Alliance Wars Matchmaking (Discussion & Developer Response)


#2689

I do think there is less data here now because matchmaking has significantly improved- fewer folks have a reason to post now.
That doesn’t mean its perfect.
For our team the matching is much improved. You might criticize the initial roll out as being underdeveloped but SG has certainly made attempts to improve.
My first post here is #191 back in May of last year. I posted stats from nearly every war, often from both my main and my alt account, because when we moved to top 30 hero matching we got smoked almost every war. Under titan or alliance matching we won and we lost but it always felt competitive.
Now that they have adjusted 5* hero power ratings down, added troops, and weighted top 5 hero’s higher it’s better.
We have an oddity in that we almost never-like 1 in 25- have a higher average defense than our opponents.
That doesn’t tell the whole story bcs as others have said you can’t see behind the front team. We have beat opponents that average 165 more than us on D and just lost to one that only averaged 24 more.
I still track every war and will post the most recent below. Use the data to determine which strategy (We have 3 different ones) we will use. We lost this by 400 and they left 20 flags on the table - we used all.
5* means the number of 5* heros at level 80. Troops means the number of 4* at level 11 or above (level 11 being the first that reduces tiles for mana when using mana troops).
This was an odd match only in that we lost the war before this one and this was a harder team in terms of defense, so our match didn’t get easier after a loss. We will see what the next brings.


#2690

That’s awesome… Wish you wouid have tracked player level variance as well to see how that factors into the matches you win vs the ones you lose.

Compare to when you see a perfect match.


#2691

Our alliance once lost 23 times in a row (prior to the current war system). War mismatch was certainly a problem in our eyes (we could point to a consistent difference in war Def scores between our and their alliances)…

If one wants to say no fair—and be heard—bring data. Best way to answer the question.

As always, keep in mind:

Alliance War Matchups are currently based on:

  • Top 30 heroes of both alliances
  • Top 5 heroes (weighted)
  • Top troops
  • Total number of players of both alliances
  • Wins/losses record of both alliances

It is NOT based on:

  • Alliance Score
  • Titan Score
  • Trophy/Cup Score

#2692

Could you please drop the player level issue.

Some time when you are running out of arguments you bring it up again.
Then when people show you it cant be used for pairing or estimating player strength it is not the only factor, and so on………

Then Kayo checked your data and found it contradictory to what you reported.

We know pairing is not perfect, ( especially not in your case it seems ) but please refrain from bringing up player levels since you have ben proven wrong in this thread.

Better luck with your next war Pairing.


#2694

Keyo did not even address levels. And actually showed that we were mismatched by -100 across all 24 teams. Then asked if that was a mismatch.

He also admitted he used alliance data not actual war data… Admiitting it was likely off some.
Not sure why you’re are so insistent in hiding the truth.


#2695

Just a brief reminder to keep posts respectful and civil, as per Forum Rules.


#2696

Happy to address levels, but first:

What correlation do you believe levels to have an effect on War matchups?


#2697

This post was flagged by the community and is temporarily hidden.


#2698

They play a part ina few ways… until you hit a certain point is my belief…

It’s actually too much to say again in one post… and I’ve already pointed it out… Along with other common sense things that come along with leveling.

But in simple terms it pits different levels against each other without taking a consolidation of power into account…

Not overall power within an alliance… in a more refined manner…
so you end up with a level 40 facing 10 level 40’s. Which is 10’s of millions in experience…
it allows for 3 levels of power to be matched. Yet the true match is only one #…

Your alliance would not see the type of mismatch I refer to. As y’all are all so developed.

In essence it can easily match a lvl 58 with a 15 5* bench…

Against a 35 hodge podge bench and see it as even if they are only carrying 15 4* and 15 2* feeders…
give or take… as the math would look the same in top 30 power
As one only keeps the best fished 15

The other is still working up to fill the 30 in a far different approach


#2699

You know, I was a skeptic about the “player level” argument, but I think there is some merit to it. Here’s why:

Suppose two players have similar top-30 heroes, so from a war-matching perspective they are equal. But success in war isn’t really just about top-30 heroes, but about the whole bench. I often dip into my 70-plus roster below the top-30 heroes to bring in a color stack or a good skill. (Rigard, for example, is not in my top 30.) that depth is correlated with level, and the beyond-30 depth is not measured by the current war score.

Frankly, I think there remain some issues that contribute more to the imbalance. In particular I would point to the gap between an alliance with a few very powerful members leading a group of weaker players, versus an alliance with more uniform mid-tier teams.


#2700

This post was flagged by the community and is temporarily hidden.


#2701

Even more obvious than that is when I was mid-high 20’s but had no food I had perhaps 15 full and partial 3-4* hero’s…

But the main issue was no food so I would sit on a ton of feeders…

It wouid count those 15 as my top 30.
While a higher level player usually has a true 30… a 50 has a full-full bench… where a low 40 would be in the center of the two usually… Given natural progression.

So matching them is a flaw. Especially when adding a level match on top of current, can only result in a better match… and is impossible to draw a match backwards.


#2702

lurks :wink:

20 warnings to follow Forum Rules


#2703

In theory you’re saying that you’re not using your best heroes but still succeeding. I don’t know if that’s a leveling thing, but a good knowledge of how to play, and a deep bench.

While that level of User experience may help you succeed, realistically you could provide an alt account who’s experience is beyond its “level” to thrive. And your roster could potentially be equally achieved by Tom, Dick, or Mary if they have deep pockets - not leveling.

In the case of the single example that’s been provided by balustic, this is the info on levels:

Average level
His alliance (HA): 31
Other alliance (OA): 36

TP per level
HA: 110.3
OA: 100

Lowest TP per level
HA: 91.8 (level 40), 3,670 TP
OA: 75.7 (level 32) 2,421 TP

Highest TP per level
HA: 136.8 (level 26), 3,556 TP
OA: 144.3 (level 24) 3,462 TP

HA highest TP holder
level 35
OA closest in TP to HA
Level 46

Members level 20 - 29
HA: 11
OA: 7

Highest level
HA: 41
OA: 49

HA, 3 people at level 26, TPs
3,556
3,112
3,002
(554 point swing)

OA, 2 people at level 24, TPs
3,462
2,794
(668 point swing)

Highest and lowest TP level
HA: 35 and 22
OA: 44 and 32

Lowest TP from either side
OA, level 32, 2,421 TP (1 level higher than HA average level)

If there are any numbers that I’m not thinking about, let me know and I’ll run them.

It would also be helpful to get alliance info of his most recent matchup.


#2704

Matching is done on hero power. A 1/1 Grimm is less than half as powerful as a 4/70 Grimm.


A maxed 3* is more powerful than a 1/1 5*:

So having unleveled heroes waiting to be fed shouldn’t skew results.


#2705

Exactly so you see the problem there then…
one has true power… like it would be seen on a mature team… The other simply goes as a penalty against the lower player that has it as part of his 30 as pointed out above… either still in a learning curve or hold for the future to actually build. But clearly one is a scrub the other a hero. Yet both go to the 30…

Great example…

But you miss understood on the 15 counted against the 30…

Say I only have 15 5*… as my total team…
that then would find an equal in the way I explained… by finding a lower player that just has an equal power total… could be all feeders… and 5 3* that make up the 30…

A zero vs a 30 is equal in experience points in comparison
To a 40 vs a 58…
both see a 9 million point disadvantage…

But some would think a lvl 1 vs a lvl 30 was unfair… no matter the power total, as common sense would tell you if they match in power, it’s still not equal power.

That is the current system… power is power is power = match…


#2706

I think this quote summarizes your argument perfectly. Despite both sides being equal… they’re not really equal. To agree with you, I not only have to accept that 1 is not really equal to 1 - but also that it’s somehow just simply true, or in your words that it’s common sense.

Ideally they would take into account more than just the top 30… but that could make things worse for players with smaller benches. I’m just not convinced adding levels into the mix will make things better - it’s super easy to game. An alliance that avoids leveling (by for example exclusively farming low xp stages) will end up with a huge advantage over other alliances in the long run.


#2707

Wow that shows it perfect…
the top consolidation of power…

So a team like yours simply has 5 teams at the very bottom that pads the bench with low players…

Those bottom 5 then opt out of war… and the lopsided
top 5 crush the rest of the team… as they have the power overload…

I hope you continue to look at the effect the mismatch on the top levels plays, especially when the bottom drop off…

Unless the tops drop no equal match can ever occur anyway… That the problem…

That’s why the tops must match…


#2708

EDIT: This is in response to Balustic.

To summarize ALL your posts, I think what you’ve been saying is - there will never be a perfect matchup. And to that I agree. To that extent, If we’ve come to that conclusion, the next steps are either to deal with it, or opt out.

Unless you have an algorithm that will perfect the matchmaking system, in which case I’d recommend putting that in the ideas & features thread. I’d vote for a perfect system, if you have it.


#2709

I’d vote for an ELO based system - taking previous wars into account was a great step in the right direction and imo is the only “fair” solution. That said, I’m satisfied with the current matchmaking and don’t see this as a high priority issue.