsorcs

Team B goes with 8x magdens

That's a 56k loss to team DPS. You're effectively losing a 9th DD by going with 8x magden. That makes a difference at any level, whether it's score pushing or progression.

Again, there is a reason why in a sample of over 8k, only 59 people played magden. If you have a fully levelled character from any other class, you are always better off playing that class, no matter what content you're doing. Progression guilds aren't in the business of wiping. If you're proficient with a magplar or magsorc, you're expected to play the magplar or magsorc. There is no reason to ever play a substandard class.

To be fair, literally every mag spec is substandard compared to stam(esp necro/nb) rn in vSS, vHoF, vMoL, vCrags, and vArenas rn Also substandard in vAS for top end score pushing and maybe CR soon. :feelsbadman:

I know. That was just a simplified example.

Something like you said masel suggested (unique class buffs that improve total group dps) seems like it would work great and I support it but then we have ZoS here with their heads in the clouds thinking major vulnerability is a balanced unique buff -.-

T3hasiangod

MLGProPlayer wrote: »

T3hasiangod wrote: »

MartiniDaniels wrote: »

Runefang wrote: »

cheifsoap wrote: »

T3hasiangod wrote: »

MLGProPlayer wrote: »

@ at all the folks saying this data is misleading because it doesn't fit their narrative that the game is perfectly balanced.

Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have. This is the best we're going to have, and it's good enough to draw conclusions from. It's almost as if there is a different standard for collecting data for a dissertation and publication and analyzing video game balance.

No, this is not good enough to draw definitive conclusions. It is enough to say that this is the DPS trend for Sunspire, but not for drawing conclusions on overall DPS balance.

Good video btw, bro. Quick comment; why should anyone care about the "low end" dps for a specific class. No one cares about someone who messes up their rotation or has a laggy connection, as a matter of fact if anything it would be a bad metric to collect. IMO, you want the higher end of the dps scale because they are squeezing out the most dps they can on said trial. If ZOS were to balance with the low end in mind, you would have these uber players pulling crazy dps parses and what does that really do for "balance"? Not a whole lot

Nobody cares what suboptimal set ups (be it group, gear or rotation) pull in terms of dps. You just can’t conclude much from 59 Magden parses other than not many people who are in groups with publicly loaded logs run a Magden.

What you definitely can’t assume is that the top 1% of Magden parses (1 person) is equivalent to the top 1% of Stamblade parses (20 people). That top Magden result is highly unlikely to be representative of the class dps ceiling. Heck the Stamblade is also unlikely to be representative of the ceiling.

Aside from obviously bad data interpretation the other reason to disagree with the conclusion that the game balance is that bad is that Liko pulls numbers on all the classes which are closer to each other. At least I know he is closer to the Magden ceiling on the 21m dummy than the lone Magden parse on esologs.

Again I don’t disagree that class imbalance exists, I’m just disagreeing that it’s as bad as it’s being made out to be.

Sigh, 16k dps difference, "same numbers"

The difference is actually 7k for magicka DPS. For stamina DPS, using Liko's parses, the disparity is is 8k. It is not "16k dps difference" because you cannot compare magicka to stamina DPS. Is that a noticeable difference? Yes. Is it a meaningful difference? Not really, unless you're in one of the maybe half a dozen guilds that compete for world records between both PC/NA and PC/EU.

Let's assume team A goes with 8x magsorcs

Team B goes with 8x magdens

That's a 56k loss to team DPS. You're effectively losing a 9th DD by going with 8x magden. That makes a difference at any level, whether it's score pushing or progression.

Except when Team B is full of top grade magdens and when Team A is filled with only half-decent magsorcs, Team B will still mop the floor with Team A.

Emmagoldman

I think we expect too much. Its unrealistic for use to want each class to have the same dps, same ability to tank and heal. Forget balancing pvp vs. Pve, even this aspect of the pve side is enormous.

Also, do we want unique classes or a game balanced around dps parses that is completely outside of the game's context and actual play? I understand the dummy acts as a conttled variable, but mobility and range are HUGE and ignored.

idk

MLGProPlayer wrote: »

idk wrote: »

MLGProPlayer wrote: »

idk wrote: »

.

MLGProPlayer wrote: »

SidraWillowsky wrote: »

MLGProPlayer wrote: »

@ at all the folks saying this data is misleading because it doesn't fit their narrative that the game is perfectly balanced.

Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have. This is the best we're going to have, and it's good enough to draw conclusions from. It's almost as if there is a different standard for collecting data for a dissertation and publication and analyzing video game balance.

Have you ever in your life taken a statistics course? I'm guessing no, because if you had, you'd know about confounding variables. Let's play a game:

What's wrong with drawing definitive conclusions from the following two graphs?





No, higher ice cream sales don't cause higher murder rates or vice versa- there is a variable that accounts for both increases: season. Murder rates are higher in the summer, as are ice cream sales.

The ironic thing is that YOU are the one interpreting the data to fit your narrative by choosing to ignore any factors aside from class that may be affecting those numbers.

Looking at your chart, sure, some classes are more popular and tend to be considered stronger. No one is refuting that. The issue, however, is that you're not thinking about those factors- the popular, "stronger" classes are probably going to be played more frequently by the "pros". The "pros" are probably going to be running Sunspire more frequently for the scores. What I'm saying is that the percentage of top players playing on stamDKs or stamSorcs is going to be higher than the percentage of top players using some of the lower-DPS classes. And since these guys pull insanely high numbers, those are going to inflate the mean, and inflate it more for the classes that they're playing more frequently. That's one factor that I think may be at play, but in the end we don't know enough about the data to conclude that, which applies to your OP as well.

Also, LOL:
Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have.

I'm a statistician and I work in public health. One of my projects is working with data from children who have come into contact with the child welfare system who have been tracked over time to examine the effect of the CW system on social/emotional/behavioral outcomes. The data can be really messy sometimes, but if we sat back and said "oh, we'll work with what we have without any critical thinking whatsoever", the impact of any analyses done haphazardly could very well make a negative impact on the lives of children later on. My example is much more serious than yours but jesus, everything you've stated as a fact about data and analysis has been wildly incorrect.

I'm a PhD student. I'm well aware of how to interpret data.

Good for you. However, some of your comments relating to how you have interpreted the data are called into question. The most obvious is the quote I provided below.

.

MLGProPlayer wrote: »

We have a DPS gap between the top and bottom class that is ~50%. Balance has never been worse in this game.

Not sure what firing Wrobel accomplished.

The SS you provided in the OP do not show any class getting only 50% of the damage of the top class. You would have to use the WW parses to back up your comment here and WWs are clearly not a class.

Granted, I do agree those SS do show to much difference between the classes that hopefully Zos will learn from, but your overall analysis has flaws as has been pointed out in this thread. As a MLG Pro player I expect you can see some of these points.

72k is the average parse for stamcro. 47k is the average parse for magden/magcro. That's a 53% discrepancy.

50% =/= half (if that's what you're thinking).

47k is 65% of 72k based on the old fashion math I learned way back in elementary school. That does not look like it is even close to 53% you are suggesting unless you are massaging the numbers.

50% of 72k would be 36k 36+36=72 It seems pretty simple.

Maybe you can detail that fancy math you are learning in your PhD program so us less learned people have a chance at understanding these funny numbers you are tossing around.

Again, I am not suggesting Zos is doing a great job managing this game. I think Matt has shown a lack of leadership skills. I am just pointing out the most glaring discrepancy of your analysis which is what some are calling into question here in this thread.

We are talking about the difference between two numbers...

Let's simplify things:

You need to increase magden/magcro DPS by 53% to get it to 72k (47k*1.53=72k). No one is saying that magden/magcro deal 50% of stamcro DPS. What's being said is that there is a 50% difference between them.

What're you're calculating is a 100% difference.

So you are working it backwards instead of the normal accepted manner. That explains much of your analysis and why it is being called into question by so many, and rightfully so.

Kolzki

MLGProPlayer wrote: »

idk wrote: »

MLGProPlayer wrote: »

idk wrote: »

.

MLGProPlayer wrote: »

SidraWillowsky wrote: »

MLGProPlayer wrote: »

@ at all the folks saying this data is misleading because it doesn't fit their narrative that the game is perfectly balanced.

Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have. This is the best we're going to have, and it's good enough to draw conclusions from. It's almost as if there is a different standard for collecting data for a dissertation and publication and analyzing video game balance.

Have you ever in your life taken a statistics course? I'm guessing no, because if you had, you'd know about confounding variables. Let's play a game:

What's wrong with drawing definitive conclusions from the following two graphs?





No, higher ice cream sales don't cause higher murder rates or vice versa- there is a variable that accounts for both increases: season. Murder rates are higher in the summer, as are ice cream sales.

The ironic thing is that YOU are the one interpreting the data to fit your narrative by choosing to ignore any factors aside from class that may be affecting those numbers.

Looking at your chart, sure, some classes are more popular and tend to be considered stronger. No one is refuting that. The issue, however, is that you're not thinking about those factors- the popular, "stronger" classes are probably going to be played more frequently by the "pros". The "pros" are probably going to be running Sunspire more frequently for the scores. What I'm saying is that the percentage of top players playing on stamDKs or stamSorcs is going to be higher than the percentage of top players using some of the lower-DPS classes. And since these guys pull insanely high numbers, those are going to inflate the mean, and inflate it more for the classes that they're playing more frequently. That's one factor that I think may be at play, but in the end we don't know enough about the data to conclude that, which applies to your OP as well.

Also, LOL:
Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have.

I'm a statistician and I work in public health. One of my projects is working with data from children who have come into contact with the child welfare system who have been tracked over time to examine the effect of the CW system on social/emotional/behavioral outcomes. The data can be really messy sometimes, but if we sat back and said "oh, we'll work with what we have without any critical thinking whatsoever", the impact of any analyses done haphazardly could very well make a negative impact on the lives of children later on. My example is much more serious than yours but jesus, everything you've stated as a fact about data and analysis has been wildly incorrect.

I'm a PhD student. I'm well aware of how to interpret data.

Good for you. However, some of your comments relating to how you have interpreted the data are called into question. The most obvious is the quote I provided below.

.

MLGProPlayer wrote: »

We have a DPS gap between the top and bottom class that is ~50%. Balance has never been worse in this game.

Not sure what firing Wrobel accomplished.

The SS you provided in the OP do not show any class getting only 50% of the damage of the top class. You would have to use the WW parses to back up your comment here and WWs are clearly not a class.

Granted, I do agree those SS do show to much difference between the classes that hopefully Zos will learn from, but your overall analysis has flaws as has been pointed out in this thread. As a MLG Pro player I expect you can see some of these points.

72k is the average parse for stamcro. 47k is the average parse for magden/magcro. That's a 53% discrepancy.

50% =/= half (if that's what you're thinking).

47k is 65% of 72k based on the old fashion math I learned way back in elementary school. That does not look like it is even close to 53% you are suggesting unless you are massaging the numbers.

50% of 72k would be 36k 36+36=72 It seems pretty simple.

Maybe you can detail that fancy math you are learning in your PhD program so us less learned people have a chance at understanding these funny numbers you are tossing around.

Again, I am not suggesting Zos is doing a great job managing this game. I think Matt has shown a lack of leadership skills. I am just pointing out the most glaring discrepancy of your analysis which is what some are calling into question here in this thread.

We are talking about the difference between two numbers...

Let's simplify things:

You need to increase magden/magcro DPS by 53% to get it to 72k (47k*1.53=72k). No one is saying that magden/magcro deal 50% of stamcro DPS. What's being said is that there is a 50% difference between them.

What're you're calculating is a 100% difference.

I don't think that there's even any point looking at percentage differences in dps here before considering the samples. What are these players doing?

Who's playing mechanics and who's focused on parsing? What are the group buffs? Is there a reason that some classes are less effective in the one trial we have stats for? Are the stam necros in a group of 8 chaining major vulnerability? Is the 95th percentile stat for the templar werewolf with 3 parses reliable (respect to that player)? Or the 69 mag wardens? Or 113 mag necros? Would top raid groups take these builds and give them the support they need to hit their maximum possible dps?

Kolzki

MLGProPlayer wrote: »

Kolzki wrote: »

I can only hope that the combat team consider some of the discussion here and look at the data available to them more closely before using the log graphs for balancing.

The discussion here is nothing but conjecture. Nobody has shared any competing evidence.

What we have are:

- Significant differences between the classes in solo parses
- Significant differences between the classes in group content
- Significant differences in usage rates between classes

All should be raising alarm bells in Wheeler's office (if they are doing their jobs).

What we have is incomplete data.

MLGProPlayer

idk wrote: »

MLGProPlayer wrote: »

idk wrote: »

MLGProPlayer wrote: »

idk wrote: »

.

MLGProPlayer wrote: »

SidraWillowsky wrote: »

MLGProPlayer wrote: »

@ at all the folks saying this data is misleading because it doesn't fit their narrative that the game is perfectly balanced.

Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have. This is the best we're going to have, and it's good enough to draw conclusions from. It's almost as if there is a different standard for collecting data for a dissertation and publication and analyzing video game balance.

Have you ever in your life taken a statistics course? I'm guessing no, because if you had, you'd know about confounding variables. Let's play a game:

What's wrong with drawing definitive conclusions from the following two graphs?





No, higher ice cream sales don't cause higher murder rates or vice versa- there is a variable that accounts for both increases: season. Murder rates are higher in the summer, as are ice cream sales.

The ironic thing is that YOU are the one interpreting the data to fit your narrative by choosing to ignore any factors aside from class that may be affecting those numbers.

Looking at your chart, sure, some classes are more popular and tend to be considered stronger. No one is refuting that. The issue, however, is that you're not thinking about those factors- the popular, "stronger" classes are probably going to be played more frequently by the "pros". The "pros" are probably going to be running Sunspire more frequently for the scores. What I'm saying is that the percentage of top players playing on stamDKs or stamSorcs is going to be higher than the percentage of top players using some of the lower-DPS classes. And since these guys pull insanely high numbers, those are going to inflate the mean, and inflate it more for the classes that they're playing more frequently. That's one factor that I think may be at play, but in the end we don't know enough about the data to conclude that, which applies to your OP as well.

Also, LOL:
Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have.

I'm a statistician and I work in public health. One of my projects is working with data from children who have come into contact with the child welfare system who have been tracked over time to examine the effect of the CW system on social/emotional/behavioral outcomes. The data can be really messy sometimes, but if we sat back and said "oh, we'll work with what we have without any critical thinking whatsoever", the impact of any analyses done haphazardly could very well make a negative impact on the lives of children later on. My example is much more serious than yours but jesus, everything you've stated as a fact about data and analysis has been wildly incorrect.

I'm a PhD student. I'm well aware of how to interpret data.

Good for you. However, some of your comments relating to how you have interpreted the data are called into question. The most obvious is the quote I provided below.

.

MLGProPlayer wrote: »

We have a DPS gap between the top and bottom class that is ~50%. Balance has never been worse in this game.

Not sure what firing Wrobel accomplished.

The SS you provided in the OP do not show any class getting only 50% of the damage of the top class. You would have to use the WW parses to back up your comment here and WWs are clearly not a class.

Granted, I do agree those SS do show to much difference between the classes that hopefully Zos will learn from, but your overall analysis has flaws as has been pointed out in this thread. As a MLG Pro player I expect you can see some of these points.

72k is the average parse for stamcro. 47k is the average parse for magden/magcro. That's a 53% discrepancy.

50% =/= half (if that's what you're thinking).

47k is 65% of 72k based on the old fashion math I learned way back in elementary school. That does not look like it is even close to 53% you are suggesting unless you are massaging the numbers.

50% of 72k would be 36k 36+36=72 It seems pretty simple.

Maybe you can detail that fancy math you are learning in your PhD program so us less learned people have a chance at understanding these funny numbers you are tossing around.

Again, I am not suggesting Zos is doing a great job managing this game. I think Matt has shown a lack of leadership skills. I am just pointing out the most glaring discrepancy of your analysis which is what some are calling into question here in this thread.

We are talking about the difference between two numbers...

Let's simplify things:

You need to increase magden/magcro DPS by 53% to get it to 72k (47k*1.53=72k). No one is saying that magden/magcro deal 50% of stamcro DPS. What's being said is that there is a 50% difference between them.

What're you're calculating is a 100% difference.

So you are working it backwards instead of the normal accepted manner. That explains much of your analysis and why it is being called into question by so many, and rightfully so.

Dude, this is grade school math...

T3hasiangod

idk wrote: »

MLGProPlayer wrote: »

idk wrote: »

.

MLGProPlayer wrote: »

SidraWillowsky wrote: »

MLGProPlayer wrote: »

@ at all the folks saying this data is misleading because it doesn't fit their narrative that the game is perfectly balanced.

Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have. This is the best we're going to have, and it's good enough to draw conclusions from. It's almost as if there is a different standard for collecting data for a dissertation and publication and analyzing video game balance.

Have you ever in your life taken a statistics course? I'm guessing no, because if you had, you'd know about confounding variables. Let's play a game:

What's wrong with drawing definitive conclusions from the following two graphs?





No, higher ice cream sales don't cause higher murder rates or vice versa- there is a variable that accounts for both increases: season. Murder rates are higher in the summer, as are ice cream sales.

The ironic thing is that YOU are the one interpreting the data to fit your narrative by choosing to ignore any factors aside from class that may be affecting those numbers.

Looking at your chart, sure, some classes are more popular and tend to be considered stronger. No one is refuting that. The issue, however, is that you're not thinking about those factors- the popular, "stronger" classes are probably going to be played more frequently by the "pros". The "pros" are probably going to be running Sunspire more frequently for the scores. What I'm saying is that the percentage of top players playing on stamDKs or stamSorcs is going to be higher than the percentage of top players using some of the lower-DPS classes. And since these guys pull insanely high numbers, those are going to inflate the mean, and inflate it more for the classes that they're playing more frequently. That's one factor that I think may be at play, but in the end we don't know enough about the data to conclude that, which applies to your OP as well.

Also, LOL:
Lesson in data analysis: you're rarely going to find perfect data. Sometimes you need to work with what you have.

I'm a statistician and I work in public health. One of my projects is working with data from children who have come into contact with the child welfare system who have been tracked over time to examine the effect of the CW system on social/emotional/behavioral outcomes. The data can be really messy sometimes, but if we sat back and said "oh, we'll work with what we have without any critical thinking whatsoever", the impact of any analyses done haphazardly could very well make a negative impact on the lives of children later on. My example is much more serious than yours but jesus, everything you've stated as a fact about data and analysis has been wildly incorrect.

I'm a PhD student. I'm well aware of how to interpret data.

Good for you. However, some of your comments relating to how you have interpreted the data are called into question. The most obvious is the quote I provided below.

.

MLGProPlayer wrote: »

We have a DPS gap between the top and bottom class that is ~50%. Balance has never been worse in this game.

Not sure what firing Wrobel accomplished.

The SS you provided in the OP do not show any class getting only 50% of the damage of the top class. You would have to use the WW parses to back up your comment here and WWs are clearly not a class.

Granted, I do agree those SS do show to much difference between the classes that hopefully Zos will learn from, but your overall analysis has flaws as has been pointed out in this thread. As a MLG Pro player I expect you can see some of these points.

72k is the average parse for stamcro. 47k is the average parse for magden/magcro. That's a 53% discrepancy.

50% =/= half (if that's what you're thinking).

47k is 65% of 72k based on the old fashion math I learned way back in elementary school. That does not look like it is even close to 53% you are suggesting unless you are massaging the numbers.

50% of 72k would be 36k 36+36=72 It seems pretty simple.

Maybe you can detail that fancy math you are learning in your PhD program so us less learned people have a chance at understanding these funny numbers you are tossing around.

Again, I am not suggesting Zos is doing a great job managing this game. I think Matt has shown a lack of leadership skills. I am just pointing out the most glaring discrepancy of your analysis which is what some are calling into question here in this thread.

No, that's not how percentage difference calculations work.

[Final - Initial] / Initial * 100

That gives you the percentage that Final is above or below compared to Initial.

Using multiplication, a percentage increase is calculated as followed:

Value * (1 + percent/100)

So a 50 percent increase to 47k would be:

47 * (1 + 50/100)
47 * (1 + 0.5)
47 * 1.5
70.5

Working backwards to confirm...

[70.5 - 47]/47 * 100
23.5/47 * 100
0.5 * 100
50 percent

Bouldercleave

Has it occurred to anyone that perhaps the combat team isn't all that interested in complete equality of all the classes and really don't care if more people play Stamina vs Magicka?


I'm just playing devil's advocate here, but I'm not sure that anyone at ZoS REALLY cares that there are more of one class over another or if the DPS is equal across all the classes.

They mat tell you they care, but what really matters to them is the total number of players that keep logging in. As long as that number is acceptable and remains acceptable, why would they prioritize changes? If there are 1,000,000 players (just an example number) would they really care that 990,000 of them are the same?

This game is still raking in big bucks for them and may have already outlasted their expectations (MANY games don't make it anywhere near the 5 year mark). This game is still going strong and I'm not convinced that stuff like this is even on their radar anymore.

WhisperLFE

To be fair to the OP, there are seemingly dozens of moving goalposts being tossed around: look only @ top dps; discard top dps; look only at bottom dps; look at the avg player; look @ taget dummies; look @ actual content; use the same player for all parses; use different players with similar skill on the same run for all parses; etc.

And the concerns being raised are generally more methodological than statistical, since really the only statistical analysis being done is calculating the mean, but then I'm just arguing semantics.

It's interesting and sparks discussion, which is good. But as others have said, there are substantial caveats. Would be interesting to know the SDs around those numbers, for example, since just eyeball mean comparisons can be very misleading.

With large enough sample sizes, you're of course generally assuming that differences between the folks playing the classes cancel out, but that's also assuming there aren't any non-random factors at play (such as selection bias, which was already mentioned). If folks are thinking that stam classes will out-dps regular classes, the "best" players may all play stam classes, for example. So the average skill of the mag vs. stam classes is disparate. So cross-class comparisons are borked.

If the "best" players are all in stam classes, or are all in necro right now, then it might even be more informative to dump the top, say, quartile of players to potentially get more equality across each class's players.

And of course, it would be helpful to have more info across different trials.

MLGProPlayer

Bouldercleave wrote: »

Has it occurred to anyone that perhaps the combat team isn't all that interested in complete equality of all the classes and really don't care if more people play Stamina vs Magicka?


I'm just playing devil's advocate here, but I'm not sure that anyone at ZoS REALLY cares that there are more of one class over another or if the DPS is equal across all the classes.

They mat tell you they care, but what really matters to them is the total number of players that keep logging in. As long as that number is acceptable and remains acceptable, why would they prioritize changes? If there are 1,000,000 players (just an example number) would they really care that 990,000 of them are the same?

This game is still raking in big bucks for them and may have already outlasted their expectations (MANY games don't make it anywhere near the 5 year mark). This game is still going strong and I'm not convinced that stuff like this is even on their radar anymore.

They pay a team of full-time workers to work on balance. Might as well get the most bang for your buck by creating as balanced a game as possible. You're not going to lose players with more balance. You only stand to gain players/lose fewer.

MartiniDaniels

Bouldercleave wrote: »

Has it occurred to anyone that perhaps the combat team isn't all that interested in complete equality of all the classes and really don't care if more people play Stamina vs Magicka?


I'm just playing devil's advocate here, but I'm not sure that anyone at ZoS REALLY cares that there are more of one class over another or if the DPS is equal across all the classes.

They mat tell you they care, but what really matters to them is the total number of players that keep logging in. As long as that number is acceptable and remains acceptable, why would they prioritize changes? If there are 1,000,000 players (just an example number) would they really care that 990,000 of them are the same?

This game is still raking in big bucks for them and may have already outlasted their expectations (MANY games don't make it anywhere near the 5 year mark). This game is still going strong and I'm not convinced that stuff like this is even on their radar anymore.

Exactly. New combat team is 90% focused on PVP, this is clearly seen both from racial and class/weapon rebalance. From PVP perspective all those changes may be not ideal, but they make sense.
And what happens in PVE is secondary from Brian's opinion, since what's the difference if boss is killed in 5 minutes or 6 minutes, any class have enough dps potential to clear any content.