Jump to content
Sign in to follow this  
Xen

Who is the Person that Comes Up with these Values?

Recommended Posts

So then where would you put it? Do you match it with the the 62A? The 140? Do you put it with the 430 since they share the same DPM with the same depression and turret weakness? Or do you raise it because it has more hp? Lower it because lolammorack? Last I checked 907 was considered better than the 430. But how much better? 10 DPG, 20, 30, 200? Who gets to decide that number, if we have nothing to support it?

This is theoretical btw, I'm not arguing, just curious what you guys would do.

 

All of that is irrelevant. You are talking about details. Whether you choose a light or a dark blue carpet does not matter, the first priority is getting rid of the huge pile of shit all over the floor.

Share this post


Link to post
Share on other sites

There's obviously a difference between the three RU meds. But there's no glaring difference that can account for a whole 134 damage difference between the T-62A and the Obj. 140. The T-62A has more health, higher traverse speeds, and a better gun whereas the Obj. 140 has a bit better speed limit, better power-to-weight ratio, and one more degree of gun depression. I really don't see how the Obj. 140 gets 134 more expected damage with those minimal differences. Matter of fact, the argument could certainly be made that the T-62A should have a higher expected damage because of the previous comparison. And, furthermore, there's no way that the Obj. 907 has about 316 more expected damage than the T-62A. The T-62A has better traverse speeds and better accuracy whereas the Obj. 907 has better health, better speeds, and a better rate of fire. Again, those differences are extremely minimal when one looks on Tank Compare. But there's no way in hell that makes the Obj. 907 have 316 more expected damage than the T-62A. This is terrible data and tons of people all over the WoTLabs and World of Tanks community have been consistently pointing out the errors in the RU medium expected values and it's just amazing that it hasn't been adjusted yet. That's what one would call being stubborn, and it's about time the flaw is sealed.

Share this post


Link to post
Share on other sites

So then where would you put it? Do you match it with the the 62A? The 140? Do you put it with the 430 since they share the same DPM with the same depression and turret weakness? Or do you raise it because it has more hp? Lower it because lolammorack? Last I checked 907 was considered better than the 430. But how much better? 10 DPG, 20, 30, 200? Who gets to decide that number, if we have nothing to support it?

This is theoretical btw, I'm not arguing, just curious what you guys would do.

The tank has very similar characteristics to the other 3. Data cannot be used because its a biased sample. Simple yet effective solution is to use the mean of the values for the other 3 tanks (140, 62A, 430.)

Share this post


Link to post
Share on other sites

So then where would you put it? Do you match it with the the 62A? The 140? Do you put it with the 430 since they share the same DPM with the same depression and turret weakness? Or do you raise it because it has more hp? Lower it because lolammorack? Last I checked 907 was considered better than the 430. But how much better? 10 DPG, 20, 30, 200? Who gets to decide that number, if we have nothing to support it?

This is theoretical btw, I'm not arguing, just curious what you guys would do.

That's honestly not relevant at all, since those are small details: If you think the tank is more farmable because unicums have it, put it on par with the one with the biggest numbers among the others until you get enough data from enough sources to make a proper rating. What matters in a skill rating is accuracy, not whether you anally follow some mathematical formula from a terrible sample size that no one dealing with statistics IRL would publish. If you have to artificially "fix" a tank or two until you get enough solid data from a varied amount of skill levels, so be it.

 

Basically, if your well is poisoned, you don't drink from it.

Share this post


Link to post
Share on other sites

That's honestly not relevant at all, since those are small details

To be honest I think this whole damn thing is irrelevant because everyone knows the small details make WN8 insufficiently accurate. Anyone who really cares that much about a couple digits of WN8 (lol) can just as easily go and see that you've been playing the 907 instead of the 62A, and get the full picture. I guess I just take it less seriously than most.

Share this post


Link to post
Share on other sites

To be honest I think this whole damn thing is irrelevant because everyone knows the small details make WN8 insufficiently accurate. Anyone who really cares that much about a couple digits of WN8 (lol) can just as easily go and see that you've been playing the 907 instead of the 62A, and get the full picture. I guess I just take it less seriously than most.

 

The difference is actually quiet significant. I guess the difference to get to my ~3k recent between T-62A and 907 is around 500 damage per game with the current expected values (just took the numbers out of my head so they might not be 100% accurate).

Share this post


Link to post
Share on other sites

To be honest I think this whole damn thing is irrelevant because everyone knows the small details make WN8 insufficiently accurate. Anyone who really cares that much about a couple digits of WN8 (lol) can just as easily go and see that you've been playing the 907 instead of the 62A, and get the full picture. I guess I just take it less seriously than most.

 

We are talking about the Obj. 907 in comparison to the T-62A having 18.5% higher expected DPB and 28% higher expected KPB. In what fucking universe is that a small detail?

Share this post


Link to post
Share on other sites

To be honest I think this whole damn thing is irrelevant because everyone knows the small details make WN8 insufficiently accurate. Anyone who really cares that much about a couple digits of WN8 (lol) can just as easily go and see that you've been playing the 907 instead of the 62A, and get the full picture. I guess I just take it less seriously than most.

Its a pretty significant difference. Also, this is more about improving the accuracy of WN8 because there's a reason people don't take it very seriously... this is an example of a reason why its a joke sometimes. 

 

People won't take something like WN8 seriously if there are blatant flaws in it, the goal here is to fix the flaws so that the rating becomes more accurate.

Share this post


Link to post
Share on other sites

May as well use this as a thread to point out wonky expected values.  A few that I can think of:

 

Should the panther II and hitler panther really have the same expected values?  I thought the hitler panther was just a weaker Panther II without the gun depression but with a little better gun stats for the most part.

 

StuG III and StuG IV should probably have different values too.  Their stats are pretty similar for the most part, but one has to deal with the L/48.  The StuG IV may have pref, but the StuG III is still a capable tank in tier VII battles as far as tier 5 vehicles go.  Still need to play my StuG IV more, but I feel like the III is generally the better tank.

 

STA-2 expected values need to drop like they're hot.  It's just a premiumified STA-1, but somehow even shittier.

 

The Fury is pretty much the same as the easy 8, but slightly worse (rip turret armor).  Fury should have lower expected values than the E8

 

ISU-130 is shit and expecting it to have a higher DPG than the ISU-152 is goddamn hilarious.  No, seriously, kick whichever idiot changed it to that because I'm laughing so hard I've dislocated my fucking sides

 

Don't know if the AT 15A needs an increase or not ever since it got that really big buff

 

The 7201K seems to suffer from the same issue the 907 does

Share this post


Link to post
Share on other sites

So then where would you put it? Do you match it with the the 62A? The 140? Do you put it with the 430 since they share the same DPM with the same depression and turret weakness? Or do you raise it because it has more hp? Lower it because lolammorack? Last I checked 907 was considered better than the 430. But how much better? 10 DPG, 20, 30, 200? Who gets to decide that number, if we have nothing to support it?

This is theoretical btw, I'm not arguing, just curious what you guys would do.

907 is technically worse than the 140/62... so shouldn't it technically have at least the same expected as them? if not lower.

 

I mean, 2K is fucking high, even though only good players play it

Share this post


Link to post
Share on other sites

907 is technically worse than the 140/62... so shouldn't it technically have at least the same expected as them? if not lower.

 

I mean, 2K is fucking high, even though only good players play it

 

And hence its higher expected values. It was given out to the at least competent players, where the 62-A/140 are played by every skill level. 

Share this post


Link to post
Share on other sites

And hence its higher expected values. It was given out to the at least competent players, where the 62-A/140 are played by every skill level. 

Why would that matter?  Good players should get good stats for playing good.  Expecting bad players to get the same WN8 for playing worse kinda ruins the purpose of it all.

Share this post


Link to post
Share on other sites

And hence its higher expected values. It was given out to the at least competent players, where the 62-A/140 are played by every skill level. 

 

The expected values are supposed to reflect the capabilities of the tanks not of the players that play them.

Share this post


Link to post
Share on other sites

And hence its higher expected values. It was given out to the at least competent players, where the 62-A/140 are played by every skill level.

the M60 has LOWER expected than the M48, and it was arguably given out to a much tighter proportion of top players only. The VK 7201 is under 100 piints higher than the E100 as well. Not enough to make a huge difference, and even then im yet to see someone under green in one. yet the 907 is still 200 points higher than the 140, and i have even seen several sub 50% players in it, i can garuntee even more bad players got it on EU/RU, where more were given out.

So again, why is it so high?

Share this post


Link to post
Share on other sites

I ve been arguing against nonsense like this for months and given up. Praetor was all for numbers + brain.exe, but currently we have blind followers of number in charge.

 

Same problem as with reward tanks and rare premiums.

Share this post


Link to post
Share on other sites

I and others tried pointing this out a while ago and what we got it a cordoned off maths section where most can no longer post even though I have post earlier in the exact same topic.

Share this post


Link to post
Share on other sites

As an owner of an Obj 907 - what did you expect?

 

Did you really think that the tank that was handed only to the 11K best CW players in the EU server out of 1,5M active players would have the same average expected dam/frag values as a normal T10 Russian med that anyone can get, regardless of skill?

 

Fair statistics are fair.

Share this post


Link to post
Share on other sites

As an owner of an Obj 907 - what did you expect?

 

Did you really think that the tank that was handed only to the 11K best CW players in the EU server out of 1,5M active players would have the same average expected dam/frag values as a normal T10 Russian med that anyone can get, regardless of skill?

 

Fair statistics are fair.

 

 

If its more or less the same tank as T62 and Obj 140, it should have the roughly the same expt. stats. What we have now in case of the 907 is as we cut of the bottom 70% of raw data (arbitrarily number) instead of the usuall 50%, which leads to higher expected values, which is unfair because the userbase for this tank isnt equal to T62, Obj 140 and 430.

Share this post


Link to post
Share on other sites

First problem:

 

WN8 should not be considered the be-all-end-all of skill evaluation. It is intended to be looked at next to win-rate, battle count, average tier, and will never replace the inspection of the Service Record or platooning with someone or watching their replays

 

One of the first lines on the WN wiki.  Despite that very clear statement, people still want to treat WN8 as an absolute.  This very thread has people calling WN8 laughable and unreliable, while at the same time fretting over details that have a minor impact on the overall rating.  

The perfect metric is solo win rate. 'Nuff said.

 

If only that figure were available...

 

Its a pretty significant difference. Also, this is more about improving the accuracy of WN8 because there's a reason people don't take it very seriously... this is an example of a reason why its a joke sometimes. 

 

People won't take something like WN8 seriously if there are blatant flaws in it, the goal here is to fix the flaws so that the rating becomes more accurate.

 

Sadly, that's an impossible task.  There will never be a single metric that accounts for the differences in performance over the life of the game, across multiple servers, without questions regarding its validity.  That's why these measures are intended to be quick references only.  They will never tell the whole story, nor are they intended to.

 

 

For the record, I actually agree with the original sentiment here.  The 907 should not be rated so far above its contemporaries.  But who decides what its value should be?  And if we let judgement influence that vehicle's number, how far along the slope do we slip?  Change one value, and people who think the values of other tanks should change will still question WN.  Change multiple tanks, and people will insist that certain values were fine, while others insist that more tanks must be altered.  Change everything, and it is no longer mathematically sound.  There will always be outliers, and there will never be a consensus.  So until someone proves my "There will never be a perfect metric" theory wrong, use WN as the quick and dirty ballpark figure it was intended as.

Share this post


Link to post
Share on other sites

Excuse me if I am being a bit blunt, Deus, but how can someone in his right mind possibly seriously suggest that whenever there is no absolutely perfect solution, sticking with the naive and evidently blatantly obviously flawed but already established solution instead of going for partial improvement is the right choice? Please, explain this totally mindbogglingly retarded logic to me. That way of thinking is essentially active resistance against any kind of progress.

 

No problem in the entire wide world will ever have an absolutely perfect solution. Perfection is what scientists strive for in everything they do, but there is always something they can improve. ALWAYS. Literally always without a single exception. Does that mean they should all just quit?

Share this post


Link to post
Share on other sites

 

 

For the record, I actually agree with the original sentiment here.  The 907 should not be rated so far above its contemporaries.  But who decides what its value should be?  And if we let judgement influence that vehicle's number, how far along the slope do we slip?  Change one value, and people who think the values of other tanks should change will still question WN.  Change multiple tanks, and people will insist that certain values were fine, while others insist that more tanks must be altered.  Change everything, and it is no longer mathematically sound.  There will always be outliers, and there will never be a consensus.  So until someone proves my "There will never be a perfect metric" theory wrong, use WN as the quick and dirty ballpark figure it was intended as.

 

The 907 is not your typical tank though, the data that is backing the expected values for the 907 is borked because the players who are playing the 907 are not your average pubshit, the players are from the upper end of the skill curve. What this means is that the data is not a sound reflection, every other tank is based on a sample of the entire wot population, while the 907 is from upper end players.

 

A tank that has so many similarities to another 3 tanks, yet has a significantly higher expected values? That is  such a blatant flaw in statistical analysis.. I actually weep for WNx.

 

I understand that metrics are far from accurate, I personally don't pay much attention to them for the most part, but a blatant mistake like that is just retarded. We can atleast try to make is slightly more accurate rather than just saying "eh fuck it, its not like the rating matters anyway its just a ballpark figure.." 

Next time we update the tables, lets just use a sample from 100 super unicums and and assume they are your typical pubshit players and create our tables off that.

 

I don't mean to be rude or anything.. I'm just getting rubbed up the wrong way by this whole thing. Obvious errors make me frustrated.

Share this post


Link to post
Share on other sites

The 907 expected values should be an average of the ones for 140, 62a and 430. Honestly there's no reason a tank's expected values should be distorted just because it was a reward for better players. You are using an extremely biased sample of the population in terms of the 907...

yeah except you know

IS-7 side armor

expected for it should be like 3k

Share this post


Link to post
Share on other sites

I take it this is yet another case of "lost in transmission" as we have seen them before (e.g. banned mod list with the minimap directions).

 

Originally, the expected values were supposed to reflect a tank's capabilities in the current meta. In order to determine the capabilities of a tank you use a method to crunch the numbers using serverwide stats. With regular tanks where the samples are consistent that is not a problem. With tanks like the Object 907 it is.

 

Now, the reason why you think this is acceptable is that you appear to have forgotten that the expected values are supposed to reflect the TANK's strength and NOT the PLAYERS' strength. You guys literally went from "doing proper maths" to "mindless number crunching for retards". This is a rather minor difference, but extremely important when looking at premium tanks, brand new copy&paste tanks and special tanks.

Share this post


Link to post
Share on other sites
Guest
This topic is now closed to further replies.
Sign in to follow this  

  • Recently Browsing   0 members

    No registered users viewing this page.

×
×
  • Create New...