Jump to content


Why Christmas Box RNG is badly coded: Statistics say something is funky


  • Please log in to reply
72 replies to this topic

flyintiger #1 Posted 13 January 2019 - 02:36 PM

    Private

  • Player
  • 19869 battles
  • 47
  • [_GITS] _GITS
  • Member since:
    01-01-2013

 

So I got 500 ornaments from boxes, and the results are: 

 

Tier 5: 1

Tier 4: 17

Tier 3: 68

Tier 2: 148

Tier 1: 257

 

Ok, so what else is new, drop rate for T5 ornaments sucks badly. BUT. What is more worrying is there is a huge number of duplicates of same ornament, that statistically shouldn't be there!

 

For example, at tier 4 I had 4 duplicates of same ornament. If there's a 17/500 chance of getting a T4 ornament, and RNG is pure random, chance of getting four of the same T4 ornaments is (17 / 500 ) ^ 4 = 0,000013 - or about 1/750.000. Sounds probable? Correct, it's not. 

 

What I think is happening: instead of building a "true random" (or technically "weighted random" I guess) RNG system where each ornament is pulled from a random list with single pull (probablility is adjusted so that each T5 ornament gets 1 ticket, each T1 ornament gets 99 tickets -> 1 ticket is pulled from total ticket list), guys at WG have build it like this: 1) Pull which tier -> 2) Pull which collection -> 3) Pull which of the T5 ornaments in that collection. 

 

So, statistically "pull tier * pull collection * pull which T5 ornament" is still quite random. But, if your "1) pull tier, 2) pull collection" ends up in point "Tier=5, Collection=A", you are now about 3x more likely to pull the same ornament twice in same collection, or about 70x more likely to pull the same ornament four times in same collection - since the chance for last pull is only about 1/18.    

 

Even still, this "1-2-3" method still gives us only about 1/10.000 chance of pulling 4x same T4 ornaments, so I suspect there is also something funky about how the algorithm generates the random seed for running the RNG. It might be that only one seed is generated at the moment you click "open boxes", or something else that might give you same seed number more than once.

 

Someone who knows about statistics and coding true random RNG's might be able to explore this further. 



jabster #2 Posted 13 January 2019 - 02:49 PM

    Field Marshal

  • Beta Tester
  • 12721 battles
  • 25,713
  • [WSAT] WSAT
  • Member since:
    12-30-2010

View Postflyintiger, on 13 January 2019 - 01:36 PM, said:

 

So I got 500 ornaments from boxes, and the results are: 

 

Tier 5: 1

Tier 4: 17

Tier 3: 68

Tier 2: 148

Tier 1: 257

 

Ok, so what else is new, drop rate for T5 ornaments sucks badly. BUT. What is more worrying is there is a huge number of duplicates of same ornament, that statistically shouldn't be there!

 

For example, at tier 4 I had 4 duplicates of same ornament. If there's a 17/500 chance of getting a T4 ornament, and RNG is pure random, chance of getting four of the same T4 ornaments is (17 / 500 ) ^ 4 = 0,000013 - or about 1/750.000. Sounds probable? Correct, it's not. 

 

What I think is happening: instead of building a "true random" (or technically "weighted random" I guess) RNG system where each ornament is pulled from a random list with single pull (probablility is adjusted so that each T5 ornament gets 1 ticket, each T1 ornament gets 99 tickets -> 1 ticket is pulled from total ticket list), guys at WG have build it like this: 1) Pull which tier -> 2) Pull which collection -> 3) Pull which of the T5 ornaments in that collection. 

 

So, statistically "pull tier * pull collection * pull which T5 ornament" is still quite random. But, if your "1) pull tier, 2) pull collection" ends up in point "Tier=5, Collection=A", you are now about 3x more likely to pull the same ornament twice in same collection, or about 70x more likely to pull the same ornament four times in same collection - since the chance for last pull is only about 1/18.    

 

Even still, this "1-2-3" method still gives us only about 1/10.000 chance of pulling 4x same T4 ornaments, so I suspect there is also something funky about how the algorithm generates the random seed for running the RNG. It might be that only one seed is generated at the moment you click "open boxes", or something else that might give you same seed number more than once.

 

Someone who knows about statistics and coding true random RNG's might be able to explore this further. 

 

If you generate one random seed for each time you go to open boxes then where’s the problem?

Baldrickk #3 Posted 13 January 2019 - 02:54 PM

    Field Marshal

  • Player
  • 32165 battles
  • 16,693
  • [-MM] -MM
  • Member since:
    03-03-2013
Whoever said that drop rates for ornaments are identical?

Balc0ra #4 Posted 13 January 2019 - 03:01 PM

    Field Marshal

  • Player
  • 72718 battles
  • 20,666
  • [WALL] WALL
  • Member since:
    07-10-2012

You got 1 level 5? Thought luck. I got lvl 4 the least. As that's where I lacked the most on all 4 collections at the end and had to spend lots of gems. And I had half on one collection on lvl 5. As it did seem the drop the most for me on the weekly missions vs the daily mission boxes that hardly gave me anyone at all. 

 

But as also pointed out. 

View PostBaldrickk, on 13 January 2019 - 02:54 PM, said:

Whoever said that drop rates for ornaments are identical?

 

Just because you got 1. Don't mean everyone else did, or had the same drop rate. Some might have had 17/500 too ofc, but some might have had 50/500 on lvl 4 too. Or less then 10 like I did. Same with pulling duplicates. 

 

Then again at the end of the day it's still random, and the idea is still... to farm low tiers for gems and build higher tiers. just like you had to mix 5 lower levels to get one higher level 2 years ago. It was not a given you would get all the lvl 5 in the boxes. As it's also designed for those that play often. As you got 500 decorations, that's what? 166 small boxes? So you played for half the days then? And when you don't play almost daily to farm enough. That means that you also have less of a chance to get more lvl 5's randomly. 


Edited by Balc0ra, 13 January 2019 - 03:09 PM.


Mikalaras #5 Posted 13 January 2019 - 03:26 PM

    Private

  • Player
  • 1684 battles
  • 3
  • Member since:
    03-05-2016
That's not bad coding, it's gambling and it's implemented like this on purpose. And guess what; the house never loses.

jabster #6 Posted 13 January 2019 - 03:29 PM

    Field Marshal

  • Beta Tester
  • 12721 battles
  • 25,713
  • [WSAT] WSAT
  • Member since:
    12-30-2010

View PostBaldrickk, on 13 January 2019 - 01:54 PM, said:

Whoever said that drop rates for ornaments are identical?

 

I’ve always assumed that different decorations are weighted towards drop rates but I'm not so sure whether it’s linked to what you already have.



Baldrickk #7 Posted 13 January 2019 - 03:57 PM

    Field Marshal

  • Player
  • 32165 battles
  • 16,693
  • [-MM] -MM
  • Member since:
    03-03-2013

View Postjabster, on 13 January 2019 - 03:29 PM, said:

 

I’ve always assumed that different decorations are weighted towards drop rates but I'm not so sure whether it’s linked to what you already have.

Well, it appearing that way would be kind of obvious.

If you roll a dice trying to get a 6, you've likely already tried and failed at least once. 

The chance of getting a 6 is fairly low, so the next time you roll, you have a higher chance of getting a result you already have (!6) than the result you need (6).



Nishi_Kinuyo #8 Posted 13 January 2019 - 04:06 PM

    Lieutenant General

  • Player
  • 9036 battles
  • 6,239
  • [GUP] GUP
  • Member since:
    05-28-2011

View PostBaldrickk, on 13 January 2019 - 03:57 PM, said:

Well, it appearing that way would be kind of obvious.

If you roll a dice trying to get a 6, you've likely already tried and failed at least once. 

The chance of getting a 6 is fairly low, so the next time you roll, you have a higher chance of getting a result you already have (!6) than the result you need (6).

Depends a bit on which die you use for it though.

Chessex are notoriously bad for rolling anything other than a 1 for example.

From a case study, the chances of rolling a 1 on a chessex was around 30%, leaving just 14% on average for the rest assuming that remains equal.

 

Also, expected a whine about the large boxes instead of the small ones.



Homer_J #9 Posted 13 January 2019 - 04:13 PM

    Field Marshal

  • Beta Tester
  • 32284 battles
  • 35,325
  • [WJDE] WJDE
  • Member since:
    09-03-2010

I kept a track of the last few days where I played enough to get most of the missions done and my tier V drop rate was about 1 in 50 (ornaments not boxes)

 

Tier IV was about 1 in 10 and it increased down the tiers.

 

 



TakoezaEU #10 Posted 13 January 2019 - 04:18 PM

    Lance-corporal

  • Player
  • 42462 battles
  • 57
  • [SPIKE] SPIKE
  • Member since:
    06-23-2014
I agree with flyintiger 100%! The coding is bad. If you code for a random number, you should not get the same number 6 times in a row. That is not random. I just need 3 decorations to fill my New Year Album. I gathered enough shards and set the Large Holidron Decorider 2.0 to Pendant for the decoration, IV for the level and New Year for the collection. What I need is a Yuri Gagarin Pendant to complete my level IV collection. What I generate is a Silver Star Pendant. SIX TIMES IN A ROW! Really? That is random? To all the people above who commented, that is like rolling a 6, not just 6 times, but 6 times in a row! Any Casino will tell you it is nearly impossible and not random. I got none of the other pendants that were also available. I got the Silver Star six times in a row. I am going to keep trying and will update this thread if this "streak" continues. I took a screen shot of this because it is just not randomness. It smells like bad coding. Then again given Wargaming's track record for fairness, I probably should not complain.    

jabster #11 Posted 13 January 2019 - 05:08 PM

    Field Marshal

  • Beta Tester
  • 12721 battles
  • 25,713
  • [WSAT] WSAT
  • Member since:
    12-30-2010

View PostBaldrickk, on 13 January 2019 - 02:57 PM, said:

Well, it appearing that way would be kind of obvious.

If you roll a dice trying to get a 6, you've likely already tried and failed at least once. 

The chance of getting a 6 is fairly low, so the next time you roll, you have a higher chance of getting a result you already have (!6) than the result you need (6).

 

My point was more the chance of getting decoration X was different than decoration Y regardless of what you require but is what you require factored in. So to take the die example, if you’ve rolled a six and need another six are the odds still 1 in 6 or instead ‘tweaked’ to be less than that.

 

I’ve no reason to assume it is but with this being WG it wouldn’t completely surprise me if this was the case.


Edited by jabster, 14 January 2019 - 08:30 AM.


__Eric__ #12 Posted 13 January 2019 - 05:22 PM

    Second Lieutenant

  • Player
  • 34487 battles
  • 1,217
  • [T-D-U] T-D-U
  • Member since:
    03-15-2013
You should have bought 200 Large boxes like me. Got almost all tier 5 ornaments rightaway (not to mention the tanks, gold and other stuff). The few I didn't have, I crafted.

Nishi_Kinuyo #13 Posted 13 January 2019 - 05:42 PM

    Lieutenant General

  • Player
  • 9036 battles
  • 6,239
  • [GUP] GUP
  • Member since:
    05-28-2011

View Postjabster, on 13 January 2019 - 05:08 PM, said:

 

My point was more the chance of getting decoration X was different than decoration Y regardless of what you require.

You need a 6, grab a chessex die and start rolling trying to get it.

Lets see how quickly you manage. ;)



malachi6 #14 Posted 13 January 2019 - 06:24 PM

    Brigadier

  • Player
  • 50920 battles
  • 4,250
  • Member since:
    04-14-2011

Getting the same number 6 times in a row is not random?

 

So, random is guaranteed not getting the same number 6 times in a row?



Blubba #15 Posted 13 January 2019 - 06:53 PM

    Captain

  • Player
  • 60017 battles
  • 2,344
  • Member since:
    05-30-2011

View Postmalachi6, on 13 January 2019 - 05:24 PM, said:

Getting the same number 6 times in a row is not random?

 

So, random is guaranteed not getting the same number 6 times in a row?

 

That's the problem with random I guess.

I don't really know how many items there were and what the odds to get the same decoration is but I reckon if I am trying to get a decoration were there is a 50/50 chance of getting it and miss 7 times, then do it again the following day and miss again 9... yes NINE times, coupled with all the other duplicates I have had, my faith in WG's definition of 'random' wavers.

Now I am prepared to accept some confirmation bias on my behalf but the mere number of chances I have tried and failed is disturbing.

I can only say that I am somewhat disappointed by the whole thing.


 



Nishi_Kinuyo #16 Posted 13 January 2019 - 08:42 PM

    Lieutenant General

  • Player
  • 9036 battles
  • 6,239
  • [GUP] GUP
  • Member since:
    05-28-2011

View PostBlubba, on 13 January 2019 - 06:53 PM, said:

 

That's the problem with random I guess.

I don't really know how many items there were and what the odds to get the same decoration is but I reckon if I am trying to get a decoration were there is a 50/50 chance of getting it and miss 7 times, then do it again the following day and miss again 9... yes NINE times, coupled with all the other duplicates I have had, my faith in WG's definition of 'random' wavers.

Now I am prepared to accept some confirmation bias on my behalf but the mere number of chances I have tried and failed is disturbing.

I can only say that I am somewhat disappointed by the whole thing.


 

Mandatory dilbert:



SoupFork #17 Posted 13 January 2019 - 10:24 PM

    Staff Sergeant

  • Player
  • 11397 battles
  • 386
  • Member since:
    11-07-2017

It’s not badly coded.

It’s coded exactly how WG want it, to encourage you to buy more boxes to complete more collections.

 

 



brisha #18 Posted 13 January 2019 - 10:28 PM

    Lieutenant

  • Player
  • 17127 battles
  • 1,535
  • [DID0] DID0
  • Member since:
    10-11-2012
Well had 5 tries with the last of the baubles to get 2 tier 5 for the prize and no, rng fucked me once again at 1440 a shot.

Soup_Potato #19 Posted 13 January 2019 - 10:53 PM

    Staff Sergeant

  • Player
  • 282 battles
  • 302
  • Member since:
    09-22-2018

View PostNishi_Kinuyo, on 13 January 2019 - 08:42 PM, said:

Mandatory dilbert:

 

This brought to mind a thread that was here some time ago. (Last summer maybe) Which turned into people answering in comics rather than having a proper discussion, possibly due to OP of that thread being bit special. Anyone remember or have that saved? I recall it was a good read. I should start bookmarking good threads.

 

E: Nvm found it. Gotta say the forum search function does indeed work brilliantly.


Edited by Soup_Potato, 13 January 2019 - 10:58 PM.


Baldrickk #20 Posted 14 January 2019 - 12:20 AM

    Field Marshal

  • Player
  • 32165 battles
  • 16,693
  • [-MM] -MM
  • Member since:
    03-03-2013

View Postjabster, on 13 January 2019 - 05:08 PM, said:

 

My point was more the chance of getting decoration X was different than decoration Y regardless of what you require.

Ah,  right.  Agreed. 

I was just taking it a bit further,  to explain why someone would most likely be trying to get the least likely one






1 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users