[MASTER] War mismatch - 136k vs. 108k war score (and other similar mismatches from other alliances)

We normally get pretty decent matches, but not this weekend. 27 per side. Our teams range from 3522-4549 Their low starts at 4323 (higher than our average) and tops out at 4703. Most are in 4600s.

I expect they will end up over 6k, we will be lucky to break 4K

Their alliance is 30d old, with most members under 2 weeks.

Edit: war scores — 85498 us vs 88152 them.

Associate win/loss to the players, not the alliance and calculate war score off the average win/loss of all the players for the last X wars. It’s not that hard.

1 Like

What makes a good or bad matchup? Yes initially, ANY change in the matchmaking criteria is going to result in mismatches. But it doesn’t mean you shouldn’t make the changes.

The only thing you can do to minimize this is to throttle the changes week to week, to blend the old matchmaking in with the new. Currently, the weighting is very heavy on win ratio. If you change that weighting to be more based on roster, you have to do it slowly.

The week you changed the matchmaking was one of our best wars yet. We got matched against a highly skilled alliance, instead of an alliance with 200-300 TP larger than ours. The score was very close.

1 Like

Is there an explanation of the goal here? Is it to just improve matchmaking in general, or is it to specifically counter the “new alliances” issue?

If the latter, I made a suggestion 2 posts above that requires some changes to the player data model, but not massive changes to the matchmaking algorithm.

If the former… that’s a much, much bigger challenge, and I certainly don’t feel like I can make suggestions without far more data.

1 Like

That’s pretty much exactly what we did when we started the project. Unfortunately the point difference of the wars was clearly higher than in the old model.

I’m planning to post an update on this topic on Thursday or Friday.

10 Likes

Hi,
We have the same issue in the war today.
I have screen shot to explain the issue regarding the score war game and the number of attack!
We have 52k against 58k as war score
We have 45 attack against 54 for them

So how it is possible? Tell me how we can have a chance to not loose?
I have try to open a ticket to the support but with the update, it’s less easy! Just have a chat in the game, no answer and no mail…
I try in the forum, hope to have an answer

Ps: sorry if I have made some mistake, I’m French :slight_smile:
Reps: not possible to put more than 1 media in the post as I’m a new member…
I can send all the information if needed

1 Like

Here’s another update.

We tried something completely different since my last post. This time we copied the alliance specific war history to every member of the alliance. In other words, we ran the matchmaking based on the old data (the war history of the alliance) but using the new implementation that reads it from the player history. After this initial seeding, we kept making matches using the new player-specific history for the past two wars.

So far the results are encouraging. Our metrics show that the matchmaking quality has been comparable to the old implementation. As a reminder: The benefit of the new implementation is that it solves the issue where multiple experienced players start a new alliance and get too weak opponents.

We’ll monitor the situation to make sure that the results don’t degrade over time.

21 Likes

Maybe I missed it, but what exactly did you do the last time? I would have expected that this is approach that you should have taken the first time around.

1 Like

Hi,
Regarding my previous post, I think you have forget to take into account an important thing: the number of participant in the war and so the number of attack!

In the last war, is it logical that the team have not the same number of participant? And so the same number of attack?
In my case, my team is weaker, we have 3 members less and we have less attack (18).
You have a bug in your match war program because you don’t take into account the number of attack of each team!

Except in our case. This new method lowered our war score below the opponent we just beat in war… Looks like we will get an unfair war once again as we are lower than some weaker allies now

1 Like

Thank you for the update! Is the new algorithm already live? It would be also nice, if you could share how many past wars is being taken into account. Is it only last 2 wars or more?

@mhalttu we have 2 players in our alliance that aren’t worth nearly as many points as everyone else. Our war score has been off for about a week with no help from you guys finding the problem. There is obviously a problem with the new algorithm with the individual scoring. Our best guess is these 2 players somehow had their 20 war penalties reset as they are each worth about 3k while rest of us are worth about 4.6k. This just happened in the last week and they have been with us for months now. Is it possible they got reset somehow by leaving after war to merc titans?? If so this is a huge bug that people will use to manipulate the new system. If what we think is true we will be facing much weaker allies than we should be as our war score is not accurate. We had a 140k war score then mysteriously lost 3k after winning our war. Today’s war we only gained 80 points or so which shouldn’t be the case when you’re only at 137k.

1 Like

The algorithm has been live in the past four wars. We keep track of much longer history than the past 2 wars.

4 Likes

Thank you for the report. I’ll have the team look into it and fix it. Stay tuned.

6 Likes

@mhalttu, my theory is that our members went to merc after our matchmaking was completed, but before ALL match making was finished. Since it takes a couple hours for all the matchups to happen our members could have gotten matched for a “2nd” war in a smaller alliance they were mercing at. If this 2nd war ended up being a loss it could negatively impact their individual war score?

1 Like

Yesterday we suffered the biggest mismatch ever, the opponent had
5 members at 4600
8 members at 4500
4 members at 4400
1 member 4200
In the other hand we had
2 members at 4500
2 members at 4400
everyone else is in the lows 4000s plus we are coming from a losing strick
I noticed that all members of that alliance had recently joined the alliance
We lost by more than 2000 points
Should I report this? I do have the video to prove it

2 Likes

Yeah, that’s what I’d do (if I didn’t write it, I thought it really loud!). And then run simulations of old vs new to ensure matchmaking for most alliances hadn’t changed.

The latter gets tricky, since you can’t just exclude any alliance that had someone join or leave in the last X days, you have to actually compare roster[today] to roster[last_war] because otherwise you exclude alliances that have people merc.

Once I was confident I hadn’t broken things THEN I’d let it accumulate war data for 20 wars to put EVERYONE on the same level. Only then would it go live.

With it live now without the history, we can expect “war alliances” to have an advantage for up to another 9 weeks.

Still, headed for improvement.

I think we have right now very unfair war. How this can be even possible? We have 29 members our alliance (nobody didn’t left or didn’t visit ours alliance)
And othet side, they have 30members, no ex-members on the list nothing!! So how this can be possible?

1 Like

They can’t always match teams with even players. And, unfortunately, you got stuck.

To give a simple example, imagine if there’s only 6 teams in the game, half with 30 players, half with 29.

4 teams get matched against the same number of players. The other 2… there’s no other option.

But yes, it’s unfair. And I think it would be better for there to be a game generated opponent in that case, but there’s issues with that too. They may resort to that if the very last team (a 1v1) has no opponent available. Don’t know.

We will see what SG answer to me :woman_shrugging:

Warscore right now is almost the same.

Cookie Settings