Jump to content


[WN7] What is it and how does it work?

statistics stats noobmeter performance efficiency metric compare quality rating Garbad

  • Please log in to reply
2483 replies to this topic

BanzaiBonsai #1161 Posted Jan 25 2013 - 00:28

    Captain

  • Players
  • 6125 battles
  • 1,724
  • Member since:
    08-11-2012

View PostNoobMeter, on Jan 25 2013 - 00:23, said:

"he just has higher global stats than you".

That is the part that threw me off, combined with the rest of the sentence

View PostPraetor77, on Jan 24 2013 - 23:53, said:

but the bashing for playing low tiers bring him to the same WN6 level as you (7 points WN difference is insignificant).

The only part that is true is that 7 points are insignificant, but in this case it shouldn't be 7 points.
And I'm not only talking about me here, but about all others with similar WN6 scores as well.

tHebUm #1162 Posted Jan 25 2013 - 00:34

    Corporal

  • Players
  • 10929 battles
  • 30
  • Member since:
    04-12-2011

View PostNavySnipers, on Jan 25 2013 - 00:28, said:

I'm confused. I ran across this today:

http://wotlabs.net/s...orceUpdate=true

Posted Image

If his stats are better than mine in almost every way, how is my WN6 rating higher?

They aren't, your overall WN6 is ~450 less.  The WN6 is better for the last sixty days, which means you've played similarly to his level for the last sixty days (by WN6's formula).

Edited by tHebUm, Jan 25 2013 - 00:37.


TheM0untaineer #1163 Posted Jan 25 2013 - 00:35

    Major

  • Players
  • 21670 battles
  • 11,908
  • Member since:
    10-18-2011

View PosttHebUm, on Jan 25 2013 - 00:34, said:

They aren't, your overall WN6 is ~450 less.  They're better for the last sixty days, which means you've played similarly to his level for the last sixty days.

I meant to say 60 day.

Serbs_Diaper_Fetish #1164 Posted Jan 25 2013 - 00:41

    Major

  • Players
  • 20135 battles
  • 5,753
  • Member since:
    08-26-2011

View PostNavySnipers, on Jan 25 2013 - 00:28, said:

I'm confused. I ran across this today:

http://wotlabs.net/s...orceUpdate=true

Posted Image

If his stats are better than mine in almost every way, how is my 60 day WN6 rating higher?

His stats are marginally better. He normalized xp per game 495 which is pretty close to yours. I remember the older versions of WN dont really discriminate much between a 60%er and a 70%er (something I feel is a flaw but that is for another discussion). You have similar damage per game and you may be playing different average tiers for the last 60 days and K/D has a lot to do with survival rate and may be indicative of playstyle and also have nothing to do with WN6. So essentially you have pretty similar stats (as far as what WN tracks) and 30 points isnt much of a gap.

Edited by JohnGaltCobraCommander, Jan 25 2013 - 00:41.


NoobMeter #1165 Posted Jan 25 2013 - 00:50

    Staff sergeant

  • Players
  • 9 battles
  • 441
  • Member since:
    11-14-2012

View PosttHebUm, on Jan 25 2013 - 00:27, said:

Not sure if this is the ideal place to post this, but I know the Noobmeter person and I think the WoTLabs person both read this so I figured I'd post it here.

The NoobMeter person thinks the NoobMeter calculation of 1964 is correct :)


TheM0untaineer #1166 Posted Jan 25 2013 - 01:08

    Major

  • Players
  • 21670 battles
  • 11,908
  • Member since:
    10-18-2011

View PostJohnGaltCobraCommander, on Jan 25 2013 - 00:41, said:

His stats are marginally better. He normalized xp per game 495 which is pretty close to yours. I remember the older versions of WN dont really discriminate much between a 60%er and a 70%er (something I feel is a flaw but that is for another discussion). You have similar damage per game and you may be playing different average tiers for the last 60 days and K/D has a lot to do with survival rate and may be indicative of playstyle and also have nothing to do with WN6. So essentially you have pretty similar stats (as far as what WN tracks) and 30 points isnt much of a gap.

Thanks. I forgot about normalized exp.

Tank_Killer1986 #1167 Posted Jan 25 2013 - 01:17

    Staff sergeant

  • Players
  • 68 battles
  • 344
  • Member since:
    12-23-2012
I like my 60 day WN6 score of 2930, sadly my real rating is 2264.

Boom_Box #1168 Posted Jan 25 2013 - 01:25

    First lieutenant

  • Beta Testers
  • 11752 battles
  • 905
  • [REL-V] REL-V
  • Member since:
    11-17-2010
Got a question for Noobmeter and other hosts who perhaps utilize per tank data.

After comparing WG API against in-game data, it is apparent that M103 and Marder II data is incorrect in the API. That goes for many tanks in the API but I'll just use these two as an example.

Meaning, the only way for a web site to get correct data is if the user uploads their dossier cache.

One question I have is this: if any of those sites perform analysis on dossier data (specifically, per tank performance) and calculate a score based on that, how can they score the user base for those who do not upload their dossier?

Implied: if NM PR is looking at particular tanks with its 'algorithms' in order to base an analysis, does it know that a) API is incorrect, b) does it use dossier data from other sites, c) why would a particular vehicle (or set of vehicles) be sufficient to base that analysis on, and d) why doesn't it provide sources so it can be certified?

I'm just curious.

VaporGator #1169 Posted Jan 25 2013 - 01:32

    First lieutenant

  • Players
  • 9786 battles
  • 809
  • [5PANZ] 5PANZ
  • Member since:
    09-19-2011
Also Boom Box, the tools like Active Dossier can only upload the results if you look at the final battle results screen.  So people can easily "ignore" the loses and only open the wins so the Active Dossier client uploads only the good stuff.  I try to upload all of mine, win or loss, so I can use it to see my performance.  I am sure I miss some that finish while I am in another game playing a different tank.  I think the dossier thing is a no go unless there is a way to automatically read the battle reports... but apparently there is not.

Neverwish #1170 Posted Jan 25 2013 - 01:37

    Captain

  • Players
  • 8637 battles
  • 1,979
  • [LABS] LABS
  • Member since:
    06-10-2012

View PosttHebUm, on Jan 25 2013 - 00:27, said:

Not sure if this is the ideal place to post this, but I know the Noobmeter person and I think the WoTLabs person both read this so I figured I'd post it here.

The WoTLabs person thinks the WoTLabs calculation of 1510 is correct.

I made a script where you can check your WN6 score in detail including all the breakdowns along with the formula:
http://wotlabs.net/t...name=tHebUm&d=1

You can change the d parameter for any number of days you wish to grab your stats and calculate your WN6.

tHebUm #1171 Posted Jan 25 2013 - 02:17

    Corporal

  • Players
  • 10929 battles
  • 30
  • Member since:
    04-12-2011

View PostNeverwish, on Jan 25 2013 - 01:37, said:

The WoTLabs person thinks the WoTLabs calculation of 1510 is correct.

I made a script where you can check your WN6 score in detail including all the breakdowns along with the formula:
http://wotlabs.net/t...name=tHebUm&d=1

You can change the d parameter for any number of days you wish to grab your stats and calculate your WN6.

It was 1517, not 1510 (rounding error though, maybe).  Also, I thought the main change between WN4 and WN6 was targeted at dropping seal clubbers' WN rating.  The WN4 for the same session (showing on Noobmeter) is 1958 and, as stated previously, the average tier was 9.74 so they should be fairly similiar (and the Noobmeter WN6 of 1964 is quite so) since that's about as far from seal clubbing as you can get.

Also, if it helps, I am pretty sure Noobmeter and WoTLabs agreed fully with WN4 so the WN4 for the session should be correct.

Either way though, I'm going to start playing for the night and that'll throw it off.  Mission accomplished of informing both parties of the issue, I'll leave it to you guys to get to the bottom of it.

BeLiAL124 #1172 Posted Jan 25 2013 - 06:00

    First lieutenant

  • Players
  • 26950 battles
  • 600
  • [-ZOO-] -ZOO-
  • Member since:
    08-28-2011
So basically WN6 is in line with efficiency being easy to manipulate.

Best way to tell how good someone plays is to run with them.

Otter_von_Bismarck #1173 Posted Jan 25 2013 - 07:43

    Captain

  • Players
  • 23096 battles
  • 1,457
  • [OTTER] OTTER
  • Member since:
    02-18-2011
What's the Super Unicum Winrate Bracket? 65%+ or 70%+?

NoobMeter #1174 Posted Jan 25 2013 - 10:23

    Staff sergeant

  • Players
  • 9 battles
  • 441
  • Member since:
    11-14-2012

View PostBoom_Box, on Jan 25 2013 - 01:25, said:

Got a question for Noobmeter and other hosts who perhaps utilize per tank data.
I will answer even though depending on the definition of "per tank data" I don't really fall into the addressee set.

Quote

One question I have is this: if any of those sites perform analysis on dossier data (specifically, per tank performance) and calculate a score based on that, how can they score the user base for those who do not upload their dossier?
No one uploads dossiers to NoobMeter so that's easy - I calculate it for the empty set who upload their dossier in the same way as for everyone else who doesn't :).

Quote

Implied: if NM PR is looking at particular tanks with its 'algorithms' in order to base an analysis, does it know that a) API is incorrect,
Yes, it doesn't use the per tank damage / etc. data in the API for which WG for some reason returns incorrect results.

If I would, it would return pretty inaccurate results - which it doesn't.

Quote

b) does it use dossier data from other sites,
No, that would be a hassle.

I'm toying around with the idea of letting people upload profile data, but again since I would only get some people to do it it seems a feature that isn't that useful.

Quote

c) why would a particular vehicle (or set of vehicles) be sufficient to base that analysis on, and
This is a question which is based on you jumping too far to incorrect conclusions before asking it.

Quote

d) why doesn't it provide sources so it can be certified?

"Certified" is a strong word. I didn't realize you were certifying rating sites now.

We could also ask why WotLabs code doesn't remain open-source considering his code is based on the open source tanks of science code (with an unknown license) - but we won't, as who cares.

We could ask why wot-news.com didn't initially open the code for their ER formula and people had to reverse engineer it, or why they don't open the code for their "new ER" formula - but we won't, as who cares.

We could ask why the latest version of Google PageRank algorithm isn't available freely to download on code.google.com - but we won't, as it is somewhat obvious.

But in case of NoobMeter.com PR it is of course "zomg he does not satisfy our curiosity, bad NoobMeter person". But I've listed the reasons a couple of times at the NoobMeter thread, we can take this discussion there.

In addition, pointing out where ratings return seemingly incorrect results can easily be done without knowing the actual formulas/algorithms used. As an example, see the multiple examples of ER malfunctioning, or the 8 examples that I posted regarding WN6 (which were summarily dismissed as "small differences" despite 500 point spreads, and it was explained that WN6 cannot fix them as it only uses "global rankings").

Boom_Box #1175 Posted Jan 25 2013 - 10:43

    First lieutenant

  • Beta Testers
  • 11752 battles
  • 905
  • [REL-V] REL-V
  • Member since:
    11-17-2010
Yes, certified was a bit strong. I have no interest on validating the work of other folks and besides, I've seen some of your code which leads me to believe that you do know what you're doing. The questions arose from your own and being an old school programmer myself, just got a bit curious on how you arrived at some rankings.

I will drop the inquisition right here and continue to show NM PR in my script, if that is acceptable to you, and even provide a convenient link for users so they can visit your site with a single click.

Cheers. [needs beer smiley]

Boom_Box #1176 Posted Jan 25 2013 - 10:48

    First lieutenant

  • Beta Testers
  • 11752 battles
  • 905
  • [REL-V] REL-V
  • Member since:
    11-17-2010

View Post25B5VF7, on Jan 23 2013 - 14:12, said:

FWIW, I'm using the 0.7.5.58 script on Waterfox 16.0.1 and I can see the label Performance Rating Calculations, however, I don't see any data.  If I turn the medals on, I loose the column headings, too.

Cheers!!

You found a race condition of sorts. Medals could be populating before that stat table. That should be fixed, now page requests aren't performed until after the stat table has been completed.

NoobMeter #1177 Posted Jan 25 2013 - 10:53

    Staff sergeant

  • Players
  • 9 battles
  • 441
  • Member since:
    11-14-2012

View PostBoom_Box, on Jan 25 2013 - 10:43, said:

Yes, certified was a bit strong. I have no interest on validating the work of other folks and besides, I've seen some of your code which leads me to believe that you do know what you're doing. The questions arose from your own and being an old school programmer myself, just got a bit curious on how you arrived at some rankings.

I will drop the inquisition right here and continue to show NM PR in my script, if that is acceptable to you, and even provide a convenient link for users so they can visit your site with a single click.

Cheers. [needs beer smiley]

Sorry, Boom_box, I just get a bit bothered whenever I feel there is a double standard applied to NoobMeter.com vs the rest of the world, whether in regards to "show us all your code" or "we expect you to run a website which costs $150 /month in hosting, but don't you dare try to recoup some of it by showing ads".

What code of mine have you seen? The JS that draws graphs + all the HTML? I must then say that HTML+JS is not my strong suit, I am mostly a server-side programmer.

Boom_Box #1178 Posted Jan 25 2013 - 11:16

    First lieutenant

  • Beta Testers
  • 11752 battles
  • 905
  • [REL-V] REL-V
  • Member since:
    11-17-2010

View PostNoobMeter, on Jan 25 2013 - 10:53, said:

Sorry, Boom_box, I just get a bit bothered whenever I feel there is a double standard applied to NoobMeter.com vs the rest of the world, whether in regards to "show us all your code" or "we expect you to run a website which costs $150 /month in hosting, but don't you dare try to recoup some of it by showing ads".

What code of mine have you seen? The JS that draws graphs + all the HTML? I must then say that HTML+JS is not my strong suit, I am mostly a server-side programmer.

Looked at the html and a little of how you implemented the visualization stuff. Not much the client can see to be honest. The fact that you capitalize on these types of libraries is pretty cool. Something I might look into one of these days but I'm not much of a web developer. Had to stare at the html though before coming up with a quick pattern for PR match.

If you break it tomorrow I'd understand lol. For a real fun time go look at some russian html. As for ads etc, personally I do not care one bit how you run your site. Hopefully it pays the bills and I'd happy to link you in my greasy little script.

Praetor77 #1179 Posted Jan 25 2013 - 12:52

    Captain

  • Players
  • 29723 battles
  • 1,675
  • Member since:
    05-01-2011
Banzai, I don´t understand why you feel insulted... it´s the truth he has higher global stats...

BB
Damage per battle: 1822.41
Frags per battle: 1.75
Spotted per battle: 2.25
Defense per battle: 1.3
Average Tier: 7.25

Zeal
Damage per battle: 995.19
Frags per battle: 2.18
Spotted per battle: 1.92
Defense per battle: 2.3
Average Tier: 4.63


I´m not saying he is a better player than you, I am simply explaining why he gets the rating he gets. Since WN only takes into account global stats, there is a limit to how precisely it can measure a player´s skill. Like I said, playing with OP tanks a lot will artificially inflate your WN if you play enough matches with it.

In the case of Zealpath, he farmed a lot of stats with his Marder, not so much with his Hotchkiss. He has then played  some games with high tier tanks raising his average tier, but not enough to make your better performance with high level tanks make a difference in global stats. Nevertheless, you should keep in mind you are both unusual accounts, outliers if you will, and therefore hard to measure accurately by a global account method like WN.

Edited by Praetor77, Jan 25 2013 - 13:05.


VaporGator #1180 Posted Jan 25 2013 - 15:21

    First lieutenant

  • Players
  • 9786 battles
  • 809
  • [5PANZ] 5PANZ
  • Member since:
    09-19-2011

View PostPraetor77, on Jan 25 2013 - 12:52, said:

Banzai, I don´t understand why you feel insulted... it´s the truth he has higher global stats...

BB
Damage per battle: 1822.41
Frags per battle: 1.75
Spotted per battle: 2.25
Defense per battle: 1.3
Average Tier: 7.25

Zeal
Damage per battle: 995.19
Frags per battle: 2.18
Spotted per battle: 1.92
Defense per battle: 2.3
Average Tier: 4.63


I´m not saying he is a better player than you, I am simply explaining why he gets the rating he gets. Since WN only takes into account global stats, there is a limit to how precisely it can measure a player´s skill. Like I said, playing with OP tanks a lot will artificially inflate your WN if you play enough matches with it.

In the case of Zealpath, he farmed a lot of stats with his Marder, not so much with his Hotchkiss. He has then played  some games with high tier tanks raising his average tier, but not enough to make your better performance with high level tanks make a difference in global stats. Nevertheless, you should keep in mind you are both unusual accounts, outliers if you will, and therefore hard to measure accurately by a global account method like WN.

This is where we need context to ratings Praetor.   Like the issue with VE where I started looking at charts per tier, if we can come up with a way to put the player's scores in context to their average tier, it would make everything better.  Players would be compared to players of the same tier, so the scores are relative to that context...  Zeal and BB in this case would not be in the same tier, so there would be no arguing who is better and the validity of the score remains.   It is how we interpret the score that is the issue.





Also tagged with statistics, stats, noobmeter, performance, efficiency, metric, compare, quality, rating, Garbad

2 user(s) are reading this topic

0 members, 2 guests, 0 anonymous users