Jump to content


Gross Credit Calculation

credit calculation metrics

  • Please log in to reply
13 replies to this topic

Ajeroth #1 Posted 16 March 2016 - 06:05 PM

    Private

  • Players
  • 3360 battles
  • 6
  • [DMT]
  • Member since:
    12-21-2015

Hello All,

 

I have been working on a pet project to approximate the gross credit calculation, and I am getting close . . .

 

First off, I have done my homework and I am aware of http://wiki.wargaming.net/en/Battle_Mechanics#Tank_Experience_and_Credits, as well as the numerous threads that, bluntly, have misleading or incorrect information.  Once you do the math, you quickly understand the misconceptions most have.

 

I have been able to isolate and determine most of the independent factors that contribute to the gross credit calculation as follows:

 

  1. Tank Tier
  2. Unique tank factor (Used to derive the credit earning coefficients from http://forum.wotblitz.com/index.php?/topic/29676-credit-earnings-for-all-tanks-tier-5)
  3. Damage
  4. Assist Damage
  5. Spotting
  6. Cap point contribution of win by cap.
  7. Victory factor

 

My derived algorithm works MOST of the time.  There are instances where the actual credit earnings is off by greater than the +/- 0.1% I usually see due to rounding.  The outliers are on the order of +/- 33-50% error.  I have explored some medals, time in match before death, distance traveled, and I can't isolate the rogue stat.  My guess is its a medal, or some step function if you team does really well or something. I recently had a match that I was awarded 47% below my calculated earnings.  It was a victory, damage dealt, and no medal.  It was so bad that the gross credit I was awarded was under what I would have got if I joined and went afk.

 

Does anybody have any ideas?

 

Edit:  I will expand the testing to all of the 1-4 non-premium tanks, and any 5+ I have.  Any outliers I'll screenshot all details.  You should have heard the flames deriding my weird antics in battles to isolate the metrics hehe.

 

Edit:  I have attached my testing of Tier 1, I will continue with T2 then probably just test the tanks I'm currently driving . . . Getting 10 rounds of each T1 felt like shoving bamboo under my nails.  Have I said how much I hate the Vickers Medium 1?

Spoiler
Spoiler

 


Edited by Ajeroth, 17 March 2016 - 10:00 PM.


Archie761 #2 Posted 16 March 2016 - 06:39 PM

    Senior Sergeant

  • Players
  • 28015 battles
  • 953
  • [SPUDX]
  • Member since:
    10-19-2014

I never made it very far when I was looking at the calculation but I am really interested in seeing what you have come up with. 

How often are you getting outliers in you current calc and what is the variability in % error? That might help in at least trying to figure out the number of variables missing. 

 

Edit: good luck and please share more when you are ready. It will satisfy the math nerd in me


Edited by Archie761, 16 March 2016 - 06:43 PM.


Oralflame #3 Posted 16 March 2016 - 06:46 PM

    First Sergeant

  • Players
  • 19110 battles
  • 3,318
  • [-AIDS]
  • Member since:
    12-16-2014

I also had asked this a while, specifically if anyone had full calculations.

 

Vsavsa1 was the only responder. And he just sent me pictures of his Algebra homework.



Ajeroth #4 Posted 16 March 2016 - 07:28 PM

    Private

  • Players
  • 3360 battles
  • 6
  • [DMT]
  • Member since:
    12-21-2015

View PostArchie761, on 16 March 2016 - 06:39 PM, said:

Edit: good luck and please share more when you are ready. It will satisfy the math nerd in me

 

I came close to using the  Moore-Penrose pseudoinverse to solve the overdetermined set.  I guess I can try using this but I doubt it would get usefull information since the errors I'm seeing are rare and significant . . .



af24snipes #5 Posted 16 March 2016 - 07:46 PM

    First Sergeant

  • Players
  • 10893 battles
  • 3,256
  • Member since:
    06-28-2014
+1 for math that I can't do.

Ajeroth #6 Posted 17 March 2016 - 10:03 PM

    Private

  • Players
  • 3360 battles
  • 6
  • [DMT]
  • Member since:
    12-21-2015
Updated the original post to include the first cut of the algorithm

tlouie #7 Posted 17 March 2016 - 10:06 PM

    First Sergeant

  • Players
  • 9642 battles
  • 5,665
  • [TAT]
  • Member since:
    04-07-2015
+1 for doing the hard maths work.

Is it possible that the outlying datapoints are instances where WG's credit calculation is buggy?

gogo_gadget_arms #8 Posted 17 March 2016 - 11:02 PM

    Junior Sergeant

  • Players
  • 34610 battles
  • 120
  • [SRT-]
  • Member since:
    07-05-2014
What about damage you do where someone else is assisting? I believe you end up sharing the XP. Perhaps you also share the credits?

Ajeroth #9 Posted 18 March 2016 - 07:20 AM

    Private

  • Players
  • 3360 battles
  • 6
  • [DMT]
  • Member since:
    12-21-2015

View Postgogo_gadget_arms, on 17 March 2016 - 11:02 PM, said:

What about damage you do where someone else is assisting? I believe you end up sharing the XP. Perhaps you also share the credits?

 

That toggled my memory from the WoT grid of credit calculation. Specifically:

 

Action Credit Award
Damaging enemies that your team is spotting, but not yourself 50% of the above
Damage done to targets you are spotting, by team members who are not spotting them themselves 50% of what the team member doing the damage would earn if he spotted the target himself, divided by the number of team members spotting the target.

 

In the case of the first, it would explain the errors where the algorithm estimates high.

 

This point brings up an unresolveable problem.  There is no way to determine the damage you inflicted on account to someone else spotting, with the exception of knowing your hitting someone outside your view range, remember how much you hit for.  I have verified with a TD in a camp spot in Mines.  I hit a tank across the map once for 48, and the algorithm estimated high by 20%.



Ajeroth #10 Posted 18 March 2016 - 07:24 AM

    Private

  • Players
  • 3360 battles
  • 6
  • [DMT]
  • Member since:
    12-21-2015

I will update the algorithm after some Tier 5+ tests.  But I more or less stand by the relative accuracy with the exception of the assisted damage issue . . .

 

It is enough to now do some API data mining to determine a tanks credit effectiveness/minute (CE/M).  :justwait:



olddatsunfan #11 Posted 04 May 2016 - 08:10 PM

    First Sergeant

  • Players
  • 34394 battles
  • 2,645
  • [-CFR-]
  • Member since:
    06-02-2015
Ajeroth, if you are still working on this I can help with data.  What vehicles are you missing so far?
QO6Y0A8.gif

1991Syclone #12 Posted 20 June 2016 - 04:12 PM

    Senior Sergeant

  • Players
  • 46093 battles
  • 931
  • [DEF]
  • Member since:
    07-29-2014
With the reworked credit earnings, does this spreadsheet formula still apply?  If not, we all need to start contributing to determine the best earning tanks in the game and play those to get the credits.

MayoNasalSpray #13 Posted 20 June 2016 - 04:18 PM

    ( ͡° ͜ʖ ͡°)

  • Players
  • 34351 battles
  • 4,111
  • [MSELF]
  • Member since:
    01-27-2015
It's funny when people complain about not making credits when the answer is simply your spotting is below 1 but don't listen to me I like making money more apparently...

evil00genius #14 Posted 20 June 2016 - 07:41 PM

    Senior Sergeant

  • Players
  • 45232 battles
  • 510
  • [III-S]
  • Member since:
    07-30-2014
One factor to explore is that your damage inflicted and damage assisted ignore the tier of the opposing tank that was damaged.  I know that this impacts exp earned, perhaps it also has an impact on credit earnings.  You get more exp for damaging tanks that are a tier above and less for a tier below, could a similar co-factor in the damage assessment change the credit earnings?




1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users