[From NA Forum] Interesting statistical analysis of 400 matches outcomes based on Match Maker and players Win Rates...

[Leo_Apollo11]

Hi all,

 

[From NA Forum] Interesting statistical analysis of 400 matches outcomes based on Match Maker and players Win Rates...

 

 

"Will An Improved Matchmaker Result In Fewer Blowouts - An Analysis"

 

https://forum.worldofwarships.com/topic/243942-will-an-improved-matchmaker-result-in-fewer-blowouts-an-analysis/

 

 

Highlights:

 

Quote

Here is the process I used.   I kept track of team balance for 400 battles.  The thing I tracked was the number of so-called "red" players (the color assigned to players by the MMMonitor app who have less than a 47% account win rate).  This is the group of what the MMMonitor would suggest are the weakest players.  I used this because in my previously posted study over hundreds of battles it was obvious that the team with the least number of poor (red) players had the best chance of winning that battle.  In that data the team with 1 less red player won 61% of their matches and the team with 4 fewer red players it won 94% of the matches regardless of anything else.  This was for mainly high tier matches because that way each player tended to have enough matches to make his/her account win rate statistically meaningful (ship win rates are not as accurate imo because many times there aren't enough battles in each ship to be so). 

I've arbitrarily defined a blowout as one in which the winning team had at least 7 (i.e. over half) of their ships still alive and that the match ended in 13 minutes or less (or shorter than "normal"/closer matches).   And then I kept track of how many red players each team had in each match.  Here is the summary of that data:
 

- In these 400 battles 32% of them were blowouts. 
- In the matches where one team had 4 or more red players than the other team (very unbalanced teams based upon account win rates) the team with fewer red players blew out the other team 42% of the time
- In the matches were there were equal numbers of red players on the teams (very balanced teams based upon account win rates) there were still blowouts in 45% of the matches
- And interestingly of the blowouts, 12% of them were actually won by the team with more reds, i.e. the team with more weaker players blew out the other team.
 

The highest percentage of blowouts occurred when there was no difference in red players between the teams. It sure seems that blowouts will occur just as much regardless of using a more balanced matchmaker, at least it was that way for these 400 battles.

 

P.S.  The data by difference in red players between the teams was:
          0 difference in reds - 45% blowouts
          1 difference in reds - team with fewer reds blew out the other team 28% of the time
          2 difference in reds - team with fewer reds blew out the other team 17% of the time
          3 difference in reds - team with fewer reds blew out the other team 37% of the time
          4 or more difference in reds - team with fewer reds blew out the other team 42% of the time

 

Quote

Games in each category (% of each of the 400 total)
0 difference in reds - 14%
1 difference in reds - 31%
2 difference in reds - 23%
3 difference in reds - 19%
4 or more difference in reds - 13%

1 difference in reds,  unfavored team blew out the favored team 2% of the time
2 difference in reds, unfavored team blew out the favored team 6% of the time
3 difference in reds, unfavored team blew out the favored team 9% of the time
4 or more difference in reds, unfavored team blew out the favored team 3% of the time

Not any pattern that I can see.

 

I didn't keep data on the number of purples in the previous study for one very simple reason......the number of reds in each battle vastly outweighs the numbers of purples in each battle (in many battles there are no purples at all).  So looking at that doing "just" hundreds of battles seemed like a waste of effort.  But maybe the reason for a team with more reds winning could be that it had way more very good players.    In order to understand that and the "actual" implication of various win rates on winning it would take a lot of data and a multiple regression effort beyond my means.

 

Quote

When a team has 1 less red player it wins 61% of the time on average and when it has 4 or more less red players it wins 94% of the time on average. 

 

 

Leo "Apollo11"

[mtm78]

400 is to low sample size

I uses player account win rate as only identifier, but it doesn't cover which ships they are in or the rosters which are extremely influential. 

 

Having two destroyers in your team who are really really bad, and you're playing domination -> influence grows

Having two cruisers in your team who are really really bad, but you're in a DD and you're playing domination -> influence limited

 

20 minutes ago, Leo_Apollo11 said:

I didn't keep data on the number of purples in the previous study for one very simple reason......the number of reds in each battle vastly outweighs the numbers of purples in each battle (in many battles there are no purples at all).

This is also bad, since you would expect the presence of a purple has vastly more influence on a battle as the influence of a red. 

 

I am not sure what this 'analysis' is actually analyzing, it seems a bit WeeGee to me. 

 

Also, @Excavatus reminded us yesterday to not start new matchmaking discussion threads ( you know, like with the CV and now submarines, this topic has it's own pinned thread as to not clutter the forum ). 

[Leo_Apollo11]

Hi all,

 

5 minutes ago, mtm78 said:

400 is to low sample size

I uses player account win rate as only identifier, but it doesn't cover which ships they are in or the rosters which are extremely influential. 

 

Having two destroyers in your team who are really really bad, and you're playing domination -> influence grows

Having two cruisers in your team who are really really bad, but you're in a DD and you're playing domination -> influence limited

 

This is also bad, since you would expect the presence of a purple has vastly more influence on a battle as the influence of a red. 

 

I am not sure what this 'analysis' is actually analyzing, it seems a bit WeeGee to me. 

 

Also, @Excavatus reminded us yesterday to not start new matchmaking discussion threads ( you know, like with the CV and now submarines, this topic has it's own pinned thread as to not clutter the forum ). 

 

I agree... but it does show the significant trends (i.e. regardless of ship type the "RED" player has)...

 

BTW, I would not quality this as pure "matchmaking discussion thread" and thus I didn't put it there... I hope @Excavatus agrees!

 

 

Leo "Apollo11"

[Molly_Delaney]

Interesting Stats.

But how many of the better players were running MMM and 'threw the game' by not playing to win but farming damage (or rage quitting?).

As of today my random win rate is 46.99% so be interesting to see how this plays out.

 

[Yedwy]

Well aside the blowout part the part where the +4 tomatos brings defeat to their team in 94% of the cases supports what many of us think and sometimes utter as well - that the tomatoes are game braking to all other players and should be either banned or barring that at least restricted by the MM to their own "league" and everybody would be happier (yes even them as they would win more often)

[Margin_Walker]

Interesting exercise.

 

Sample size is still pretty low though and a better measure would have been the average account win rate of the team. You often see games where there are more red players on one team but the average win rate tilts the other way. 

[Figment]

It probably also matters WHERE these reds start. Are they on the same side? Chance of flank failure increases. Are they opposite to other reds? If a flank of reds fails due to an enemy red side, how does this affect things? Etc.

[Figment]

3 minutes ago, Yedwy said:

Well aside the blowout part the part where the +4 tomatos brings defeat to their team in 94% of the cases supports what many of us think and sometimes utter as well - that the tomatoes are game braking to all other players and should be either banned or barring that at least restricted by the MM to their own "league" and everybody would be happier (yes even them as they would win more often)

You can’t both separate them and have a winrate based MM…

 

You can at most apply recent WR from randoms to determine ranked leagues.

[PaxtonQuigly]

40 minutes ago, Leo_Apollo11 said:

In the matches were there were equal numbers of red players on the teams (very balanced teams based upon account win rates) there were still blowouts in 45% of the matches

"tHnX mM fOr gIvInG mE a LoSeR tEaM." - every potato who cries about his team.

[Yedwy]

2 minutes ago, Figment said:

You can’t both separate them and have a winrate based MM…

 

You can at most apply recent WR from randoms to determine ranked leagues.

Yes you can, when ones WR drops to tomato levels he/she should play there until it rises

[Leo_Apollo11]

Hi all,

 

3 minutes ago, Molly_Delaney said:

But how many of the better players were running MMM and 'threw the game' by not playing to win but farming damage (or rage quitting?).

 

 

The latest "Match Making Monitor" has user statistics - right now just 600 users are using the app... so I don't think that there are any "match trowing" going on at all...

 

BTW, regardless of the "Match Making Monitor" I always play for victory - it doesn't matter how good / bad my team is - I only use it to know how will I (possibly) be adjusting my tactics when trying to win!

 

 

Leo "Apollo11"

[LemonadeWarriorITA]

Could it also be dependent on the ship the Super Bad player is playing?

When I am playing a battleship and all my Super Bad destroyers suicide it is a [edited] to carry, but when they are playing a Thunderer and I am the destroyer it is a lot easier.

 

I am personally struggling a lot to win with the not-so-talented team late game carrier matches. Not to mention my will to win is almost zero during a carrier match...

[Figment]

27 minutes ago, Yedwy said:

Yes you can, when ones WR drops to tomato levels he/she should play there until it rises

The WR in either "league" does not have the same value. Unicums fighting just unicums would end up with a 50% average over their combined matches, just as noobs vs noobs would end with a 50% average over their combined matches. Someone has to lose and win, so what you're going to do is skew and muddle their statistics.

 

Furthermore, you're going to create longer queues and extra population issues by further segmenting the already dwindling and spread out pop further. It would be self-destructive to do this.

 

This is further worsened by the amount of matches someone has played. I can win one match and have a 100% WR, or 20 out of 30 and have a 66% WR. In time this could still go to 50 over 100 and have a 50% WR. But what if I don't continue playing to get to the 50/100?

 

Should people start wondering if they should keep playing if they have a current good enough WR for promotion but if they continue they might shoot themselves in the foot? Sorry, but this isn't a football league where everyone plays the same amount of matches against the same amount of people under the same time and match constraints. You can't compare stats that easily and when you're going to create selective ranking, where do you put the barriers? And what if there's more people playing? How large will you make the leagues? It's way more complex than you're suggesting.

[Yedwy]

11 minutes ago, Figment said:

The WR in either "league" does not have the same value. Unicums fighting just unicums would end up with a 50% average over their combined matches, just as noobs vs noobs would end with a 50% average over their combined matches. Someone has to lose and win, so what you're going to do is skew and muddle their statistics.

 

Furthermore, you're going to create longer queues and extra population issues by further segmenting the already dwindling and spread out pop further. It would be self-destructive to do this.

Maybe they would and maybe not, it depends how good they play now doesnt it, what this system brings in however that even with a +1 red player you are at a disadvantage and it is basically a guaranteed loss in a game where you get +2, +3 or +4 tomatoes as opposed to the enemy... For each such game one looses he/she needs (lets assume one wants good stats) to win 2 or 3 games to balance it out, and WG keeps wondering why nobody takes their games seriously as e-sports titles... 

[mtm78]

24 minutes ago, LemonadeWarriorITA said:

When I am playing a battleship and all my Super Bad destroyers suicide it is a [edited] to carry, but when they are playing a Thunderer and I am the destroyer it is a lot easier.

 

Obviously, which is why I thought the analysis to be rather meager. If the conclusion you want to reach is that red's are bad for your win chance, well then that's what you're going to find because it's rather obvious. It would be way more interesting if you actually looked at in which scenario's this influence grows or shrinks and made a statistical analysis of that so you could perhaps better gauge how much class balance has an influence in correlation with random matchmakes, or even go deeper and analyze ship influence in correlation with player skill and battle influence. 

 

You'd need such a big dataset for this though.. and so much free time :)

 

2 minutes ago, Yedwy said:

For each such game one looses he/she needs (lets assume one wants good stats) to win 2 or 3 games to balance it out, and WG keeps wondering why nobody takes their games seriously as e-sports titles... 

But this thread is about random battles and not competitive which shouldn't be influenced by 'random matchmaker' outside of the top clans maybe not selecting you if your overall win rate is not sufficient. And I expect those clans to rather have you play a few games in division as trail as to just look at stats?

 

I mean, what's the influence of this bad match maker on competitive directly? 

 

 

[_HomTanks_]

1 hour ago, Leo_Apollo11 said:

When a team has 1 less red player it wins 61% of the time on average and when it has 4 or more less red players it wins 94% of the time on average. 

The question is, what replaced this 1 or more less red players? A dark purple, a purple, a blue player, dark green, green, yellow, orange? If you don't know the replacement, you can't get a meaningful picture out of it.  

[Yedwy]

18 minutes ago, mtm78 said:

I mean, what's the influence of this bad match maker on competitive directly? 

The influence on competitive prob not much, but influence on your enjoyment of game when you see all your efforts go to waste is immense, after a while one just stops caring...

[bloodynicknames]

funny, upon just reading the title i knew one of first posts would be "bUt Y0uR s4mPl3 S1Ze iz t00 smAlL" and low and behold, i wasnt disapointed.

 

some people just prefer to put their head in the sand about the atrocious matchmaking and the fact that, yes, no matter you claim, it is rigged at times. 

people truly seem to believe that a pattern where winning streaks alternate with loosing streaks and you can already tell by looking at MMM is somehow random. 
heres the thing, WG doesnt want you to be happy. WG wants you to be frustrated and spend money on signals and premiums. WG wants even the biggest potato to win at times because otherwise they'll walk away and dont spend. In the same way WG wants good players to not get too far ahead of the curve stats wise cause if their super happy with their techtree ships, why bother buying signals and premiums. 

this game has, by far, the worst MM i've seen in more than 20 years of gaming, even War Thunder is better in that regard and thats saying something. 

 

add to the that the atrocious game design where the loss of a single ship(thanks to no respawns) will already start the snowball and you end up with the shitshow were seeing every day.

of my last 20 games 17 were complete steamrolls with the highpoint being a 950:0 victory killing of all reds. 

 

but as long as theres monkeys that spend money on this these vodka-soaked fools will never change their ways. 

[Lord0]

400 is not enough

[mtm78]

54 minutes ago, bloodynicknames said:

funny, upon just reading the title i knew one of first posts would be "bUt Y0uR s4mPl3 S1Ze iz t00 smAlL" and low and behold, i wasnt disapointed.

 

some people just prefer to put their head in the sand about the atrocious matchmaking and the fact that, yes, no matter you claim, it is rigged at times. 

people truly seem to believe that a pattern where winning streaks alternate with loosing streaks and you can already tell by looking at MMM is somehow random. 

 

See @Leo_Apollo11 this is why I think @Excavatus would start urging this is becoming the same old same old rant about rigged matchmaking. 

[CptBarney]

1k samples should be the minimum really when it comes to this sort of thing anything north of 3k and that becomes more and more solid.

 

Hopefully we get more data sets from this and updates, so we can see the trends even out a lot more.

[UnterSeeBot]

2 hours ago, Leo_Apollo11 said:

I agree... but it does show the significant trends (i.e. regardless of ship type the "RED" player has)...

 

but what does this tendatious analysis hope to achieve? The OP makes no constructive argument for an alternative to current arrangements, but picks on the easy notion of low win rate"red" players. This is a giveaway, an honest study would not focus exclusively on lower WR players (the -47% according to the OP), behind the facade this is the usual toxic bs. I quote

Quote

" The thing I tracked was the number of so-called "red" players (the color assigned to players by the MMMonitor app who have less than a 47% account win rate). "

Why not track the 57%+ players too? What about the 70%+?

I can only see this kind of discussion going round in a circle, going nowhere.

 

What we can see is that both Red tomato and Purple unicum players are equally responsible for blowouts, either for playing too badly or for playing too well. How does that help us? 

it doesn't.

[FryingTonight]

Interesting data, but I would suggest you include percentage of blowouts that included:

 

a) Lemming trains

b) BBs camping away from caps all battle, no matter the danger of points situation

c) DDs yoloing into caps early and getting deleted

 

More interestingly, would like to see percentage of blowouts where none of the above are a factor. :)