Jump to content

Recommended Posts

Its now a drag to calculate games per tier without boombox´s script... :(

Any ideas where I could get that info to copy paste?

I added that in the updated version of (Userscripts.org).

And I also included the "experimental" WN8, though, I'm not completely sure I added the most updated formula, and I'm still doing the normalization part for it.

 

 

Link to post
Share on other sites

Wow Orrie that is freakin awesome! Can you now collect data about players? Like an excel with total battles, avg frags, avg dmg, spots, eff, wn7, wn8 etc?

Also, what formula did you implement, and did you include the normalizer term and if so how did you implement that?

 

Viruzz, let me look at that and check, I tried to make it so top 1% in tier 10 gave about equal WN8 to tiers 7-10 and going down from there, but we can always tweak some kind of small penalty to frags at low tier to compensate for any differences there sitill might be, but yeah I agree we should look at that.

Link to post
Share on other sites

frag = 465*avgFrags,
dmg  = avgDmg*630/(287*Math.exp(0.203*avgTier)+25),
spot = avgSpots*125,
def  = Math.min(1.4, avgDef)*70,
win  = ((185/(0.17+Math.exp(((avgWin*100)-35)*-0.134)))-500)*0.45,
pen  = -Math.abs((((5-Math.min(avgTier,5))*125)/(1+Math.exp(avgTier-Math.pow(battles/220,3/avgTier))*1.5))),

"pen" is the low tier penalty calculation taken from WN7, its only temporary, and will be replaced with the normalizer when I manage to do that.
And -Math.abs is being used to transform pen into a negative numeric for easier display handling.

Note that I'm not taking anything at all from the API, and this only information fetched from the player page itself.

I wanted to see how far I could get before being forced to use the API, as you know, it is awful and random.

The only benefit I get from changing to the API is per-tank stat, and a non-rounded per-tank winrate.

 

I'll see if I can manage to do the normalizer before I fall asleep.

Link to post
Share on other sites

Have you had any contact with the wot-news guys on updating from wn6 to wn8 when it comes out?

 

For xvm

Link to post
Share on other sites
battlesNorm    = 1104 // Total Battles Tier 1-3
battlesNormTop = 465  // Total Battles Top 3 Tier 1-3

frag = 465*avgFrags,
dmg  = avgDmg*630/(287*Math.exp(0.203*avgTier)+25),
spot = avgSpots*125,
def  = Math.min(1.4, avgDef)*70,
win  = ((185/(0.17+Math.exp(((avgWin*100)-35)*-0.134)))-500)*0.45,
norm = -Math.abs(Math.max(0.22*(battlesNormTop-battlesNorm*83/(1+Math.pow(((0.15-0.000072)*battles),10))),0)),
wn8  = frag + dmg + spot + def + win + norm;

// Without normalizer: 1588.67
// With normalizer:    1486.37
// Normalizer diff:     102.30

// | Formula Type | Total   | Scaled | Destroyed | Damage | Detected | Capping | Defense | Victories | Tier Penalty |
// | WN7          | 1576.03 | 75.99  |   553.96  | 543.41 |  177.21  |    –    |  111.18 |   190.28  |      0       |
// | WN8pre       | 1486.37 | 70.69  |   554.20  | 589.16 |  177.21  |    –    |   77.82 |   190.28  |   -102.3     |

Oh boy, my blue color is gone. =(

 

The part where you fetch out the Tier 1-3 battles isn't that many lines, but I can't really be done in a nice formula kinda way, ( I'm probably wrong there, but I'm a scrub, soooo.. )

Link to post
Share on other sites

IMHO, Doesn´t make much sense to create something specifically for tier 10s if you need a dossier file to analyze it. If you want per-tank stats, you just use wot statistics and/or ingame E rating for tier 10 tanks. :D

 

Orrie let me know how the normalizer term comes along. If yo can get that working that would be a HUGE help! :D

Link to post
Share on other sites

I took a look at the tier 10 and tier 6 tanks ingame, XVM tracks averages and top 1% (I believe it's top 1%, it could be a different percentile), and using the top 1% numbers, it seems the top 1% scores higher WN8 at tier 6 than at tier 10.

 

On average, top 1% scores ~0,1 kills per game more than tier 10, and their damage is worth ~150 WN8 points more than the top 1% tier 10 players. spotting also tends to be higher for top tier 6 players than tier 10s

 

All in all, plotting a rough average into excel for top tier 6s and top tier 10s gives the tier 6 players ~200 WN8 more than the tier 10 players. Given that this is the same percentile of skill, shouldn't it give the same WN8 score?

I only looked at tier 10 and 6, I don't know is 5, 7, 8 and 9 paint a different picture.

 

Since everybody looks at 60 day stats, if I decided in 60 days I wanted to apply to a clan with a certain WN8 requirement, It would be easier for me to reach that requirement the more tier 6s I played, obviously they wouldn't take someone who plays exclusively 6, but if I play 30% tier 6 and 70% tier 10 it would be easier for me to reach my goal than if I played 20%/80%

 

All this of course assumes that XVM's numbers are accurate, which I can't verify since I don't know where they come from, it also assumes that the top1% at tier 6 is the same level of skill as the top 1% at tier 10, which I think is a reasonable assumption seeing as it is literally the top 1% numbers for both brackets. I hope this makes sense, even reading it back it seems a bit confusing.

 

 

Be very careful when looking at only the top 1% of a distribution.  Statistics can start to break down when you are doing such things.  Its not wrong per se--but if most people are bad at general stats, looking at particular portions need a VERY careful study.  I wouldn't be comfortable doind such a thing without some heavy research/refresher into non-parametrics and the like.

Link to post
Share on other sites

Be very careful when looking at only the top 1% of a distribution.  Statistics can start to break down when you are doing such things.  Its not wrong per se--but if most people are bad at general stats, looking at particular portions need a VERY careful study.  I wouldn't be comfortable doind such a thing without some heavy research/refresher into non-parametrics and the like.

 

I don't have the entire distribution to look at, I only have the average and the top 1%.

I use what I got!

 

One thing I should take into account that may have a large effect is popularity of tanks, I just made a very rough average counting each tank as equal, even though KV-1S is probably many times more common than a 30.01H., but I dont have anything to work with that would give me anything remotely accurate, so I can't :P

Link to post
Share on other sites

Wow Orrie that is freakin awesome! Can you now collect data about players? Like an excel with total battles, avg frags, avg dmg, spots, eff, wn7, wn8 etc?

Also, what formula did you implement, and did you include the normalizer term and if so how did you implement that?

 

Viruzz, let me look at that and check, I tried to make it so top 1% in tier 10 gave about equal WN8 to tiers 7-10 and going down from there, but we can always tweak some kind of small penalty to frags at low tier to compensate for any differences there sitill might be, but yeah I agree we should look at that.

 

After reading this thread initially I decided to work on a script in my spare time that would pull user info and output a csv file which could be loaded into Excel with most of that information (currently it doesn't calculate eff/WNx/etc... but that's easy to do in Excel once the information is in there.  Figured it would be a decent excuse to learn Python and get a little bit of experience with web APIs/JSON (I normally develop in the .NET non-web world).  It's in a very basic state right now but I've done a little testing and it is working (at least for me  8) ).  If you'd like to try it out and give any feedback/yell at me because it's broken feel free: https://www.dropbox.com/s/v5ikwgvqjxzhc7y/WoTStatGrabber.zip

 

No GUI right now, you have to run it from the command line and give it two arguments.  The first argument is a file with the names of people you want to get stats for (one name per line).  The second argument is the output file (where it will save the results), if you name it with a .csv extension Excel should automatically pull it in when you open it.  Right now the names have to be exact (though capitalization doesn't matter), I thought about a way to do a random sampling of players so you wouldn't have to supply all the names but I'm not sure if I can get the API to cooperate for that.

 

Example of calling it (first open the command line then navigate to the folder the .exe is in):

WotStatGrabber.exe c:\temp\names.txt c:\temp\stats.csv

PS: For anyone who does try this and wants to provide feedback please PM me rather than cluttering up Praetor's thread.  :thumbup:

Link to post
Share on other sites

That is pretty awesome hibachi, I´ll let you know how it works. One key thing we would need the script to do in order to be of help for furtehr developing WN8 is that it would need to be able to parse the tier and the number of games played in the 5 most played tier 1-3 tanks.

Link to post
Share on other sites

It's currently calculating average tier and the # of games played in the 3 most played tier 1-3 tanks.  I'll update that this weekend (maybe I'll have it take the number of tanks as an optional parameter).  Alternatively if you're familiar with Python and want to change it yourself you can pull the source from here (I need to stop being lazy and fill out that empty README file, lol): https://github.com/AltairDusk/WoTStatGrabber/tree/develop

Link to post
Share on other sites

I'll add a "dev" function where you can add your own custom formula for the script, that you can modify yourself on the go.

And the normalizer should be in 0.8.6.18 calculated as WN8pre+norm, and ignores currently only the top 3 low-tier tanks.

WN8pre does not have the normalizer.

Link to post
Share on other sites

I added that in the updated version of (Userscripts.org).

And I also included the "experimental" WN8, though, I'm not completely sure I added the most updated formula, and I'm still doing the normalization part for it.

I tried running this and get the error "The program can't start because MSVCR100.dll is missing from your computer. Try reinstalling the program to fix this problem."

 

That file is present on my system, so I just copied it to the same folder where I'm running the executable from and now get this error. "The application was unable to start correctly (0xc000007b). Click OK to close the application."

 

Any thoughts?

Link to post
Share on other sites

I was just curious as ITDude posted  on the WGForum how to see WN8pre and WN8pre+norm on our WoT Profiles.  WN8Pre looks cool, it takes me up north of 1300.  WN8pre+norm almost drops me below 1000.  I had no idea that enjoying a healthy mix of tanks from tier 1-10 would be penalized that heavily.  After seeing that, I really am starting to doubt the "+norm" as a good metric, as it punishes those of us who do not play exlusively in tier 8-10 matches.  

 

I am still curious as to why play in mid-lower tiers is punished so heavily in the +norm formula.

Link to post
Share on other sites

Orrie and Hibachi, you guys are awesome. Give me some time to play around with these two new toys. :)

 

Orrie that dev function should be awesome, since I think there is some mixup in the normalizer formula, results are not matching my excel, and it seems everyone I checked is penalized, which shouldnt happen.

Hibachi, I know more PERL than Python, but I will give it a whirl. :D

 

Privpub, I think the norm formula is botched somewhere, many people who arent supposed to be penalized are being hit hard. Patience. Also, measuring whether a formula is good or bad based on how your rating changes is a bit biased ;)

 

Ruestir, post that here:

 

Link to post
Share on other sites

Can you give me a few examples of expected results with and without the normalizer, and I'll see what I did wrong in the formula. ( Though, I already think I know why it displays a different number. )
 
EDIT:
Here is the style for the custom formula:
3GMPt.jpg

 

Just have to finish up the saving part, and It'll be good to go.

 

Anything more needed?

Link to post
Share on other sites
Privpub, I think the norm formula is botched somewhere, many people who arent supposed to be penalized are being hit hard.

 

I am one of those that has been hit HARD. Probably the hardest. But while I play many games in Tier1-3 I also play many games in ALL tiers.

LGpPMK0.png

 

Shouldn't the formala take that into account?

 

Link to post
Share on other sites

 

 

Privpub, I think the norm formula is botched somewhere, many people who arent supposed to be penalized are being hit hard. Patience. Also, measuring whether a formula is good or bad based on how your rating changes is a bit biased ;)

 

Obvioiusly I'm biased towards wanting a better internet-points-so-my-epeen-is-larger-score!  

 

But still, while I knew that the WN8+norm was gonna ding people who played 'too many' low tiers *aka abusing seals*, I had no idea that it would hit people like me who have a lot of games played between tiers 3 and 6 because I found tanks of those tiers that are extremely fun to play every now and then as a break from grinding up new lines.

 

Honestly, comparing my battles to ITDude's, I have an almost identical breakdown of my tiering and amount played (because I have at least 1 tank of every tier and enjoy playing every one).  I'm probably getting hit by the same formula snafu that is hitting him.

Link to post
Share on other sites

What about a tier specific WN7.  Something like this:

 

High Tier only:  Tiers 8-10, calculated normally

Tier 10 only: 

 

This way you could at a glance screen out the jsnazz padders and focus on what matters most -- tanks that are used in or similar to in playstyle tanks that are used in CW.

 

Another option might be:  damage whore quotient -- focus ONLY on kpg, dpg, and normalized exp/game (assume prem).  Win rate, capping, and spotting are ignored.  The goal would be comparing damage whore output.

Link to post
Share on other sites

What about a tier specific WN7.  Something like this:

 

 

 

Learn2Read.

Link to post
Share on other sites

Im in the lab atm, will get back to you on that tommorrow Orrie. 

 

Guys don`t panic. Let Orrie and I iron out the problems with the formula. THEN you can complain to me about your low WN8 scores. :D :D :D

 

Garbad, don`t quite get what you mean, if you mean use only tier 10 tank stats, not possible ATM due to API problems.

Link to post
Share on other sites

if you mean use only tier 10 tank stats, not possible ATM due to API problems.

 

Which poor preator has only had to say about.... oh .. 50 times.

Link to post
Share on other sites

Im in the lab atm, will get back to you on that tommorrow Orrie. 

 

Guys don`t panic. Let Orrie and I iron out the problems with the formula. THEN you can complain to me about your low WN8 scores. :D :D :D

 

Garbad, don`t quite get what you mean, if you mean use only tier 10 tank stats, not possible ATM due to API problems.

 

No!  The internet demands your capitulation on this as of yesterday and this fixed as of 30 years ago!

 

But seriously, take your time and do it right.  Anything worth doing is worth taking the time to do it properly.

Link to post
Share on other sites

 

Hey Praetor :) I was wondering if this excel spreadsheet uses the up to date WN8 formula. I was wonder because I just filled it out (thank you for having me as one of the default tabs it made filling it out much quicker ;).... anyways..  it says my penalty was ~-125 but this thread: http://forum.worldoftanks.com/index.php?/topic/259618-wn8-work-in-progress/page__st__120#top the person that checked my WN8 penalty was around around -285.

 

Anyway have a good one and keep up the good work  :thumbup:

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...