Warscore Manipulation (not matchmaking)

Just to clarify: you weren’t in error in coming up with more than 1500 pts per side. That’s super common due, I believe, to rounding in assigning points and bonus points to teams. It was simply the magnitude of the points above 1500 in that second war that made me wonder if maybe there was an error that had crept in.

I don’t know that I would say anyone is actively researching the question of matchmaking per se, but there are several of us who try to understand what is implemented and some of the implications.

In addition to Kerridoc whom you tagged, @General_Confusion is probably the preeminent expert on the subject. @Gryphonknight is also very versed in the relationship between wins/losses, roster and war matchimg. @DBC has been exploring the question of whether there might be some oddity in how hero power and troop power interact in matchmaking.

Some relevant posts that you might find useful:

I’m also happy to talk through my understanding of the strengths and weaknesses of matchmaking.

2 Likes

@Garanwyn I assumed that when teams took more than 1 flag to kill, then by the time they are dead the attackers gain slightly more points (in total) for the kill. I have yet to find the time/energy to collect data, but this seems likely based on observation (perhaps due to teams healing)

I would love to get a firmer grip on the exact mechanics of war scores vs fair matches but there are so many variables :frowning: Some are impossible to measure i. e. player skill and board luck. If anyone can find a precise way to measure that I want to know!

I do think that simply comparing strength of visable defence team strength is a poor way to judge “fairness” But it does seem that the closer to 30members an alliance is the more likely a match is to be reasonably close (in end score)

As a group, our alliance has been through a huge learning curve, and I can assure people that group strategy and high flag usage should be examined before shouting “unfair”. Having said that, we have seen plenty of wars where the match seemed heavily weighted in their favour (one we won due to their low flag usage, so I can assume their history has lowered their score a fair bit) One “fair” match we lost because their strategy was poetry in motion… and ours wasn’t :wink: But I have yet to see a war where any obvious imbalance was weighted in our favour. But logic dictates that some alliances must get easy matches (or there can be no imbalance)

Losing a close war is frustrating, but loosing by a landslide is simply demoralising, and if there is any way the algorithm can be tweeked to avoid the later, I want to know what it is so I can yell about it :smile:

1 Like

I track all our wars, so I actually have very extensive data on this. Even multi-flag kills usually equal the reported total. Occasionally, there will be some rounding in damage that will up the total by a point or 2.

An example from our last war:

Screenshot_20190406-021401_Sheets

Kilroy77777 is worth 51 points. You can see the first two flags sum to 51 (44+7) for a kill. The next flag is a one-shot for 51.

Kübel is worth 54 points, but a little rounding has driven the first two-flag kill up to 55 points (25+30). Subsequent flags are unaffected at 54 points.

I completely agree on both points. Smaller alliances seem to get more uneven matches, and I’m really unclear as to why.

2 Likes

Thanks for that, when we clear the board at the start of the war, we always end up with more than 1500points which was something I was working up to investigating in more detail :slight_smile:

1 Like

You’re welcome. Of interest probably, though, is that the teams always start with a bit more than 1500 points. Our last 3 wars, the points on the board before first kill were:

1528 vs 1526

1525 vs 1530

1522 vs 1527

All for co-sized alliances.

In beta, I have done a 1v1 war, and the points were 1500 exactly for each “side.”

1 Like

Because math can be used to take advantage.
I’m gonna reveal my trick hoping devs do not fix it :grin:

I made a new alliance with an alt account which i do not level. It stays at lv 12 with not leveled heroes and make him partake in wars.
My main account is over lv 60 with several legendaries.

The matchmaking seems to force a focus on other 2 member alliance before anything else (i always faced 2 opponents, never more until now) and then try to match the strenght of both.
It result that i pretty much always face two lv 30-40 players with a not so great bench.

My alt account of course get hammered everytime, but his value is about 300-400 points (out of 1500) and there is a limit on how many times he can respawn.
So at some point they must attack my main account, and most of the time, get crashed and waste attacks.

On reverse i have zero problem of one shot their stronger defence, so i usually go for a 4-2 (4 times the stronger, 2 times the weaker) which result in more then 4500 points for sure.

So the math is tricky.
We probably have similar strenght if we summ all the heroes, but i can take advantage of the reset function, while they cannot.

This way i’m up to 20 wins straight in war.

8 Likes

Thanks for sharing this! It aligns with what several of us have taken to calling a high-variance strategy. I believe @General_Confusion made a proposal in APL which would have addressed the underlying issue. But I don’t think he got any traction.

4 Likes

History of War Matchmaking

Their are several obstacles in your way ( I should probably write a history of war matchmaking post ).

Many many many many players complained due to broken war matching so lot of debris from previous attempted fixes making search unusable ( nature of a forum).

People hated war so they set messed up defense teams, deliberately spent zero war energy, and in hundreds of ways broke matchmaking. So Devs had to fix that and “Cheating in war” rants ( more like “Shooting self in foot with a Nuclear missile” ) are liberally salted throughout the matchmaking topics. This lead to Tourneys but that is another discussion.

The Devs had a .
. .
. . .
. . . .
Vision about war matching ( that I very strongly disagree with ).

But a legitimate concern that starting from scratch would consume resources and not actually be any better. Each tiny fix was, is this a fast emergency tourniquet? Has the bleeding stopped? This took time, which increased the forum debris a huge amount.

They jerry rigged the matching until is works poorly but eliminates the problem on 104 losses per year.

So by a realistic, observable, quantitative, visible measure they actually succeeded. But neither the Devs nor the players are happy with it.

War matching simplified

Your attacking teams and defense are measured- this is complicated. There is really no realistic reason to explain more details - the Devs have moved on a long time ago.

Alliances are compared with a steap penalty for miss matched number of players that have participated in the last 2 wars.

A fudge factor is applied based on winning too many in a row or losing too many in a row.

Laws of War

Nope. Laws of War prevent this:

( Tips & Strategies for Alliance Wars (AW) - #22 by Gryphonknight )

Points not adding up

It can usually be traced to rounding war points.

Frankenstein’s Monster Matchmaking versus Elo’s Math

Remember our reroll conversation?

Smaller alliances bork the matchmaking because they are less flexible:

Synergy Imbalances

Imbalances in hero synergy are magnified with fewer total heroes ( smaller membership and smaller rosters each ) and minimized with more total heroes ( larger membership and larger rosters ). This is seen in newbie alliances versus elder alliances but evens out once an alliance has a decent number of 4* 3.60 heroes.

This effect is the basis for matching war center colors. Example all blue centers.

If all you have is a hammer you will have trouble tightening bolts. If you have a hammer and your teammate has a wrench your team can tighten bolts and hammer nails but not glue wood. By settings all their centers to the equivalent of hex screws, alliances hope you will run out of rosters with Allen wrenches.

This is exacerbated by the existence of three War Rules ( Attack Boost, Arrow Barrage, Field Aid ) each which prioritize a different roster for war defense and war attack.

Unlike Elo’s math, the current system assumes a 100% synergy.

Targets

Small alliances get matched with small alliances. This gives them fewer targets. If the only remaining target for them is an attack team with center Kasshrek, but all they have is blue heroes left, this will go poorly for them.

Just like with Synergy imbalances, small alliances are not as flexible in wars.

Unlike Elo’s math, the current system assumes a 100% appropriate targets.

Revive

The less hitters you have, the more enemy team revive depresses you score. You just do not have the flexibility of a large number of potential attacks to that fit your schedule.

Unlike Elo’s math, the current system assumes a 100% availability when a potential target is available.

Schedule

A real life emergency, a work shift, seeing friends, all impact your war success rate, but unlike Elo’s math, the current system assumes a 100% participation and a 100% war energy spent.

If you have 30 members, this has a smaller impact than if you have 5 and three go to the bar to watch Game of Thrones.

Knowledge base

Smaller alliances, especially non- English speakers and unaware of the forum, have a smaller Knowledge base. They do not know as much about synergy, color stacking, over writing buffs, etc.

Larger alliances can discover something by accident, on the forum, or from a merc, and spread that knowledge to the rest of the alliance.

Unlike Elo’s math, the current system assumes a 100% game knowledge.

Lack of rewards

The war chest is awesome, but war loot, not so much.

Sometimes players will skip war. This has less an impact in a large alliance than a small.

Unlike Elo’s math, the current system assumes a 100% participation.

Peer pressure

Large alliances can excert more peer pressure to participate in war.

Unlike Elo’s math, the current system assumes a 100% participation.

Replaceable

Small alliances either have a hard time recruiting, or are deliberately not merging with other small alliances. So if they remove a teammate for not participating in wars, they may permanently lose a member hitting titans. While the matchmaking auto opts out after 2 consecutive war skips, it does not deal with players who only hit only on Saturdays or who constantly use less than 6 war energy.

Unlike Elo’s math, the current system assumes a zero participation or a 100% participation.

The perfect is the enemy of the good

But due to sunk costs, development cycles needed elsewhere, and diminishing returns the current war matchmaking system is unlikely to change.

I had hoped they were rebuilding matchmaking from the ground up for Tourneys, and could then apply it to wars. But it looks like they took the sensible business approach of applying the war matchmaking to the Tourneys to conserve development cycles.

2 Likes

There are many alliances with 29 or 30 members, so the algorithm can pretty easily find a closely matched 29-30 member opponent.

If you have 17, then the algorithm is looking for an opponent with 16-18 members and a similar war score. That’s a smaller universe tomdraw from and will result in more mis-matches.

3 Likes

Just another reason alliances should aim for full, and I’ve seen it work as motivation for more recruiting, merges, or even disbandings. Which makes competition better as a whole

3 Likes

Thank you very much! I will look over these threads and join the discussion

Woah lot to unpack here lol…what is “Frankenstein’s Monster Matchmaking versus Elo’s Math”…

Opponent pool

Good point about the opponent pool.

Loophole

Unless they want to write a patch solely for this loophole, it is not going away.

This sounds like an extreme variation of one of the first ways the matchmaking was broken and then patched together:

(Analysis obsolete war strategy ARCHIVE for reference only - #7 by Gryphonknight)

Next band

Either the name of my next band or the next movie from Universal Pictures

The Devs thought Trophy dropping was because Elo’s math is broken instead a combination of maximizing rewards for least effort and prisoner’s dilemma ( see Notes ).

The patchwork of sewn together parts they ended up with is a miracle of life from death, but not very pretty and largely brute force.

Notes

1 Like

I will add, based on the answer to a few questions we have asked the devs, that it would appear that some of the scaling you would hope would happen to scores doesn’t seem to be happening. For example targeting the same number of opponents and related penalties to war score for having a numeric advantage. They may have fixed that, but at the time we asked it was a bit of an issue.

1 Like

Yep, and I would say no traction. My assumption is that it is such an edge case that they aren’t focused on it yet. or maybe they have their own solution they are working on, certainly mine is but one approach to the problem using some rather standard statistical tools.

1 Like

That seems like a principally reasonable explanation. It would be interesting to see a plot of alliance size vs number of alliances vs war score.

I predict that the War score is going to obscure a lot of the data’s fine details.

I would be more interested in a plot of number of 5* 2.60 / 5* 3.70 / 5* 4.80 / 4* 3.60 / 4* 4.70 / 3* 3.50 versus war score compared to the number of alliances versus war score ( seven overlaid plots ).

Many players come to the forums when they get sucked in by the heavy advertising for 5* heroes, especially HotM. Complaining about 3* / 4* ascension items or food and not realizing the problem with Hero XP and 5* heroes.

2 Likes

If we were trying to improve the algorithm, that would be super helpful. I’d just like to understand if the match pool is super thin, even at the aggregate level.

And yeah, it’s hard to keep even my alliance mates on the straight and narrow with heroes.

I’m currently leveling Ameonna, with Hel, Sartana and Aeron staring at me, hungry for feeders. It takes a lot of willpower…

1 Like

So shiny. . .

I’ll just throw my 2 cents in here. I’m in a smaller alliance as well (only 12 participating in war) and we routinely see exactly the same sorts of mismatches described in the original post and I can typically look at the match-up during the prep phase (before the opponents teams show up on the board) and have a pretty fair idea whether we are going to be competitive or if one alliance (usually us) are going to get blasted just by looking at the player levels. If the opposing side has several players whose levels are 10 or more levels higher than our highest, we are going to get beaten pretty badly unless those players opt out.

As @Kerridoc suggested the reason is almost certainly depth. These players best team is probably no better than any of ours, or only marginally so, but their deeper benches mean that their 3rd - 6th teams are way, way better. Unfortunately the algorithm intentionally devalues this difference by weighting heroes 6-30 less than heroes 1-5. As for this being a more significant issue in smaller alliances this is because in a war between small alliances one or two high-level players with deep benches are capable of one-shotting the top part of the opposing alliance and are going to have a much bigger impact on the outcome than they would in a 30-on-30 war where the top players on each side can only reliably take out a small percentage of the opposition. If one player gets 3 one-shot kills against an alliance of 12, this is 25% of the opposition, whereas the same player doing the same thing against an alliance of 30 is only taking out 10% on one set of flags.

This depth issue probably extends to troops as well.

2 Likes