Jump to content


RibbleStripe - does matchmaker analyze player skill?

MM Matchmaker

  • Please log in to reply
289 replies to this topic

CA_vampire #261 Posted 21 February 2018 - 11:52 PM

    __Banned__

  • Players
  • 31039 battles
  • 1,264
  • [SLIT]
  • Member since:
    11-21-2016

A lot of players who are not good in the game would like to have some plausible excuses. So, they prefer the explanation that MM+RNG are rigged. 

 

A lot of players who are good in the game want to believe that everything is completely random, and it is only their superior skill that matter. They are scared in the possibility that MM+RNG are rigged (which is nothing new, it is standard practice in gaming softare, hence the papers and the patents). 

 

In reality, neither group the foggiest idea what the software code is really doing, haha!!!  

 

 

(And actually, referring to a player's skill is used by many as an "ad hominem" attack.   ;) 

 


No Mercy / No Malice

CA_vampire #262 Posted 22 February 2018 - 12:21 AM

    __Banned__

  • Players
  • 31039 battles
  • 1,264
  • [SLIT]
  • Member since:
    11-21-2016

View Post__Crusader6__, on 21 February 2018 - 07:09 AM, said:

 

Diclosure: I am neither a stats major (stats 301derailed my original attempt at a major) or computer programmer. 

 

However from my understanding ABAB and other simple randomizing sequence methodology are only effective for separating in a batch with one dimension. It could only be used to assign same tier tanks to one map.   

 

We know that Blitz MM isn’t one dimensional. 

  WG has effectively admitted to tank weighting (Honeymoon/Not Honeymoon) 

    There are also Platoons, Type Limits, and we suspect but WG has been silent on it weighting of map sequence, high/low tier efforts and more. 

 

 

 

 

 

Disclosure: I am a professional programmer, working mainly on scientific programming. Semi-retired. 

 

The standard procedure is the following:

1. At any moment, you have a pool of players that pressed "Battle". You pick 14 players according to tier and waiting time. 

2. You split the 14 players into two groups. How? One way is ABAB (simple playground algorithm that creates balanced teams, best player goes to A, second best to B and so on). Another way is random. Another way is ... well, there are a lot of ways to do it! 

 

The non-standard procedure is:

1. You create two teams A and B.

2. You pick 7 players from the pool and place them in team A and 7 players in team B. 

 

Usually the second method is used when you want non-random and non-balanced teams. For example, you designate team A as "winners" (to be helped by RNG) and you pick a couple of players with long losing streaks in A to help them get a win, for the rest you pick good players. In team B (losers) you pick players who are playing for a long time (ie addicted).  Sometimes, you do not really have concrete rules for how you pick winners and losers. It is done by a machine learning algorithm (machine learning is software that learns, it is actually easy to do it, it is not as sophisticated as it sounds, for example it has been used for decades in character recognition from scanners).

 

What is the goal? To maximize player retention and player spending. That is, not for a particular player, but for as many players as possible. There are a lot of scientific studies about this. There is also scientific research into maximizing the euphoria hormones while gaming, but I don't know if any company has applied any of the findings, they are currently doing research only, as far as I know.... 

 

Contrary to what some people seem to believe, none of that has the goal to make "gaming skill" irrelevant. This would be counterproductive.  Skill is important. And yet, it is definitely a "rigged system" of MM+RNG. The game plays you, haha! 

 

 

 

 


No Mercy / No Malice

minitel_NA #263 Posted 22 February 2018 - 01:03 AM

    First Sergeant

  • Players
  • 454 battles
  • 6,257
  • [EUREF]
  • Member since:
    12-19-2015

To say that we have stripped the official MM thread of a good 15 pages. It breaks my heart.

http://forum.wotblitz.com/index.php?/topic/86788-official-matchmaking-discussion-thread/


Minitelrose visiting from EU, occasional player/forumer in the NA

 


minitel_NA #264 Posted 22 February 2018 - 01:11 AM

    First Sergeant

  • Players
  • 454 battles
  • 6,257
  • [EUREF]
  • Member since:
    12-19-2015

Need to say CA vampire as well as crusader got some nice posts above here. 

 

About ABAB procedure I need to say this : 

it’s great for training a football team. it’s not great at surprising a player out of his routine. Making a crazy easy game followed by a crazy hard game is much better to get his blood pumping. ABAB can’t very much be qualified as random : It really is not. WG always stated that the MM is random. (Whatever that means. There are also so many requirements for the mm like failtoons, etc.. crusader got the list). This doesn’t rule out that some ABAB is applied. I remember for a period of time, I couldn’t play with my unicum friend without having unicums in front of us. I’m talking thousands and thousands of games, until my friend quit for good.

 

Hmmm... more thinking now...


Edited by minitel_NA, 22 February 2018 - 01:19 AM.

Minitelrose visiting from EU, occasional player/forumer in the NA

 


Mayjaplaya #265 Posted 22 February 2018 - 01:36 AM

    [YYK]

  • Players
  • 16046 battles
  • 1,494
  • [MBTS]
  • Member since:
    07-05-2016

View Postzeur0sis, on 21 February 2018 - 02:53 PM, said:

 

...when you have been proven wrong and have nothing else left to say.   

 

Classic response in any hobby/game/sport. "Well this is just a stupid game that doesn't mean anything" :rolleyes:

 

(Also the Tiger I is now one of the most overpowered tanks and one of the best heaviums in the whole game. If you're not putting in really fat numbers in that, then you can't be helped.)


YYK

 

Sig credit: Panbun


4sfield #266 Posted 22 February 2018 - 01:53 AM

    Space Lord

  • Players
  • 33618 battles
  • 5,474
  • [-1]
  • Member since:
    07-08-2014

View PostMayjaplaya, on 21 February 2018 - 08:36 PM, said:

 

Classic response in any hobby/game/sport. "Well this is just a stupid game that doesn't mean anything" :rolleyes:

 

(Also the Tiger I is now one of the most overpowered tanks and one of the best heaviums in the whole game. If you're not putting in really fat numbers in that, then you can't be helped.)

 

  You forgot, “I play for fun”.

 

  Most nubs think I’m trying to dull their shine when I’m giving instructions they say no and I argue back at the beginning of the match. “It’s a game is a common answer to a lot of comments. 

 

  I did have a team at Port Bay today that listened to everything I said. We were at 2-5 heavy disadvantage and I called the river. The reds went town and it gave me time to give good direction. After we killed a lone straggler, “don’t chase town”. Then, “Turn around”. We killed 2 tanks coming around, then 2 coming up the road from town. Then I said, “Now town all’. I got affirmitives every time from everyone. None of them were >48%ers and we win easily just because they listened. I got 6 upvotes after the game, one from every team mate. I felt like I just gave birth to 6 kids. When they listen and do it right, it doesn’t matter who or what’s in the other team. The problem is, most of the time they don’t do either.

 

  It helped a little that I was running a Tiger I. It is wicked OP, I agree. 


 

Snowflakes - 1, Good Guys - 0

 

FREE _stealy!!!


7ero #267 Posted 22 February 2018 - 02:03 AM

    First Sergeant

  • Players
  • 46239 battles
  • 1,925
  • [SRT]
  • Member since:
    09-17-2016

View PostMayjaplaya, on 21 February 2018 - 07:36 PM, said:

 

Classic response in any hobby/game/sport. "Well this is just a stupid game that doesn't mean anything" :rolleyes:

 

(Also the Tiger I is now one of the most overpowered tanks and one of the best heaviums in the whole game. If you're not putting in really fat numbers in that, then you can't be helped.)

 

Ikr.    You could almost hear his feet stomping on the ground when he posted that.   



 

“ Mm sucks just like u r face. Now Go cry on tissue not here.” ~ raj

 

 

 


CA_vampire #268 Posted 22 February 2018 - 08:50 AM

    __Banned__

  • Players
  • 31039 battles
  • 1,264
  • [SLIT]
  • Member since:
    11-21-2016

View Postminitel_NA, on 21 February 2018 - 06:11 PM, said:

Need to say CA vampire as well as crusader got some nice posts above here. 

 

About ABAB procedure I need to say this : 

it’s great for training a football team. it’s not great at surprising a player out of his routine. Making a crazy easy game followed by a crazy hard game is much better to get his blood pumping. ABAB can’t very much be qualified as random : It really is not. WG always stated that the MM is random. (Whatever that means. There are also so many requirements for the mm like failtoons, etc.. crusader got the list). This doesn’t rule out that some ABAB is applied. I remember for a period of time, I couldn’t play with my unicum friend without having unicums in front of us. I’m talking thousands and thousands of games, until my friend quit for good.

 

Hmmm... more thinking now...

 

 

The thing about ABAB is that it is a verifiable algorithm. Verifiable by *any* player. This is important! You can open the wotb Assistant, and find the "personal rankings" for all the players in your team and for the red team, and verify that ABAB was used (or not used). 

 

I think you always have "surprises" in multiplayer games because we are humans. Great players do not always play well. Teams do not always cooperate. 

 

ABAB is used all over the world (especially in playground soccer and basketball, 11 vs 11, or 5 vs 5, or any vs any) because it is simple, fair, and leads to balanced matchups and interesting games. You rarely get 7-0 results in playground soccer. 

 

However, when software developers want to have "rigged" games (like the "dynamic difficulty adjustment" class of algorithms) then you don't use ABAB because it restricts the freedom of the system to "manipulate" the MM. IHMO, that's why we have so many 7-0 battles in blitz.  

 

 

 


No Mercy / No Malice

minitel_NA #269 Posted 22 February 2018 - 10:37 AM

    First Sergeant

  • Players
  • 454 battles
  • 6,257
  • [EUREF]
  • Member since:
    12-19-2015

View PostCA_vampire, on 22 February 2018 - 08:50 AM, said:

 

 

The thing about ABAB is that it is a verifiable algorithm. Verifiable by *any* player. This is important! You can open the wotb Assistant, and find the "personal rankings" for all the players in your team and for the red team, and verify that ABAB was used (or not used). 

 

I think you always have "surprises" in multiplayer games because we are humans. Great players do not always play well. Teams do not always cooperate. 

 

ABAB is used all over the world (especially in playground soccer and basketball, 11 vs 11, or 5 vs 5, or any vs any) because it is simple, fair, and leads to balanced matchups and interesting games. You rarely get 7-0 results in playground soccer. 

 

However, when software developers want to have "rigged" games (like the "dynamic difficulty adjustment" class of algorithms) then you don't use ABAB because it restricts the freedom of the system to "manipulate" the MM. IHMO, that's why we have so many 7-0 battles in blitz.  

 

 

 

 

your post was very interesting, because it was informative, and it was leading the discussion to a technical standpoint, and not an opinion or personal experience (“perception”;) standpoint.

 

but  I need to disagree here for 2 points :

 

1

Simple,  Fair and balanced are again, interesting to train a team. They might not be exciting in a 1v all situation like in the public games. As WG said, randomness provides a more diverse and unexpected range of situations, which I agree with. If you could tell the level of the enemy team each time before playing, like everyone wants, you would have a very even set of choices and strategy. And that’s why I’m against XVM. Being able to asses your own team behavior in the first seconds, discovering the enemy strength and weaknesses as the game develops, and adapting to the situation... I very much value this experience of the game play, it’s very unlinear and interesting to study, and improve upon.

Besides the goal of ABAB is to make everyone 50% victory and force players to improve. Regardless of players levels, unicums or noobs. I don’t think many peaopl are ready to accept both of these at the same time. 

Now about WG and their exact definition of randomness... and the gap between what they say and what they do... I let people decide that for themselves. XD

 

2

And about 7-0 in WoTb compared to soccer... I think if we removed one player out of a team each time that team let the ball in, it would show a lot more 7-0 !

lolz. 

if anything, the mechanics of the game leads to unbalance situation from the initial first shot. That’s how unicums are sometimes defeated. Make their team 6v7 or 5v7 asap, if not rushing them first.

And that’s probably why other comparable games provide a respawn mechanics, giving the losing team a possibility to recover from some initial bad luck due to RNG (first shot failed) or just bad turn of event (spotting gone wrong, falling of a cliff), or wrongly turned strategic choice (unfateful meeting around a corner, red team not going expected side...).


Minitelrose visiting from EU, occasional player/forumer in the NA

 


_Crusader6_ #270 Posted 22 February 2018 - 02:42 PM

    There is no Society- go back to sleep.

  • Players
  • 74467 battles
  • 16,819
  • [-EXP-]
  • Member since:
    12-08-2014

CA Vamp, you making some key errors more Programers do. 

   I find many in your field seem to try to simplify or ignore other variables. 

Garbage in = Garbage out. 

 

ABAB could be used for certain aspects, but why? Secondly it wouldn’t allow for randomness in map assignments 

  So if you already skew the player pool. 

 

Secondly - balanced games?  How fun are 7 Super Unicum V 7 Unicum games ? 

   Like watching paint dry.  That seems to be what you propose unless I am reading you wrong. 

Like Mintel said, great for breaking up kids sports teams for balance - but poor for randomness of play in Blitz. 

 

RNG is set for Penetration and Dmg at moment of incident. 

  Despite what you seem to insinuate, most losses are easily dissected and a player has ability to impact the vast majority of their games.   

   

  Anyone with some objectivity can look at my stats, or anyone else’s. 

You can see tanks and tiers where folks struggle - and if you watch a few videos or their game play one can usually see why. 

   My issue is I often realize what I did wrong about 1 second too late. 

 

I accept this is a ME issue, and I don’t go blaming WG for the fact I don’t have a 60%+ WR in all my tier X tanks. 

    


I hate Annihilator spammers...  

 
Tank Hoarder: 436 tanks in Garage:  449/451 Played Tanks Aced
 
I need more tanks...
 
    Wallet Warrior: Loyal Original M60 owner
 
 

 


fedsnbr #271 Posted 22 February 2018 - 02:47 PM

    Senior Sergeant

  • Players
  • 58707 battles
  • 999
  • [STRYK]
  • Member since:
    11-26-2014
@CA_vampire I’m lost on your main point after reading the whole thread. If you are proposing ABAB, that is essentially skill-based MM. Then what was the point for asking Ribble why they wouldn’t use ABAB earlier? They have stated that the MM is not intended to be skill related. (Let’s assume that that is not a lie in this context, otherwise it would be an entirely different discussion)

fedsnbr #272 Posted 22 February 2018 - 02:56 PM

    Senior Sergeant

  • Players
  • 58707 battles
  • 999
  • [STRYK]
  • Member since:
    11-26-2014
Also I am not sure ABAB would be completely verifiable in wotb, at least not to the point of assuring everybody that there is no “rigging”. You can only verify that the 14 players are matched up by ABAB, but as long as you don’t see the entire player queue, you still cannot verify that there is no “rigging” in the selection of these 14 players. For instance, MM could pick them so that the players it wants to win are placed on the A team, led by a unicum, while the top (ie the second best out of 14) is a 55%er.

DanAnsomeHansome #273 Posted 23 February 2018 - 12:57 AM

    First Sergeant

  • Players
  • 43882 battles
  • 1,692
  • Member since:
    06-01-2015
Yes matchmaking does certainly analyze player skill. It’s called a ratings battle.

ABN_Bearcat #274 Posted 30 April 2019 - 12:28 AM

    Senior Sergeant

  • Players
  • 41216 battles
  • 880
  • Member since:
    11-16-2013
I have a question that I think is a little more on the mark instead of MM considering player statistics. Here is the question, "Does the matchmaker try to get players a win?" Basically, are players matched to a team based on recent win-loss of all the players in the queue in an attempt to "balance" out the near/recent win-loss?


Bellatormonk #275 Posted 30 April 2019 - 04:55 PM

    Tinfoil Knight

  • Players
  • 35552 battles
  • 1,469
  • [D2TH]
  • Member since:
    04-12-2011

Is this like when you play the same tank for example, hit battle, are insta joined into matchs after 5 secs or less, lose 5x in a row to face rolls no matter what and then you hit battle the 6th time and MM has you sit there for 30 secs and you wind up on a Green team that wins with a face roll whether you participate or not??  :) 

So why and what exactly are you sitting there for 30 seconds the 6th game as opposed to the first 5 games in 5 secs?  Was MM finding you a "dream team" to make up for the first 5 train wrecks of window lickers?? :)

 

 

I know this quote is old as year old dirt however since I know Crusader will read and respond here I have to ask about this one.  Why would this be so boring?  Wouldn't it be more entertaining, fair and challenging if MM actually did this?

 

"Secondly - balanced games?  How fun are 7 Super Unicum V 7 Unicum games ? 

   Like watching paint dry.  That seems to be what you propose unless I am reading you wrong. 

Like Mintel said, great for breaking up kids sports teams for balance - but poor for randomness of play in Blitz."



_Crusader6_ #276 Posted 30 April 2019 - 05:25 PM

    There is no Society- go back to sleep.

  • Players
  • 74467 battles
  • 16,819
  • [-EXP-]
  • Member since:
    12-08-2014

View PostBellatormonk, on 30 April 2019 - 11:55 AM, said:

Is this like when you play the same tank for example, hit battle, are insta joined into matchs after 5 secs or less, lose 5x in a row to face rolls no matter what and then you hit battle the 6th time and MM has you sit there for 30 secs and you wind up on a Green team that wins with a face roll whether you participate or not??  :) 

So why and what exactly are you sitting there for 30 seconds the 6th game as opposed to the first 5 games in 5 secs?  Was MM finding you a "dream team" to make up for the first 5 train wrecks of window lickers?? :)

 

 

I know this quote is old as year old dirt however since I know Crusader will read and respond here I have to ask about this one.  Why would this be so boring?  Wouldn't it be more entertaining, fair and challenging if MM actually did this?

 

"Secondly - balanced games?  How fun are 7 Super Unicum V 7 Unicum games ? 

   Like watching paint dry.  That seems to be what you propose unless I am reading you wrong. 

Like Mintel said, great for breaking up kids sports teams for balance - but poor for randomness of play in Blitz."

 

I don’t think ABBA etc will work for Blitz simply as what do you use for sort criteria. 

WR alone won’t work.  

 

WG doesn’t use any single metric that could be decently used for that. 

 

Also 7 v 7 games are hard to rig without player input. 

  

I toon with AbnBearCat, the majority of our losses are due to us not being on voice and miscues/separations where we can’t support each other. I don’t expect any of the other 5 to help us win at all. 

 

I don’t mind playing games against skilled players - but I don’t see it being viable the entire time - as it will gradually erode better players WR and raise worse players.  

 

 

 


I hate Annihilator spammers...  

 
Tank Hoarder: 436 tanks in Garage:  449/451 Played Tanks Aced
 
I need more tanks...
 
    Wallet Warrior: Loyal Original M60 owner
 
 

 


ABN_Bearcat #277 Posted 01 May 2019 - 01:03 AM

    Senior Sergeant

  • Players
  • 41216 battles
  • 880
  • Member since:
    11-16-2013

View Post__Crusader6__, on 30 April 2019 - 12:25 PM, said:

 

I don’t think ABBA etc will work for Blitz simply as what do you use for sort criteria. 

WR alone won’t work.  

 

WG doesn’t use any single metric that could be decently used for that. 

 

Also 7 v 7 games are hard to rig without player input. 

  

I toon with AbnBearCat, the majority of our losses are due to us not being on voice and miscues/separations where we can’t support each other. I don’t expect any of the other 5 to help us win at all. 

 

I don’t mind playing games against skilled players - but I don’t see it being viable the entire time - as it will gradually erode better players WR and raise worse players.  

 

 

 

 

I will say this you are a great platoonmate and I definitely win more when we do. I also agree that in times that we lose, it is due to miscues as you say. However, since update 5.4 my win rate has gone down and many of my clan mates and other unicum friends echo this. I want to know what has changed in the game to cause this. I do know that Wargaming has tweaked the matchmaker to make it "more fair" to all players or something like that. What did they tweak/change? This is a game and it is supposed to be fun, but there is ever increasing number of just completely clueless and absolutely stupid players in this game> I had one today with a 33% win rate. How is that even possible?


4sfield #278 Posted 01 May 2019 - 03:33 AM

    Space Lord

  • Players
  • 33618 battles
  • 5,474
  • [-1]
  • Member since:
    07-08-2014

  I just want it like it used to be. Assign a value per vehicle and keep the sum of the numbers between the 2 teams within 1. I don’t care if there are 7 TDs on one team against 5 lights a medium and a medium and a heavy. All I want to see is it based on what the vehicle is worth, not on whether or not everyone has the same amount of tanks for each type. 

 

  Platoon mm needs to go away too and it needs to be like it once was. You grab a toon mate and boom, you get a game opposing team platoon or not. 

 

  As far as the tweak goes, whatever it is it isn’t fair and it needs to go away.


 

Snowflakes - 1, Good Guys - 0

 

FREE _stealy!!!


minitel_NA #279 Posted 01 May 2019 - 10:31 PM

    First Sergeant

  • Players
  • 454 battles
  • 6,257
  • [EUREF]
  • Member since:
    12-19-2015

I watched that guy in the last Q&A

“Been working on MM every day since I was born”

good to know that u change it constantly buddy, and it’s not that I want u to lose ur job but really

 

what I really want to see is the 14 next guys to go randomly into 2 teams.

you might exception to get a platoon on each side but I’m not even sure I want that.

oh and failtoons ? I don’t care. They deserve what they chose.

 

that will end the repetitiveness of current games with exactly the same tanks on both sides.

i don’t care if one side has 3kv2 and the other only light tanks. U should adapt to that.

having exactly one on each side is horrible, and it happens every single time.

same goes for autoloaders, light tanks etc...


Edited by minitel_NA, 01 May 2019 - 10:32 PM.

Minitelrose visiting from EU, occasional player/forumer in the NA

 


4sfield #280 Posted 01 May 2019 - 10:49 PM

    Space Lord

  • Players
  • 33618 battles
  • 5,474
  • [-1]
  • Member since:
    07-08-2014

View Postminitel_NA, on 01 May 2019 - 05:31 PM, said:

I watched that guy in the last Q&A

“Been working on MM every day since I was born”

 

 

   He’s a 40%er, guarantee it. Otherwise it would have been left alone a long time ago. 


 

Snowflakes - 1, Good Guys - 0

 

FREE _stealy!!!






Also tagged with MM, Matchmaker

5 user(s) are reading this topic

0 members, 5 guests, 0 anonymous users