Serious matching problem with alliance wars

We have only had one close match up wince the start of alliance wars, and it isn’t fun no matter who was the stronger. But I have not seem a match up this bad yet. You have to go half way down their roster to find a team with a power similar to our strongest team. There top 4-5 teams are maxed out 5 stars…most of our team doesn’t even have 5 stars and none of our 4 or 3 stars are maxed. This isn’t a difficult problem to solve, please do it. I’m just going to tell my alliance not to populate their defensive teams until I hear this is fixed.

1 Like

Misleading thread title! You called it a “serious” matching problem with alliance wars, then proceeded to provide a very un-serious analysis.

Really? By all means, go ahead and explain how.

After 3 months or so the matchmaking is still garbage, no improvements in other areas regarding a war leaderboard, war log or something to make wars a challenge. Already players lost interest in this feature. Last 2 wars was like a friendly match, nobody have a tactic and nobody loose time to make a strategy. Players start to give a sh… on wars and that is a big problem for the devs who are sleeping in their shoes… Another update is coming, let’s hope they do something in this one…

You might be right. Or Alliance Wars might be a nice little feature that is keeping players interested in the game.

2 Likes

They are most certainly not sleeping in their shoes (I like that image though).

There have been at least a half dozen updates specifically designed to address this issue and they have communicated these changes to the community via these forums.

This clearly shows that they recognize that war matchmaking is a problem and one worth solving.

It is most certainly not an easy problem to solve and as @Brobb offers the OP, sharing the easy solution would be a lot more helpful than stating there is one. Before sharing them though I suggest browsing the numerous threads on the subject to ensure you aren’t wasting your time suggesting a solution that has either already been tried, or rejected for well founded reasons.

2 Likes

So you don’t consider a issue where the code doesn’t meet one of the primary stated goals isn’t a serious issue. Good to know, Hopefully you aren’t a dev.

I have been lightly tracking this and can say some of the things they have tried would obviously not work. The only meaningful indication of capability of an alliance is the individual team power and size. Cups, ability against titans and all the other things they look at is meaningless. Granted even power is a flawed as a measure, but it is closer than the other factors. Take a weighted average of all the teams power (weighting to the top members) look for statistical outliers and give additional weight to super heavy hitters that are outliers. Only match those that are within a few % of each other for total (not a hard number). If they have tried this, I’ve missed it.

I’d argue that titan score is a better estimation of alliance strength than hero power levels, but not that much better. Titan score at least takes into account the collective ability of the alliance’s members to do damage, but it doesn’t account for bench depth, raid proficiency, defensive strength, or the fact that item use can skew results.

Power levels might be a rough estimate of bench depth if it weren’t for three inaccuracies:

  • Different star levels can’t be compared easily. For example, my completely unleveled Thorne has roughly the same power level as my 3^50 Valen, but they’re vastly different in effectiveness and presumably treated the same way by the matching algorithm.
  • Adding hero powers together to get team power is misleading because certain hero combinations and formations have better synergy than others, which skews the accuracy of the team power number (this alone is likely the reason for half of the raiding complaint threads I’ve ever seen).
  • Some types of heroes (i.e. healers) work far better in AW, but this also isn’t accounted for by power level.

With all the variables that factor into success, an ELO system similar to the one used for raids is probably the only way to effectively measure war power, but even that will update too slowly for it to work properly. I don’t think there’s an easy answer that will solve all mismatches, but maybe we’ll eventually get something that at least minimizes them.

The two issues with titan damage is you can greatly impact it by battle items and if you have low enough power members and high enough titans you are fighting, their damage will not even register due to near instant death. Neither of these thing are true in AW. I agree with you summary of the issues with power, and have some additional ones. But when looked at broadly and not specifics it tends to work.

There are 3 large problems with this model that will still lead to mismatches:

  1. Total power of heroes is a truly terrible indicator of the effectiveness of those heroes in a war setting. The biggest issue is that the star level of a hero comes with a total power bonus that FAR exceeds its value until the 5* are fully ascended. This could potentially be corrected for. However other issues such as varying special skills having more or less value in a war setting (often dramatically - Alby Vs. Horghall anyone?).

  2. It does not take warring skill into account. Things like good hero development strategy, good team selection, good gem play, etc. will leave some alliance perpetually behind. Maybe this is ‘fair’ in the sense that good teams will beat bad ones. But it sure sucks for those bad teams. It will always feel like they were screwed by the match-ups.

  3. Most importantly, it is not self correcting. Teams that end up on the short end of the stick of this system (because of the factors above) will almost always lose.

I hope we end with a ladder based system where teams have a war score that is a sum total of individual player’s war scores on the team. War scores could be calculated as a combination of hero strength, individual war performance and team war performance. It could be a hidden metric or a visible one although I’d prefer hidden so you don’t introduce the concept of individuals choosing self-interested tactics in war.

All that said, it’s not as easy as just putting that in place. There are a number of undefined pieces that would need to be ironed out. SG is making the plans. I am choosing to go along for the ride.

1 Like

I kind of agree with you for #1. I can say with my mid-leveled 5 stars, the power rating seems about right based on success in raids. I totally agree with the “not all powers are equivalent” statement, but again broadly it works. And in statistics its all about working broadly, not in specifics.

#2 as you say, is the point of wars. line up two teams that are close to equivalent and let the team with the superior individual players or group tactics win. I don’t think participation prizes are a good thing.

I like you addition for #3, I would do it a bit differently, but as they start to gather enough data to make statistical statements about certain hero’s or combo of hero’s having a strong correlation to beating the odd, they can tweak the algorithm. That would be a very nice refinement indeed.

Um… yes? No? I’m not sure?

What “primary stated goal” do you think it is that the code does not meet? Are we supposed to guess?

Hint. Look at subject line.

So what’s the “problem”?

That’s a smart and sensible process. I don’t think it’s likely to reduce the number of alliance war mismatches - I suspect it would actually increase the number of mismatches.

Please elaborate on why an increase. It works in college football.

Step 1. Tallying the number of players at level 12 would have some effect, I’m sure, for beginner alliances, but only for them. I suspect most alliances will have all players well beyond level 12. (I might be wrong about this - I can’t see the data.) We might just as sensibly tally the number of players in each alliance at level 20, or level 30 - or maybe I’m erring int he wrong direction, and we should tally players past level 5. It’s all arbitrary and unlikely to be any use outside a small subset of alliances.

Step 2. Using Top 5 Power as a proxy for player strength is not an awful idea. I personally prefer it to Top 30. But it should definitively produce more uneven matches, because it is specifically pairing players with roughly even defences, but radically different benches. (Evenly matched benches will sometimes occur, but under the process you describe, they will be freak occurrences - rare flukes.) The outcome of these wars will always be that the stronger benches win. (I don’t think this is a bad thing entirely, but it won’t make mismatches less common.)

Step 3. To say “match the teams within these groups” is to avoid answering the question: How? If I knew how to do it in such a way as to avoid mismatches, then I would post the explanation here to prove how clever I was. I suspect others would, too, for more selfless reasons. How to match teams is the essence of the problem and your method doesn’t address it at all.

So your suggestions are perfectly logical and sensible, but they fail to address the essence of the problem - how to match alliances - and don’t seem in any way to decrease the likelihood of mismatches. E&P certainly is not college football, though it’s worth noting that that sport is famous for mismatches.

The level 12 is the starting point for eligibility for playing in the war. That is where you start counting the number of eligible war players. All the way to level 60 or whatever is the highest level attainable or hasn’t been attained.

If two alliance have equal amount of eligible (level 12 and above) players, all top 5 heroes’ power of every player is added up, and the totals are near equal between the two alliances, then that should be pretty equal teams. These numbers change and so would the matchups. The depth of players will be pretty equal as well. Sure there are those that got lucky and have a few five stars and then mostly threes. But most of us have have a 30 hero roster that is top heavy but very close on down.

Well a new update is coming so we’ll see if it’s fixed. They probably wouldn’t take our advice even if we wrote the code out for them.

Also, have you ever seen Div I play Div II or Div III teams in football? Even Div I is divided into two; FBS and FCS. Then they are divided by conferences as well. The matchups within conferences was what I was referring to. Two teams ranked 7 and 8 nationally and are in the same conference are usually good hard fought games.