Most unbalanced ever...

[YouSatInGum]

I have MM monitor and run it most of the time (although often to look until after the battle) , so wanted to let you guys see what it looks like when the matchmaker really gets it wrong.

Usually when I see a 3% difference or greater in account WR the higher team wins 90% of the time.... 4% - 95% of the time.  Here we have over 10%.  Since we are talking about aggregate averages here, that is a HUGE amount.  In fact, this should be an extremely rare occurrence, so guess I was "lucky" to see....but not to be on the good side. 

I'm hoping that one day, even if they don't have an active skill lever matchers, they will at least implement a filter to catch gross mis-matches like this one.  If it catches one, swap 2 or 3 players from one side to the other.  Problem fixed.

...well, one can dream, right?

In case you're wondering, this one was over in 8 min.... but was actually over at 3 min mark already down by 6 ships at that point.

[CriMiNaL__]

So what you are asking for is skill based match making?

[Old_Baldy_One]

Here you go - I'll do you one a bit better (46 v 58 if you can't see it - colors are PR based on this one):

Then there was this one......(colors are account WR based)

[Grey_Paladin]

Something strikes me as very wonky here, but it may just be me not knowing MM Monitor.

Account Battles - is that the total for the account, or is it the number in the tier, or something else?  I see a player with 40 battles in a T7 ship.  Another has 124 total but 54 of those are with a T7 ship.  Am I misunderstanding something here?

 

[KaptainKaybe]

Honestly, while I was dead set against a skill based matchmaker for a while, I'm starting to change my mind on this. The overwhelming majority of games I've seen in the last year and a half have all been massive one sided steamrolls, and the game is just really not that much fun anymore because of it. I like being challenged, not just watching half my team melt in the first five minutes and realizing there's nothing I can do to salvage a win from this.

[Shigure_DD]

Stats can be easily abused divs/ rerolls etc.

[Wanturskul]

I had a guy on my team last night running that.  Before the match even started he was going on in chat how we were going to get rolled and why play.

Me in my Benham and a couple cruisers rolled our flank, turned and smashed the pushing red flank.  We won pretty good.

After the match went to look at the loudmouth's stats...  HIDDEN.

So I cant put too much faith in that mod.

[GoaJim]

22 minutes ago, Wanturskul said:

I had a guy on my team last night running that.  Before the match even started he was going on in chat how we were going to get rolled and why play.

Me in my Benham and a couple cruisers rolled our flank, turned and smashed the pushing red flank.  We won pretty good.

After the match went to look at the loudmouth's stats...  HIDDEN.

So I cant put too much faith in that mod.

How do you know he was running that? Because he said so? You believe the loudmouth? Is it possible that a loudmouth might misread the numbers?

Maybe you just shouldn't put your faith in the loudmouth.

[BrushWolf]

58 minutes ago, KaptainKaybe said:

Honestly, while I was dead set against a skill based matchmaker for a while, I'm starting to change my mind on this. The overwhelming majority of games I've seen in the last year and a half have all been massive one sided steamrolls, and the game is just really not that much fun anymore because of it. I like being challenged, not just watching half my team melt in the first five minutes and realizing there's nothing I can do to salvage a win from this.

I am still against a true skill based MM but I am open to skill balancing within the ship types. Divisions would be based of the best player in the division.

[KilljoyCutter]

I'd be OK with a bit of "soft" skill-based MM at the end of the MM process. 

After MM builds the teams, make a final pass and look at the "player rating" (whatever you set up for it) of the players on both teams -- if the disparity between the teams exceeds a threshold, MM picks two players to swap, then checks again, etc, until the disparity is under the threshold. 

By having a significant threshold, you'd be addressing the most egregious mismatches without MM spending a lot of time fixing minor mismatches, and without the problem of "everyone gets pushed to 50%". 

 

[KaptainKaybe]

13 minutes ago, BrushWolf said:

I am still against a true skill based MM but I am open to skill balancing within the ship types. Divisions would be based of the best player in the division.

Or the average? There are many many divisions where you have one really good player carrying a really poor friend.

I'm not sure how to best approach a skill based MM myself. I'm thinking could be based on solo win rate when using the same ship type in +1/-1 tiers. With a possible downgrade of equivalent skill if the player is using a ship type and tier they are very unfamiliar with, for example a dedicated battleship player sailing in a brand new Kidd they just picked up. But yeah, actual stat nerds could maybe find a solution here.

[CriMiNaL__]

14 minutes ago, KilljoyCutter said:

I'd be OK with a bit of "soft" skill-based MM at the end of the MM process. 

After MM builds the teams, make a final pass and look at the "player rating" (whatever you set up for it) of the players on both teams -- if the disparity between the teams exceeds a threshold, MM picks two players to swap, then checks again, etc, until the disparity is under the threshold. 

By having a significant threshold, you'd be addressing the most egregious mismatches without MM spending a lot of time fixing minor mismatches, and without the problem of "everyone gets pushed to 50%". 

 

I think SB MM should be a thing, but I doubt it would stop blow outs, at least at the end of the match no one can complain the other side was better than them, even tho they were.

[Wanturskul]

47 minutes ago, GoaJim said:

How do you know he was running that? Because he said so? You believe the loudmouth? Is it possible that a loudmouth might misread the numbers?

Maybe you just shouldn't put your faith in the loudmouth.

I know because he stated he was.  And no I never listened to people that concede defeat before the battle even begins. 

The mod can be used as a crutch for why certain, not all, players perform badly.  Do you see professional sports team quitting before a game because the opposing team dominates them in the pregame stats?  

Play the game, let the chips fall where they may.

[NashW8]

You guys get this all wrong. I don't want the bad players gone. I want the unicums to be doing something else while I'm in randoms. Like ranked :)

[AADusNuffin]

2 hours ago, YouSatInGum said:

I have MM monitor and run it most of the time (although often to look until after the battle) , so wanted to let you guys see what it looks like when the matchmaker really gets it wrong.

Usually when I see a 3% difference or greater in account WR the higher team wins 90% of the time.... 4% - 95% of the time.  Here we have over 10%.  Since we are talking about aggregate averages here, that is a HUGE amount.  In fact, this should be an extremely rare occurrence, so guess I was "lucky" to see....but not to be on the good side. 

I'm hoping that one day, even if they don't have an active skill lever matchers, they will at least implement a filter to catch gross mis-matches like this one.  If it catches one, swap 2 or 3 players from one side to the other.  Problem fixed.

...well, one can dream, right?

In case you're wondering, this one was over in 8 min.... but was actually over at 3 min mark already down by 6 ships at that point.

yeah, that happens occasionally and it can be pretty frustrating to see that huge gap.

[USMC2145]

1 hour ago, Shigure_DD said:

Stats can be easily abused divs/ rerolls etc.

Agree, also in the pick even the Purple account win rates were players who do not have enough games played to really judge. But, it is scary the amount of reds. 

[Admiral_Thrawn_1]

2 hours ago, CriMiNaL__ said:

So what you are asking for is skill based match making?

I would only ask that if you have 6 highly skilled players and the rest are clueless with the stats to prove it on both groups of players that the good players get even distributed between the 2 teams instead of most or all the good players sent into the match being sent onto one team. And that is my opinion if I am on the favored team or the other team. Closer well fought battles tend to be more fun after all. And this idea is a lot easier than full blown skill based MM.

[Umikami]

3 hours ago, YouSatInGum said:

I have MM monitor and run it

This is the problem; you're paying attention to this nonsense and defeating yourself before the match even starts. Play to have fun, and you'll have fun.

[Stew_Pedastle]

2 hours ago, Shigure_DD said:

Stats can be easily abused divs/ rerolls etc.

Not the stats that matter or what MM shows you . 

[Admiral_Thrawn_1]

3 minutes ago, Umikami said:

This is the problem; you're paying attention to this nonsense and defeating yourself before the match even starts. Play to have fun, and you'll have fun.

Yeah I had an awesome battle with tanks and looked at the in game stat profiles in the post battle battle report and was shocked that a I had been the only good player on my team by WR and end my players WR were all good. But I had not known that during the battle and made the fact I had won the battle all the better lol.

[Navalpride33]

3 hours ago, YouSatInGum said:

have MM monitor and run it most of the time (although often to look until after the battle) , so wanted to let you guys see what it looks like when the matchmaker really gets it wrong.

Random is Random, in all honesty there is nothing wrong with the pic shown...

%90 percentile of the player population is concentrated at the %45-%55 WR block...

%5 percentile of the player population is located at %56 and above, Finally...

The other %5 percent of the population is located at %44 and below...

---

Now that said, Random will pick SHIPS waiting in queue... It also means the %45-%55 block is more active then the other blocks...

While the other two remaining blocks play less compared to the %45-%55 block... So what to that means for MM... It means that if you have say 13k people online...

%90 of them will be %45-%55 block or 11,700 people..

The other %5 or 650 people each... Will come from the remaining blocks...

MM will only go by ships... It doesn't discard the two  %5 percentages.

Further divide that with the number of game modes we have an well we get the number you have in the pick... The good thing about your pic was, its Random... You will not have the same team on the same side the next time around...

But considering the odds based on what I stated... Its really random chance to go twice in a row with a similar team...

[cecill611]

With skill base MM how do you count for the players that hide there stats?

[Rollingonit]

2 minutes ago, cecill611 said:

With skill base MM how do you count for the players that hide there stats?

Hiding stats, doesnt hide it from the server :)

[cecill611]

Just now, Rollingonit said:

Hiding stats, doesnt hide it from the server :)

Thank you ..I knew that just not thinking.

[YouSatInGum]

2 hours ago, KilljoyCutter said:

I'd be OK with a bit of "soft" skill-based MM at the end of the MM process. 

After MM builds the teams, make a final pass and look at the "player rating" (whatever you set up for it) of the players on both teams -- if the disparity between the teams exceeds a threshold, MM picks two players to swap, then checks again, etc, until the disparity is under the threshold. 

By having a significant threshold, you'd be addressing the most egregious mismatches without MM spending a lot of time fixing minor mismatches, and without the problem of "everyone gets pushed to 50%". 

 

That is exactly what I'm referring to.  The fact that someone posts within a few min of a game with even more disparity makes me wonder if the skill disparities are happening more often than one would expect if the MM was truly "random."

Of course, it is not really random, since tiers, ship types, and several other factors.  Then there are the other MM "rules" that aren't public.  I am nearly convinced the the MM avoids putting Friesland together with CVs.