Jump to content
Gryphon_

Estimation of WN8 Expected Values for Rare and New Tanks

Recommended Posts

BUt there are always exceptions ^^

 

While obvious not every exception can be adressed (far from), some of them are fairly simply to fix, like in case of fury, which is a weaker E8 anyway

 

(i understand your point well though, just like you cant make a rating system to fit tazilon hes 40k VK28 games, you also cant fix pure arty players of a person with 3000 games in m3 leee)

 

This is not about me. I was simply pointing out an example. I am not asking for special treatment. The whole discussion made clear what has been going wrong with rare tanks. This very thread was created to create the solution, not to start over discussing the validity of having the discussion in the first place. 

Xen's thread and poll made clear how the bigger part of the active wotlabs community thinks about the issue. 

It is time to draw conclusions and to take action, not to go back to square one. 

Link to post
Share on other sites

Screw that, don't cave in.  The purpose of the system isn't for the numbers to be dictated by what the community feels.  This is math, not a democracy or group therapy.  The only reason I'd see to consolidate stats is for convenience with completely identical tanks, especially those that may not see as much use anymore like the KV-220 & KV-220 Beta Test and Panzer V/IV & Panzer V/IV Alpha.  For tanks that are not identical, this shouldn't even be a discussion.

 

I look at WN8 to show how a player compares to other players driving that particular tank, the context that certain tanks have been distributed among the more talented players is already there, loud and clear, to anybody who cares enough to look at individual stats per tank.  To flex on this, imo, would weaken the value of the system as a whole significantly.  I don't really want to go with a slippery slope argument but that's pretty much what is going to happen here.  You combine two tanks where you're making exceptions for significant differences like gun depression, 100 hitpoints, different armor values & layouts, different fire chances, different turret traverse, different aim times, accuracy, reload, ammo capacity, soft stats.......where does it stop?  Do you combine the Matilda BP and the Matilda IV?  The Churchill I and Churchill III?  The T-34-85 and T-34-85M?  Unless the tanks are identical, there is no clear way to determine what stats can and cannot be different to have a minimal impact.  Combine these two tanks because x, y, & z were determined to be insignificant, and you'll end up with a dozen more threads giving reasons to combine things that are clearly not insignificant.

 

I already see people advocating that the Easy 8 and the Fury be combined because their differences are "insignificant".  Yes, the quick way to describe the Fury to anybody that plays is "its a premium Easy 8".  However, there are differences between the two tanks that change your play style options per map between the two.  Have they forgotten already that the HD modeled Fury has half the turret armor of the Easy 8?  People were pretty pissed about that when it came out.  And that combined with the Easy 8's 2 degrees more of gun depression means that the Easy 8 averages 19.10% shots without damage vs the Fury's 13.07%?  And yet, the Fury has higher Frag, Damage, and Defense expected values than the Easy 8.  Just because a higher quality player on average bought the Fury doesn't mean that the expected values need to be adjusted because some don't feel purple enough.  They are two different tanks, and when you're playing one or the other, you're competing against different standards set by a different pool of players.

 

As Gryphon said, there isn't going to be a significant impact on a player's overall anyway.  I think its crucial to the integrity of the system to keep the numbers separate, address the issue somehow in WN9, but leave WN8 working as it was intended.

 

You agree that the expected values make no sense yet you insult us for preferring common sense to senseless pseudo-mathematical purity? Are you trolling or just stupid?

 

Learn to control yourself on this subject or you will be forced to sit it out, only warning. - Solono

Link to post
Share on other sites

If you guys don't calm down I am going to get someone to nuke this entire thread. 

 

At this point substituting subjective expected values is a hotfix using scale analysis. It gets the right order of magnitude and can accurately estimate which tank values should be higher or lower than most comparable examples. However, it not a permanent fix. It will not analytically determine anything, it won't "prove" relations using objective data and it can't be used to improve the accuracy of WN8 beyond removing a source of error which is barely significant anyways. 

 

We are approaching this from the wrong angle. The current argument is to adjust the expected values of rare tanks to improve the accuracy of WN8 when applied to those rare tanks. Basically, we have fitted a curve to our data, and are moving outliers to make the fit look better. I think we need to examine why these vehicles are outliers, figure out what additional information we need to acquire to compensate for their unusual base characteristics and incorporate the results into WN8 as a whole. 

 

The fact of the matter is that every tank has a unique set of characteristics and are not sourced from homogeneous data. There is no one set of accounts which have played every tank with equal ability - we are patching together widely disparate data to get what amounts to a linear approximation of a multivariate process with an extra dimension thrown in. We are approximating the expected values for each tank, then trying to adjust them based on the assumption that everything intercepts at one nice approximation point. 

 

I don't think this is good enough. I am an Engineer - I understand the value of practical solutions. I do not consider WN8 a practical solution. It is scale analysis being passed off as analytics. It is a 20% tolerance component which is mislabeled as a 1% tolerance component which is also not even rated for the operating conditions where it is most often used. 

 

Throw me out of the discussion if you wish - I don't have time to single-handedly build this system. I want to contribute to making a rating system which works and can handle these exceptions, but until you guys understand that a subjective solution is simply not good enough I am simply unable to help you. If WN8 itself was sound I could condone it, but since WN8 itself has considerable error adjustments like these will only compound it. 

Link to post
Share on other sites

snip

In technical terms. We are putting the fire out now. Then we make sure it doesn't burn again.

 

So fix the blatant wrong values fist and then find a way to make all better, whether it's called wn8b or wn9 or wnXen.

 

 

edit:

Fixing the values is easy and can be done short term. Fixing the rest will take significantly more time.

Edited by peregrine
Link to post
Share on other sites

Would like to remind people that flinging crap like hyperactive chimps only serves to get you in trouble and make you look like a lout. Just stop doing it or people are going to get smacked hard for being children.

Link to post
Share on other sites

I don't think fixing these expected values will help much at all if we just do so by adjusting them manually. It won't change stats significantly and we can't even agree on what to adjust them to. 

 

I want to see a lot more hard data regarding tank vs tank per account before even starting to make these kind of manual changes. 

Link to post
Share on other sites

Wow, I am deeply disappointed. I don't feel you guys are actually interested in making ANY step towards us wanting to have SOME of the values changed, based on pragmatic but also REASONABLE assumptions. 

 

But whenever someone loses his cool, which I at this point can very much understand given the sheer ignorance which is presented, you are all on top of it. Ridicule is handed out and the threat to shut down the whole conversation. Unbelievably low, I really didn't expect that.

Thoughtful and viable suggestions have been made. We all know it is not a super-long-term solution. There will be a new metric at some point anyway. 

The lack of want to cooperate and let go of some statistical purity towards a practical infused understanding of the situation is quite depressing. 

The last 5 years while working on my PhD at top-notch universities I have met absolute brilliant researchers, working with scientific standards way beyond what is applied here, and none of these scholars ever showed such stubbornness and flat-out ignorance towards their peers. 

This is not meant to personally offend you guys working on metrics, I just have to reflect back to you how your behavior comes over. To a certain degree it seems absolutely intentional.

If you don't want to make a compromise and ignore the community, just say it. So we can save you the trouble and time to have a conversation at all.
 

Link to post
Share on other sites

Quite ironically, basing calculations on flawed samples is no more mathematically pure than guesstimating random subjective values. The whole point the "mathematically pure" side is trying to make is totally moot.

Link to post
Share on other sites

The point I am trying to make is that we should figure out why our bad data is causing the problems we have and work to fix the method AND the data so that the system works better. 

 

It is also a lot easier to get better data (over time we get more and more) then it is to try and work with a bad method. 

Link to post
Share on other sites

The point I am trying to make is that we should figure out why our bad data is causing the problems we have and work to fix the method AND the data so that the system works better. 

 

It is also a lot easier to get better data (over time we get more and more) then it is to try and work with a bad method. 

 

Hence the temporary solution while you genuises figure out a WORKING and NON-PSEUDO mathematical solution.

Link to post
Share on other sites

In other words, end of discussion? 

 

We are allowed to wait to the point when you guys feel comfortable to go about the issue however you feel is right with total disregard of what we proposed and discussed? 
 

Three threads full of discussion, this one even suggested by Gryphon to get into action not just talking. 

Link to post
Share on other sites

The point I am trying to make is that we should figure out why our bad data is causing the problems we have and work to fix the method AND the data so that the system works better. 

 

It is also a lot easier to get better data (over time we get more and more) then it is to try and work with a bad method. 

http://en.wikipedia.org/wiki/Selection_bias

 

No more data won't fix this. It never will.

 

Does this need further explaination?

Link to post
Share on other sites

The fact of the matter is that every tank has a unique set of characteristics and are not sourced from homogeneous data. There is no one set of accounts which have played every tank with equal ability - we are patching together widely disparate data to get what amounts to a linear approximation of a multivariate process with an extra dimension thrown in. We are approximating the expected values for each tank, then trying to adjust them based on the assumption that everything intercepts at one nice approximation point. 

 

This was my entire point bringing up the Fury vs E8 differences.  Different tank values, however slight some might consider them, produce different results, even if you were somehow able to control all other factors.  More or less gun depression would result in a bad player cresting a hill and receiving damage while a good player wouldn't crest or may choose to not even put themselves in that situation.  The differences on the Fury and the E8 (again, for example) can reward or punish in ways we can't even begin predict, therefore it comes down to establishing what is acceptable comparisons and what is not.  The debates over whether or not 2 degrees less gun depression is worth 4 degrees faster hull traverse will be endless and inconclusive in any way that can be backed up by solid numbers.  And that's just for one case.

 

 
Throw me out of the discussion if you wish - I don't have time to single-handedly build this system. I want to contribute to making a rating system which works and can handle these exceptions, but until you guys understand that a subjective solution is simply not good enough I am simply unable to help you. If WN8 itself was sound I could condone it, but since WN8 itself has considerable error adjustments like these will only compound it. 

 

I completely agree.  I don't have your same background, nor your eloquence, but this is what I was trying to get at.  The weaknesses and outright flaws of WN8 are already a known factor.  All tanks values, rare or not, have their player pools skewed to varying degrees already as certain vehicles attract or dissuade skilled players for various reasons.  I place more stock in a system with a list of "known bugs" so to speak, than one with subjectively fudged numbers in an imperfect attempt to patch up holes.  We can all at least agree on the shortcomings of WN8 as it currently stands, but introduce exceptions to the current system and you reduce the credibility of the system even further.

 

If this discussion was all in the context of building a new system, whether that be a new WNx or another rating system entirely, and examples like the Fury or the 907 were being put forth completely as "these are outliers that WN8 did not handle well, how can we fix that for New Rating System?" then I'd practice my usual lurking.  But its not, its fruitless debate over whether the new drapes look well over the giant hole in the hull.  How to best make the 907's expected values easier to swallow is not the pressing issue.

Link to post
Share on other sites

I feel like this has just been one huge waste of time, the community is suggesting a fix to an obvious error and the maths guys are being stubborn.

 

I also, for the life of me, cannot understand what you keep going on about the data for... WN8 people like to tell everyone that the metric "distorts" after a certain point, 2600 I believe, but now you are trying to calculate the base value using players who are beyond this distortion point?

 

This is the current situation:
The ship has a massive hole in it, the crew are throwing suggestions at the captain but he keeps telling them that its "not a perfect fit, find something that matches the hole perfectly or else fuck off." The crews suggestions would be able to slow the leak and give them time to find a better patch, but the captain is being too stubborn and so his ship sinks. GG WP M8.

 

Honestly, if you guys don't want to fix the problem then just tell us, instead of wasting our time. But don't expect anyone to actually take WN8 as anything more than a joke.

 

TL;DR - If you want to fix it, then lets fix it. Atm we are just wasting time and effort to pointlessly fill up threads.

Link to post
Share on other sites

you know,  if a lot of people are yelling about it, it might possible need fixing. see: arty.

 

I agree with Pity, a short term bodge of an obvious issue is far better than spending ages trying to think up a final solution. just make the 907 the same expected as the 140... 

 

also for the love of god fix the fact the STB needs to be a bloody scout tank to get anywhere near purple. 1.55 spots per game? pfft

Link to post
Share on other sites

The weaknesses and outright flaws of WN8 are already a known factor.  All tanks values, rare or not, have their player pools skewed to varying degrees already as certain vehicles attract or dissuade skilled players for various reasons.  I place more stock in a system with a list of "known bugs" so to speak, than one with subjectively fudged numbers in an imperfect attempt to patch up holes.  We can all at least agree on the shortcomings of WN8 as it currently stands, but introduce exceptions to the current system and you reduce the credibility of the system even further.

So a flawed system is better than one where many of the flaws are minimized? With all due respect, that preference chain simply doesn't make logical sense and follows a train of stubbornness. Furthermore, a patched system is a lot more credible than a flawed one because there are less errors. If we follow the logic chain that a flawed system is better than a patched one, we should go ahead and break the system even more, and that's just illogical. The ship reference Pity pulled up fits very well: would you rather sail in a boat with a big hole in the bottom or a boat with a hole patched with hardened-resin? Sure, the resin one still is a bit "ugly," but it functions a whole lot better. I'd rather have a floating boat rather than a sinking one.

Link to post
Share on other sites

 its fruitless debate over whether the new drapes look well over the giant hole in the hull.  How to best make the 907's expected values easier to swallow is not the pressing issue.

 

Here you get real. You completely IGNORE the MAJORITY of the people who posted in three different threads over the course of the last days. Yes, to many it IS a pressing issue. That is the community here! You are a FEW people acting completely ignorant about what has been discussed.

 

You put YOUR opinion of a few over the opinion of the many who contribute here to find a reasonable compromise. 

You are stalling and shutting down the conversation. Some of you guys take it personally and you let it out on the community.

 

First it was: Oh, we don't have the IDs and there is no consensus which values to take.

We provide the IDs, we have a reasonable discussion about which values to take.

Gryphon adds what would be easier to handle. We would go for that.

What happens next?

Everything we achieved so far gets completely ignored, on top highly illogical arguments are made that leave most of us baffled. 

Pitty said it, I said it too and I repeat it again:

 

If you don't give shit about the community and the effort so many put in here already to make the case, just SAY IT.

We obviously can't force you guys who are in charge of the numbers to make the suggested changes. 

I still would prefer cooperation. Just be at least honest about it.

 

Link to post
Share on other sites

So I started derping around with tank stats for the 62A, 140, 430 and 907, and a calculator:

 

I started looking at the tank stats that have an effect on the damage per game, which were: RPM, accuracy, aimtime and gun depression. We can forget damage per shot here since they're the same for all four tanks.

No, I did not include dispersion values. I don't know where I can (easily) access those and I don't know how big their effect on the DPG is.

 

First I averaged those stats for the 62A, 140 and 430:

RPM: (9.28+9.09+9.09)/3 = 9.153

Accuracy: (0.38+0.35+0.34)/3 = 0.357

Aimtime: (2.3+2.1+2)/3 = 2.133

Gun depression: (5+6+5)/3 = 5.333

 

Then I threw them together and added some weight factors since not all stats have the same effect on DPG.

9.153*A-0.357*B-2.133*C+5.333*D results in a DPG of 1778.64.

(1778.64 is the average expected damage for the 62A, 140 and 430)

 

Assigning arbitrary numbers to those factors:

A=1.8
B=1.2
C=1.1
D=0.9

(To get the proper weight factors we need somebody who actually knows what he's doing. Aka: Not me.)

 

9.153*1.8-0.357*1.2-2.133*1.1+5.333*0.9=18.5

 

Doing the same for the 907 gives us 18.48.

 

18.48*1778.64/18.5=1776.87 expected DPG for the 907

 

Doing the same with kills:

18.5 -> 0.98 expected KPG on average for the 62A, 140 and 430.
18.48 -> 0.97 expected KPG for the 907

 

 

 

Yes, I do know it has many, many flaws, but hey, it's a start?

Edited by User
Link to post
Share on other sites
 

So a flawed system is better than one where many of the flaws are minimized? With all due respect, that preference chain simply doesn't make logical sense and follows a train of stubbornness. Furthermore, a patched system is a lot more credible than a flawed one because there are less errors. If we follow the logic chain that a flawed system is better than a patched one, we should go ahead and break the system even more, and that's just illogical. The ship reference Pity pulled up fits very well: would you rather sail in a boat with a big hole in the bottom or a boat with a hole patched with hardened-resin? Sure, the resin one still is a bit "ugly," but it functions a whole lot better. I'd rather have a floating boat rather than a sinking one.

 

I would rather have foundation that, while imperfect, has non-fatal flaws, rather than attempt to add misguided repairs to it that will finally render it structurally unsound.  Its an imperfect system but its a system with clear rules and formulas, no exceptions, that apply to all.  Attempting a "repair" or creating an exception to the system is only going to give rise to people clamoring for more, with varying amounts of people unsatisfied with each and every exception made (or not made).  As Gryphon has pointed out repeatedly the effect on overall for any of these exceptions is going to be minuscule at best.  Rather than attempting an ill-advised "repair" that is only going to give rise to more and more threads like this one, where everybody pitches and argues for why their exception has to be made, and ultimately only ends up in any remaining faith in WN8 completely eroded by adding made up numbers to the mix, why not accept that WN8 has these known flaws, and focus on the development of a new system that is robust enough to deal with outliers?  It needs to be done anyway, and arguing over how to best fix WN8 is a waste of time.

 

 

Here you get real. You completely IGNORE the MAJORITY of the people who posted in three different threads over the course of the last days. Yes, to many it IS a pressing issue. That is the community here! You are a FEW people acting completely ignorant about what has been discussed.

You put YOUR opinion of a few over the opinion of the many who contribute here to find a reasonable compromise. 

You are stalling and shutting down the conversation. Some of you guys take it personally and you let it out on the community.

First it was: Oh, we don't have the IDs and there is no consensus which values to take.

We provide the IDs, we have a reasonable discussion about which values to take.

Gryphon adds what would be easier to handle. We would go for that.
What happens next?

Everything we achieved so far gets completely ignored, on top highly illogical arguments are made that leave most of us baffled. 
Pitty said it, I said it too and I repeat it again:

If you don't give shit about the community and the effort so many put in here already to make the case, just SAY IT.

We obviously can't force you guys who are in charge of the numbers to make the suggested changes. 
I still would prefer cooperation. Just be at least honest about it.

 

First, for somebody I haven't directly addressed, ever, you seem incredibly bent out of shape and threatened.

Second, you're speaking while obviously under a staggering amount of misconceptions.

-I am not involved in the extensive math hax required to create, maintain, or implement anything remotely close to the WNx systems.  The closest I've come is keeping and tracking the various expected values mainly because of my own curiosity, which admittedly I haven't even done all that well.

-Possessing a dissenting opinion != oppressing you.  Please consider this.

-Expressing a dissenting opinion != putting my opinion over anybody else's in a literal sense.  I have no power to negate anybody's opinion by expressing my own, no power other than what you're giving me in your own mind.  As far as previous discussions over the last few days, both threads were locked by the time I got there and this seems to have gone from proposal to implementation in record time, there were several well written posts, from both sides, by Mort, CrabEatOff, bjshnog,Gryphon, MaxL, Deusmortis, and others.  If anything has suppressed discussion, its a few individuals that feel that all dissenting opinions have to be swiftly and thoroughly refuted.  The topic is to discuss how to change the values Gryphon has provided, my opinion of "combine identical tanks (i.e.: KV-220 and KV-220 Beta Test), but don't change the others, focus on WN9" is just as valid as your own.  My replies to Gryphon and MaxL were directed at them and my agreement with them is just that, agreement.

-Not the entire community shares your opinion, a community is composed of many different people with differing opinions and varying abilities to effectively have a discussion.  MaxL's differing opinion does not automatically exclude him from the community, no more than your posts make you its mouthpiece.  If you're going to point at the poll as being truly representative of the community, I believe somebody linked the Wikipedia entry for Selection Bias already.  I would personally put more stock if Crab's suggestion of a committee to decide on and handle matters like this, rather than the hot headed arguments that have happened in these threads.  Several times Gryphon, Max, and others have made legitimate points and immediately get steamrolled by accusations that they're being defensive or stubborn or having ulterior motives.  That's not laying a foundation for these changes that will result in people putting more faith in the accuracy of WN8 than they already do, if anything its only going to weaken the system because it opens the door on the discussion for how biased the manipulations are.  It shows that with enough arm twisting, the data the calculations are based on doesn't matter.  Which is my entire point.  The numbers aren't perfect, but at the end of the day at least nobody can dispute that the numbers themselves don't have a personal stake in any of it.  Had the discussion been measured on both sides, I probably wouldn't have even commented.

Link to post
Share on other sites

All this reminds me of maths at school where pupils stubbornly discussed with the teacher "Why is this wrong? I used the formula I had to use!" - "Yes, but that was a rectangle, not a triangle!" - "So what?!? The formula has to work!!!"

Link to post
Share on other sites

Xeroproject, I'm done talking to you. If you can't realize what your behavior is and what you are doing, how what you wrote must come over to everybody who has been invested in this, bless your soul.

 

Regarding this topic this community has been more forthcoming and cooperative than it should be necessary. Your accusations are basically turning things around and twisting reality. Yes, there has been a heated discussion. Nevertheless we got together what was necessary to take action. Then you come along, obviously not caring about what has been already said, probably not even have read the whole discussion to start with, not caring about the polls and rambling about data and acting like a troll.

I think you are doing it on purpose and you know it. 

You are detrimental to any fruitful conversation, it's sad that your kind gets your way by just acting like a douche. 


 

Link to post
Share on other sites

I would rather have foundation that, while imperfect, has non-fatal flaws, rather than attempt to add misguided repairs to it that will finally render it structurally unsound.

 

This is getting ridiculous. There is at least one fatal flaw in wn8.

 

Someone wants to guess which it is?

1FwgKyI.png

G7OpCWH.png

 

Hints:

  This is from vbaddict. So the effect is there twice.

Edit:

added spoilers

Edited by peregrine
Link to post
Share on other sites

 

This is getting ridiculous. There is at least one fatal flaw in wn8.

 

Someone wants to guess which it is?

1FwgKyI.png

G7OpCWH.png

 

Hints:

  This is from vbaddict. So the effect is there twice.

Edit:

added spoilers

 

I appreciate what you are doing. If there was only one decent human being listening that cares. 

Link to post
Share on other sites

I'd still like to see what expected values we get if the account data for all four USSR MT Xs are pooled. Similarly for the other tightly related tank sets.

Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


  • Recently Browsing   0 members

    No registered users viewing this page.

×
×
  • Create New...