Jump to content
You need to play a total of 5 battles to post in this section.
CSV_defender

Matchmaking Needs to Adjust Algorithm

34 comments in this topic

Recommended Posts

50
[SIDE]
Members
49 posts

I've noticed a statistically unrealistic number of games where the teams are totally unbalanced in skill.  Game after game where the game is unwinnable for one side.  The winning teams take very few losses, and in many cases, none.  Many times the losing team hits zero before the game is half over.  Wargaming has access to player stats, so I suggest they incorporate this information into the matchmaking.  I've experienced more unwinnable games, regardless of carry, in the past 3 days than I've seen in 3 years.  Just because someone gets their hands on a T 9 or 10 ship doesn't mean they are competent to play it in a random match.  Also, having two divisions on one side and not the other is another problem.  I hope WG takes a look at this. 

  • Cool 7
  • Funny 1
  • Boring 7

Share this post


Link to post
Share on other sites
50
[SIDE]
Members
49 posts

Just got out of a random.  Our team got 1 kill between 12 players.  We hit zero....AGAIN.  This either stops, or I'm out.

This has become exacerbated since the CV rework.  

Edited by CSV_defender
additional information
  • Cool 1

Share this post


Link to post
Share on other sites
412
[SOFOP]
Members
769 posts
6,881 battles

Priority 1 for Wargaming....ensure the maximum amount of matches possible.  Everything else is secondary.

That said, this comes up daily and ultimately, there is no accurate way to measure skill that all agree on so....this is what we get.

  • Cool 1

Share this post


Link to post
Share on other sites
1,233
[SYN]
[SYN]
Members
2,797 posts
10,295 battles

Yeah, the steamrolls are bad, but I don't think skill-based matchmaking is good.

Share this post


Link to post
Share on other sites
110
[WOLF6]
Members
320 posts
2,609 battles
47 minutes ago, CSV_defender said:

I've noticed a statistically unrealistic number of games where the teams are totally unbalanced in skill. 

What type of regression analysis did you run?  Are you running SPSS, or some other statistical analysis software?

  • Funny 2
  • Boring 2

Share this post


Link to post
Share on other sites
27
[SALTY]
Members
85 posts
5,013 battles
30 minutes ago, ZoomieG said:

What type of regression analysis did you run?  Are you running SPSS, or some other statistical analysis software?

Lol

Of course he is. Got a whole neural network built out just for this I bet.

On a more serious note, the games are not more lopsided because of CV’s. They are more lopsided because of the absence of good players. Snowballing effects are exacerbated when there are not good players in the game capable of reversing the losses of early ships.

Clan Battles sucks out a lot of good players and many who have left have taken their purple stats with them. Want better games, find better players. 

Share this post


Link to post
Share on other sites
5,048
[INTEL]
Members
9,822 posts
27,863 battles
1 hour ago, ZoomieG said:

What type of regression analysis did you run?  Are you running SPSS, or some other statistical analysis software?

Would you need a full blown regression? Chi square should be enough...

Share this post


Link to post
Share on other sites
1,228
[WORX]
Members
3,422 posts
14,785 battles

Once again the OP wants SORTING not Match making.

Sorting stats is a bad idea... Plenty of games used some sort of SORTING of stats to appease the few. Those game developers are no longer in operation.

Would the OP be complaining if the blow out was for his benefit ? NAAA

  • Cool 1

Share this post


Link to post
Share on other sites
3,718
[PVE]
Members
14,935 posts
9,477 battles
5 hours ago, CSV_defender said:

I've noticed a statistically unrealistic number of games where the teams are totally unbalanced in skill.  Game after game where the game is unwinnable for one side.  The winning teams take very few losses, and in many cases, none.  Many times the losing team hits zero before the game is half over.  Wargaming has access to player stats, so I suggest they incorporate this information into the matchmaking.  I've experienced more unwinnable games, regardless of carry, in the past 3 days than I've seen in 3 years.  Just because someone gets their hands on a T 9 or 10 ship doesn't mean they are competent to play it in a random match.  Also, having two divisions on one side and not the other is another problem.  I hope WG takes a look at this. 

Really? Because randomly assigning players can end up with unbalanced teams 2/3 the time and it's NOT unrealistic, statistically or otherwise. Let's look at how 8 good players of the 24 (8 good, 8 average, 8 bad let's say) in a match can be assigned. Ignoring how ship tier and type are used, there's a 2/3 chance of one team or the other getting 6+ of those 8 good players. We will call the teams Green and Red and here is all the ways how those 8 players can be distributed randomly.

 

G0 - R8

G1 - R7

G2 - R6

G3 - R5

G4 - R4

G5 - R3

G6 - R2

G7 - R1

G8 - R0

 

1/3 of the time Green gets 6+ good players and Red gets <=2 good players.

1/3 of the time Green and Red each have 3-5 good players

1/3 of the time Green gets <=2 good players and Red gets 6+ good players

 

So you can see that one team or the other getting 6+ players is very common. A player getting on a stacked team for his side will even out over time 1/3 of the time. 1/3 of the time, the player will have the stacked team on the reds. 1/3 of the time the teams will be roughly even.

 

Ever wonder how a bunch of bad players can end up on the same team? Substitute bad for good above and there you go.

Share this post


Link to post
Share on other sites
1,217
[OPRAH]
Beta Testers
5,389 posts
15,182 battles

@CSV_defender Quit using a MM monitor and don't bother manually checking others stats post game. That will solve your problem since you will not know others stats/skill level.

There will be no skill based MM so get over it or Bye!

Share this post


Link to post
Share on other sites
693
[TDRB]
Members
2,545 posts
7,337 battles

Dare I say deja vu? The same old complaint and call for stat based MM. And a repeat of the facts showing why stat based MM will not end blow out battles nor insure more wins.

 

Edited by kgh52

Share this post


Link to post
Share on other sites
563
[WDS]
Members
1,532 posts
8,270 battles
6 hours ago, CSV_defender said:

Also, having two divisions on one side and not the other is another problem.  I hope WG takes a look at this. 

This is a bit of a problem if the divisions are from the same clans not so so much if they division up before the game and don't know each other . I don't know if that could be addressed but I do notice the team with more clan divisions seems to be stronger not always of course .  But I'm always wary when I see a division of 3 all from the same clan and then another division of 2 from another clan that usually means trouble for your team if you have none . 

Edited by clammboy

Share this post


Link to post
Share on other sites
3,695
Members
22,445 posts
5,792 battles
6 hours ago, CSV_defender said:

Just got out of a random.  Our team got 1 kill between 12 players.  We hit zero....AGAIN.  This either stops, or I'm out.

Have you ever thought it might stop because you're out?....:Smile_trollface:

Share this post


Link to post
Share on other sites
275
[WOLFB]
Members
626 posts
4,982 battles
1 hour ago, Kizarvexis said:

Because randomly assigning players can end up with unbalanced teams 2/3 the time and it's NOT unrealistic, statistically or otherwise. Let's look at how...

We'll done @Kizarvexis. I never thought about it like that before. I am not enough of a mathematician/statistician to know if there are flaws with your logic, but this makes sense to me. And it feels about right too: probably a third of my games are easy wins, a third are very hard and we probably lose (but not always) and third are up for grabs.

So yeah that one third stacked against me feel rough but I just have to remind myself that it's good practice to face a greater challenge.

And I have to be honest about that other one third that is in my favor: we won partly because of match making, not because I suddenly got awesome at the game.

But that one third of games that are evenly matched: if I do play really well, then I win more than half of them and my Win Rates reflects it.

Seems like Match Making is working OK.

Share this post


Link to post
Share on other sites
3,718
[PVE]
Members
14,935 posts
9,477 battles
33 minutes ago, Jester_of_War said:

We'll done @Kizarvexis. I never thought about it like that before. I am not enough of a mathematician/statistician to know if there are flaws with your logic, but this makes sense to me. And it feels about right too: probably a third of my games are easy wins, a third are very hard and we probably lose (but not always) and third are up for grabs.

So yeah that one third stacked against me feel rough but I just have to remind myself that it's good practice to face a greater challenge.

And I have to be honest about that other one third that is in my favor: we won partly because of match making, not because I suddenly got awesome at the game.

But that one third of games that are evenly matched: if I do play really well, then I win more than half of them and my Win Rates reflects it.

Seems like Match Making is working OK.

It's an extremely simplified example just to show that a random distribution can be stacked one way or the other and be normal. Real MM would be more complicated.

Share this post


Link to post
Share on other sites
Members
4,383 posts
7,093 battles

Please see the hundreds of threads already created on the exact same subject and the issues with defining 'skill'.   Quote frankly the amount of going to zero has been reduced with the latest tweaks last year to MM.

Share this post


Link to post
Share on other sites
Members
2,420 posts
9,044 battles
2 hours ago, Kizarvexis said:

Because randomly assigning players can end up with unbalanced teams 2/3 the time and it's NOT unrealistic, statistically or otherwise.

You are more right than you know it.  Based on overall statistics we can see that the majority of the very best players in this game maintain 70% W/R, while majority of the worst players maintain 40% W/R.  We can thus deduce that on the average you will lose 30% of games regardless of your skill level, and you will win 40% of games regardless of what you do, leaving 30% of remaining games (or 1/3) where your personal skill level will dictate the outcome of the match and “move” your stats between the 40% and 70%.

Share this post


Link to post
Share on other sites
1,056
[NG-NL]
Members
5,719 posts
9,439 battles
8 hours ago, CSV_defender said:

I've noticed a statistically unrealistic number of games where the teams are totally unbalanced in skill.  Game after game where the game is unwinnable for one side.  The winning teams take very few losses, and in many cases, none.  Many times the losing team hits zero before the game is half over.  Wargaming has access to player stats, so I suggest they incorporate this information into the matchmaking.  I've experienced more unwinnable games, regardless of carry, in the past 3 days than I've seen in 3 years.  Just because someone gets their hands on a T 9 or 10 ship doesn't mean they are competent to play it in a random match.  Also, having two divisions on one side and not the other is another problem.  I hope WG takes a look at this. 

Unfortunate reality of F2P potatoes.

Personally WG could address problem simply by tweaking rewards system to be worse than co-op for players who have bad stats, but if they ragequit, fewer players for rest of us to beat up.

At very least, an option to report a player for sucking would be nice. If said player has bad stats, then it should cost that player some XP or credits, something to motivate him/her to get gud.

Bottom line is, there's no cure to the stupidity. Just laugh at them like I do.

Share this post


Link to post
Share on other sites
50
[SIDE]
Members
49 posts
16 hours ago, Navalpride33 said:

Once again the OP wants SORTING not Match making.

Sorting stats is a bad idea... Plenty of games used some sort of SORTING of stats to appease the few. Those game developers are no longer in operation.

Would the OP be complaining if the blow out was for his benefit ? NAAA

I am opposed to being on either side in a blow out game.  On the losing side the problem is obvious.  On the winning side the game is over before you have a chance to do damage or just enjoy a competitive game.  This situation has gotten out of control almost overnight.  I've never seen so many useless games.  I could care less if I am on the win or loss side, I just want to play with/against people who have a clue what they are doing.  Its never been this bad.

  • Cool 1
  • Boring 1

Share this post


Link to post
Share on other sites
17
[M-T-B]
Members
30 posts
5,043 battles

Behind all this MM talk … The game economy is slowing to a crawl.  Losing streaks are getting larger, credits and XP is reduced for the less than stellar player.  Tough on us that have moved up from the bottom third to mid pack.  No shame in saying we appreciate a carry player to prime the pump.  A wild guess is 1 out of 50 grasp tactics or teamwork.  Providing diversionary fire when DD's are capping is years off in the manual.  There haven't been many matches where the Red Team "isn't about to win" 5 minutes in.  Who are these guys?  

  • Boring 1

Share this post


Link to post
Share on other sites
15
[CEWL]
Members
102 posts
5,143 battles
On 5/4/2019 at 1:14 AM, ZoomieG said:

What type of regression analysis did you run?  Are you running SPSS, or some other statistical analysis software?

As a psychology research student this had me dying of laughter LOL. But MM does need some tweeking

Share this post


Link to post
Share on other sites
Members
89 posts
8,918 battles

I agree that skill-based matchmaking will not prevent blowouts - sometimes a collection of mistakes just happens.... But I do wonder if matchmaking has changed recently, either by changes in the system or just due to changes in player skill. 

My DD WR in high-tier matchmaking had been fairly decent - around 55%.  But after the last update, my high-tier WR has been below 50%.  I know that my performance has been off the mark - damage totals are down, kills are down, deaths-by-CVs are up.  But I wondered if matchmaking had a role to play.  Over my last 48 matches, the results are pretty depressing.

My WR when the reds had better average WR: 17% (5/29)

My WR when the greens had a better average WR: 79% (15/19) 

The median difference in WR was -8%.  That is, the reds had an 8% higher WR than the greens.  As you can see, I had 29 matches where the reds were better than greens, but only 19 where my side had the better WR.  Had  that 10 game mismatch been balanced, that would've taken my WR from 48% to 55%.    

In tracking these data, I've observed that there are 3 types of games:

1) Reds blowouts  

2) Greens blowouts 

3) Matches actually in contention 

The blowouts occur most often where there are larger differences in WR.  Over the last 48 matches, 25% of them had differences in WR of more than 20%.  That maximum WR differential between teams needs to be capped.  Then we can focus on what differentiates a good from bad player - how often can you influence the outcome in matches actually in contention.  The best players are able to expand what defines a match being 'in contention'.  If a difference in WR of 4-5% would ordinarily ensure a victory for the better team, a unicum player would be able to prevent more of those losses than a typical player (conversely, those 35% WR players negate any advantage for their team through exceptionally bad plays).  Clearly, I'm not able to overcome this -8% WR difference.  This is where I need to work on my game - how can I better carry teams?  Can I expand the WR differential that would otherwise have guaranteed a win for the reds?  But it isn't fun trying to tackle this problem when there are so many blowouts.  

 

 

  • Cool 1

Share this post


Link to post
Share on other sites
2
[ABIDE]
Members
8 posts
2,650 battles

The Random Matching Algorithm needs to be fixed. More than 95% of my games my ship is the lowest level 6 or 7 and half of the enemy ships are 2 levels higher. It's NOT FUN to be 95% of the time cannon fodder. The other 5% of the games there are only 1 level higher than mine. I NEVER got a game where my ships are 2 levels higher. 

How the algorithm works really? It seams to preference the big money.

Edited by todorp

Share this post


Link to post
Share on other sites

  • Recently Browsing   0 members

    No registered users viewing this page.

×