Jump to content
Target122

Massive discrepincy in Daily Wn8 Wotlabs vs Vbaddict

Recommended Posts

I have noticed several times now that Wotlabs will show my Wn8 for the day to be significantly lower than what Vbaddict Active Dossier Uploaded shows when I am playing. At first I thought I just missed some battle reports but that is now confirmed to be not the case. Does anyone know why this is happening?
 
Here is yesterday's session stats

VBADICT_zpsd7e5db6f.png
 

Here is what Wotlabs says

WOTlabs_zps2b04135e.png

Notice that the Win rate is the exact same along with the battle count, Also the Battles survived is the same as well. Even the Average Damage done is the same. As far as I can tell the only difference is that Wotlabs displays a Wn8 value which is 509 points lower. I originally thought it must be that Vbaddict had not updated to Version 15 of Wn8 but as far as my research took me they have.
 
Does anyone know why this is happening?

Share this post


Link to post
Share on other sites

If you don't have the mod for the ADU that uploads battles cached in the server then the ADU can "miss" battles since it's only pulling battles that you've actually looked at the results for, while WoTlabs pulls the data straight from the API.

Share this post


Link to post
Share on other sites

If you don't have the mod for the ADU that uploads battles cached in the server then the ADU can "miss" battles since it's only pulling battles that you've actually looked at the results for, while WoTlabs pulls the data straight from the API.

I "Looked" at all of the battles. If I understand this correctly if the ADU "Missed" a battle wouldn't it not count it for winrate battle count which were both identical?

 

The ADU calculates WN8 per battle and averages it. WoTLabs works off the totals of each stat.

Wouldn't the Average of each battle added up and averaged equal the total of everything and average of it?

If this is indeed the "problem" Then there is no way to fix this correct?

 

Which number is accurate as far as Wn8 for the session Vbaddict or Wotlabs?

Is there anyway to remove this discrepancy?

Share this post


Link to post
Share on other sites

Wouldn't the Average of each battle added up and averaged equal the total of everything and average of it?

If this is indeed the "problem" Then there is no way to fix this correct?

 

Which number is accurate as far as Wn8 for the session Vbaddict or Wotlabs?

Is there anyway to remove this discrepancy?

 

No. The function of an average is not necessarily equal to the average of a function.

Share this post


Link to post
Share on other sites

Services that use the API have no way to view separate battles, we only record your new stats and then subtract your old ones from it to see what changed, then apply the WN8 calculation over this. There is no other way unfortunately.

Share this post


Link to post
Share on other sites

So, is it like +x games in tank y, and looking at average stats change for those tanks? How many sig. digits are covered in this? 

Share this post


Link to post
Share on other sites

So, is it like +x games in tank y, and looking at average stats change for those tanks? How many sig. digits are covered in this? 

 

There are two different ways to do recent values which give you different results:

 

1. Calculate WN8 at point A and WN8 at point B. Use A*Na + C*(Nb-Na) = B*Nb to generate WN8 value.

2. Calculate B-A values for per-tank battles and the other stats. Calculate a WN8 value for the difference.

 

Option 1 is a lot better at handling the non-linearities and caps in the WN8 formula for low battle counts. It should even work for single games as long as you have double-precision subtraction. The normalization flaw may give some "interesting" results if you're playing far from your average tier though. All games are not equal.

Share this post


Link to post
Share on other sites

The ADU calculates WN8 per battle and averages it. WoTLabs works off the totals of each stat.

 

Correct.

 

As for the "missed" battles - for that case there were all uploaded, as we are speaking about the same amount of battles.

Installing BRR Battle Results Retriever inside of ADU will also help you to get all battles uploaded automatically, even without reviewing them.

Share this post


Link to post
Share on other sites

The ADU calculates WN8 per battle and averages it. WoTLabs works off the totals of each stat.

 

And here I am going to stop.

Each time this question is made, the answer relates to the fact that someone does the calculation on each match and someone else calculates the total of the session.

But if I have a liter of water in a pitcher, or a liter of water in ten glasses, why I shouldn't have in both cases a damned liter of water?

Share this post


Link to post
Share on other sites

And here I am going to stop.

Each time this question is made, the answer relates to the fact that someone does the calculation on each match and someone else calculates the total of the session.

But if I have a liter of water in a pitcher, or a liter of water in ten glasses, why I shouldn't have in both cases a damned liter of water?

 

I do think it should be consistent (as in calculating the session WN8 using the totals) so that it is compatible with the API results.

 

The reason they aren't equal, as I said before...

No. The function of an average is not necessarily equal to the average of a function.

 

This. For example, if the function f(x) = x2, and you do two "battles" where the results are 2 and 5 and you use both methods, these are the final scores:

  • vBAddict style (average of a function): (22+52)/2 = 14.5
  • WoTLabs style (function of an average): [(2+5)/2]2 = 12.25

Share this post


Link to post
Share on other sites

And here I am going to stop.

Each time this question is made, the answer relates to the fact that someone does the calculation on each match and someone else calculates the total of the session.

But if I have a liter of water in a pitcher, or a liter of water in ten glasses, why I shouldn't have in both cases a damned liter of water?

 

Because these other functions aren't the identity function. It does work for identity, which is the closest to your water example. 

Share this post


Link to post
Share on other sites

Because these other functions aren't the identity function. It does work for identity, which is the closest to your water example. 

 

It doesn't have to be the identity function to have that property. It just has to be any linear function or sum of linear functions (and all expected values must be equal or not included in the algorithm). If that was the case, WN8 scores would be the same using both methods.

Share this post


Link to post
Share on other sites

It doesn't have to be the identity function to have that property. It just has to be any linear function or sum of linear functions (and all expected values must be equal or not included in the algorithm). If that was the case, WN8 scores would be the same using both methods.

 

Yea, but my goal was to make a simple explanation for McHorn. But on that note IIRC Efficiency is a linear function, and so it prob works the same...so there is one thing going for it...

Share this post


Link to post
Share on other sites

Yea, but my goal was to make a simple explanation for McHorn. But on that note IIRC Efficiency is a linear function, and so it prob works the same...so there is one thing going for it...

 

 

I fear that I should go back to school, just to take off a little rust from my math...  :$ 

Share this post


Link to post
Share on other sites

the fact that vBaddict totally ignores your WR for the session stats (it sets WR as the expected value regardless of result iirc) but WotLabs does may have had a pretty large influence on the size of the discrepancy as well.

Share this post


Link to post
Share on other sites

Sorry to drag this back up, but I'm a bit confused on how these two sites differ so much. So maybe someone can explain in small words to me.

 

What is WotLabs doing that VB isn't when it comes to WN8 calculations? VB averages your Wn8 across battles...wotlabs is doing what exactly?

 

The reason I ask is last night I pulled 2900 WN8 according to VB, but only 1900 according to WoTL. I'm a bit confused on the math.

Share this post


Link to post
Share on other sites

Sorry to drag this back up, but I'm a bit confused on how these two sites differ so much. So maybe someone can explain in small words to me.

 

What is WotLabs doing that VB isn't when it comes to WN8 calculations? VB averages your Wn8 across battles...wotlabs is doing what exactly?

 

Difference from previous totals/tanks played in session. Look at post 11, it cannot get any simpler than bj's example there. 

Share this post


Link to post
Share on other sites

I'm not a math wiz, that's why I'm asking. Averaging is really simple for me to understand and makes sense. I simply do not understand what is being done here. I understand how it's getting there, but I"m more curious as to the why.

Share this post


Link to post
Share on other sites

One method VBaddict adds up your session results and compares it to the expected matrix for tanks played in that session. It averages the function. 

 

The other method uses the difference between your total at the start and end of session and compares it to the expected matrix for tanks played in that session. It a function of the average. 

Share this post


Link to post
Share on other sites

One method VBaddict adds up your session results and compares it to the expected matrix for tanks played in that session. It averages the function. 

 

The other method uses the difference between your total at the start and end of session and compares it to the expected matrix for tanks played in that session. It a function of the average. 

 

That makes a lot more sense than what was up above.  Thank you.

Share this post


Link to post
Share on other sites

I would give an appropriate example using the WN8 formula and a 2-battle sample, but I really have to sleep now. Shouldn't be hard for someone else to do, right?

 

Ninja'd. Bleh.

Share this post


Link to post
Share on other sites

We aims to please. 


I would give an appropriate example using the WN8 formula and a 2-battle sample, but I really have to sleep now. Shouldn't be hard for someone else to do, right?

 

Your basic algebra example is a lot easier to read though...

Share this post


Link to post
Share on other sites

I think WN9 will end up being a linear function - in this case most of the discrepancy should disappear. Otherwise there will always be issues. The only rigorous way to determine recent WN8 is to average your performance in each tank first, THEN calculate the WN8, then do a weighted average of per-tank WN8 based on relative games played. 

Share this post


Link to post
Share on other sites

I think WN9 will end up being a linear function - in this case most of the discrepancy should disappear.

 

If the player is playing the same tank, and doesn't go over a cap in one of the battles, it may be equal (I haven't done the analysis). However, if they play more than one tank, they may have very different expected values, and the rating tends towards performance in tanks with higher expected values (mainly damage, since it is the heaviest component in WN9).

Share this post


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...