More evidence for broken Alliance War Score algorithm ... post-update


#41

The next match for Guardians Ascending is available. The GA war score is 178088. That of their opponents, Celtic Revenge (CR) is 175151, so clearly GA should be the stronger team. Oddly the titan scores show the opposite … 39193 vs 43014. So what do the respective defence teams look like?

The GA number 1 power is within a few points of the CR number 10. Here are the top 10 from each team:

CR: 4024, 3690, 3438, 3924, 3470, 3324, 3738, 3400, 3209, 3428

GA: 3437, 3478, 3509, 3464, 3466, 3299, 3159, 3291, 3158, 3109

The suggestion was made above that these teams, which make up half of the war combatant teams, do not reflect the relative war scores, because the “attack half” of the GA’s lower defence component more than makes up for this defence weakness by having a battery of 5*s in their 30 heroes, OR the CR with its much stronger defence half, has a way weaker attack component than GA. How likely is such a contrast?


#42

great question…which is why I go back to @Rigs comment that we should be able to see the scores used for matchmaking, so we can understand and either explain or flag things we can’t explain. I can understand SG not wanting to expose too much, but the publicly available war score used for matchmaking should be viewable throughout the duration of the war.


#43

Defense Team scores for the next Alliance War

Us Them

4090 4006
3938 3935
3912 3826
3594 3713
3543 3707
3453 3672
3443 3651
3433 3577
3432 3560
3426 3527
3415 3464
3112 3459
3085 3442
3074 3424
3031 3372
2990 3300
2934 3240
2813 3120
2761 2773
2733 2643
2669 2635
2655 2602
2587 2601
2471 2579
2280 2479
2186 2355
1987

We have 27 members, they have 26.

After the first three, which are fairly comparable, ours drops off faster than theirs. They have an advantage through 4-18, which would indicate that we would likely lose, but probably close. Yet our Alliance War score is higher than theirs: 186414 to 185578 (not a significant difference).

I know from previous wars that our participation is generally around 18 members. I haven’t tracked that so I don’t know which 18. Until I know which opposing members are participating compared to our participating members any comparison of Defense scores is meaningless.

Once the war starts I’ll post a new comparison of participants.


#44

The s̶c̶o̶r̶e̶s̶ team powers are so close that it’ll totally depend on the details of those low-end benches. My guess is the discount to the other alliance because you have an extra team is going to hurt you. The problem is, those extra flags are effectively always for the lowest TP player, which generally doesn’t add much to the war effort.


#45

Ok duh I can check my alliance participants in the battlefield. Senior moment.

Our top 10 participate, after that is 50/50.


#46

My whole point in that convoluted message is that comparing defense scores for war is meaningless since we don’t know if they are participating or not in the war. Unless you have a full alliance with all 30 participants you are never going to know until the war actually starts. Then you can make comparisons. Fully 1/3 of our alliance doesn’t participate at all; one of them is I believe in top 6 for cups even with a lower defense than 10 other members. Active raider and apparently good at it but not active in wars (which we are all fine with btw).

And, is the discount based on alliance members or participating members? We’ve had wars where we had more members but one less participant.


#47

Sorry. In that post, “scores” was supposed to read Team Power.


#48

Um yeah mine too power not score


#49

This entire block is fairly close:

2934 3240
2813 3120
2761 2773
2733 2643
2669 2635
2655 2602
2587 2601
2471 2579
2280 2479
2186 2355

Assuming you have approximately equal participation , I still think benches probably dominate.

I mean, I guess you could have a really idiopathic set of no-participates. But as long as it’s fairly random, and theirs is too…


#50

War has started: here are the actual meaningful numbers.

First two are us, last two are them. Last row is the average for each alliance.

Defense War Defense Defense War Defense
4090 4090 4006 3938
3938 3826 3935 3928
3912 3913 3826 3888
3594 3701 3713 3713
3543 3527 3707 3707
3453 3421 3672 3673
3443 3467 3577 3577
3433 3441 3527 3391
3432 3432 3464 3464
3426 3489 3459 3462
3112 3112 3442 3146
3074 3026 3424 3424
3031 2975 3300 3302
2990 3106 3120 3120
2813 2803 2773 2773
2761 2762 2635 2636
2471 2461 2601 2571
2280 2228 2579 2556
3266 3266 3376 3348

18 participants each side.

For reference, for the entire alliances defense power averages are:
Us 3076
Them 3256

By the numbers this war is quite even, much more so than you would assume from the full alliance numbers. This may be the most evenly matched war we’ve had so far.

What is interesting is more than half of the participants have enough of a difference between defense and war defense power to indicate a different lineup. I’m not counting those that went up a few because that could be from leveling a hero. I know that I do not use my 5 top heroes for either. My defense team rarely changes while my war defense team changes depending on the available boost.


#51

That does look quite even. Next hurdle is flag use. Good luck!


#52

We did a survey of 5* heroes in the Gardians Ascending alliance, the one that has now suffered 6 out of 8 war mismatches, loosinjg by at least 275 points in each of the six occasions. The majority of players have zero or 1 five-* hero at best. Of the 10 top players, five have responded so far. The number of 5* heroes owned by those responding are 2, 3, 5, 7 and 14. The latter player consistently has the top one or two scores in titan battles and wars. So unless the six teams (that have overwhelmed us so notably with similar war scores) have strong defense teams but relatively much weaker 30-hero rosters than we have, then the suggestion that we have rosters bloated with 5* heroes is untenable. I have to ask myself how likely that scenario of strong defence teams (which we can see) versus much weaker attack rosters (which are invisible) is.

Lastly are all six applying the same tactic of losing a string of wars so as to have a subsequent easy ride?

So if the war score algorithm gives equal weight to the two dominant components of wars, namely defence power versus 30-hero attack power, then there appears to be no obvious explanation for this plethora of resounding defeats. Are we simply poor players? Then how come we have very respectable titan scores and cup tallies?

I remain unconvince that the war algorithm appropriately reflects our situation versus 6 out of 8 of our recent opponents. As I previously posted, the others two wars were finely balanced and great fun.


#53

How big is your alliance? I don’t have any hard data, but I’m noticing that the vast majority of war mismatch posts are coming from alliances with 28 or fewer players. I’m wondering if the algorithms don’t do worse with matching smaller alliances…


#54

We have typically 27-28 alliance members but only 17-18 participate in wars. Most of our post-upgrade wars have been fairly even, those that weren’t occurred shortly after the upgrade.

I am presuming it’s the number of participating members that is important.


#55

Its actually both. Things get super tricky for the system with low participation, because the variance on what teams are actually going to show up is so high.

But the algorithm seems to do worse with fewer than 30 in the alliance, even with full participation. I don’t have an explanation as to why just yet. But it does seem to be a pattern.


#56

@ Bedu I know how you feel, as this has happened to my alliance also.
BUT I have been told that if your alliance is not full of 30 teams that participate in the war you WILL be matched with a team as close to you as possible in BOTH war score AND participating teams. As a example if you alliance has 25 active players in the war the team you will face will have 25 active players ± 2, is the way I understand it. If someone who knows more than me sees I am wrong PLEASE lets us know.


#57

That’s how it’s intended to work. Similar war score, and plus or minus 3 players maximum difference in participating alliance members. If something different happens it should be reported as a bug so the devs can see about fixing it.


closed #58

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.


opened #59

#60

@Garawyn @nevamaor Many thanks for the input. Over the past month I have continued to monitor the war scores and defence team compositions for the alliance Guardians Ascending. The pattern established in the previous thread continues. Approximately one third of matches are even and lead to exciting stimulating contests. The other 2/3rds - a distinct majority - are characterised by opponents with war defence teams overwhelmingly stronger than ours, resulting in losses by between 250 and 600 points. To summarise previous conclusions:

  1. We have never had an overwhelmingly weaker opponent. They are either on a par with us or significantly stronger with little in between. In other words this is not a broad bell curve, but a matching skewness/kurtosis.

  2. We consistently have trophy tallies similar to or just below our opponents and frequently have titan scores exceeding theirs. In other words the war wipe-outs cannot be ascribed to poor tile play or team selection. In fact we have a programme of tutorials and defence team reviews in place.

  3. The imbalance is seen by comparing the defence team powers. The suggestion was made that the explanation must therefore lie in the other half of the forces, namely in 30 hero rosters that are bloated with unlevelled 5*. So we did a survey. Most of our players have one or zero 5* heroes, so a bloated hero roster is not the reason.

  4. Another suggestion was that our opponents had been deliberately losing wars so as to created an “easy run”. This is counter-intuitive. Another suggestion was that they might be new alliances. Since most have players with over 100 days membeship, this is also unsubstantiated speculation. And why sould 10 out of 14 opponents deliberately lose?

@Garawyn made the suggestion that the imbalance might have to do with a low numbers of players. Indeed, we usually vary between 26 and 28 players, but at times have been up to 30 for weeks at a time. The latest war data follows.