How would you rank alliances based on publically available (ie no spyop required) information?

Any of the hit stats (including nw's avg nw's tnw's at time of hits for example) can be used, all scores stats can be used. And somehow come up with a score that is usable, and provides a good ranking.

Things that need to be taken into consideration: A late-FS on a fat no-def alliance would lead to large civ counts and good HPK, but does not in anyway show good attacking ability; an early FS by one warring alliance on another should be rewarded despite the fact that the civ counts & networths would be low etc; and a high HPK does not necessarily mean that you're not good at warring, it can simply imply that your opponent is good at stonewalling; or it could indicate that you don't know what you're doing.


So, put your thinking caps on and give me some ideas ;) Even a nice starting point would be useful. Feel free to use the last few sets of data (in boxcar or elsewheres) to test your scoring algorithm.


Basically I'm looking for an alternative to ANW / TNW rankings for warring members that is actually accurate in describing warring ability & effort.

The only viable way I can see this happening is examine NW loss/Type of attack. A no def country would lose little to no NW. A country stacked full of defence would lose a measurable amount.

Say GS is the attack. Troops count towards $X/NW - So breaking a target down would bring total NW down.

BR would destroy buildings and Jets & Turrets = $X/NW + $Y/NW + $Z/NW. The only flaw with BR is acres. How can you tell if a country is fully developed or sitting on 10K free acres?

Also missile attacks would have to be calculated in as well - Acres and buildings - $X/NW & $Y/NW.

Obviously, it would have to look at success vs failure rate to take into account lemming as well.

Hmm we could probably use military losses in an attack as well actually; while not public, it would be hard to reverse engineer that information as it would aggregate fairly quickly

formula:

score = ((kills of clan A on Clan B/deaths by enemy B)+(sum(max living countries of B)/(sum(max living countries of all clan A))+(networth lost of B dead countries in M / networth lost of A dead countries in M)-(((sum(max living countries of A))/end # of A living)-1)+(ending NW of A/ending NW of B))*(case when A had first kill then 1 else 1.25 end)

A vs B
A: 50 member, 150 mill NW, 150 kills, kills 200 mill NW, ending NW 100
B: 100 member, 250 mill NW, 90 kills, kills 120 mill NW, ending NW 150
A got the FS
A had 50 most living at once
B had 100 most living at once
A had 40 living at end
B had 55 living at end

A score = ((150/90)+(100/50)+(200/120)-((50/40)-1)+(100/150))*1 = 1.6 + 2 + 1.6 - .25 + .6 = 5.55 * 1 = 5.55
B score = ((90/150)+(50/100)+(120/200)-((100/55)-1)+(150/100))*1.25 = .6 + .5 + .6 - .8 + 1.5 = 2.4 * 1.25 = 3

so even though clan B ending higher NW, with more countries, clan A killed enough and had a better restart rate, and even though they took a penalty for getting the first kill, they still ended with a higher score

i came up with this based on several measures of war performance. i think # of hits shouldn't matter as much as what was actually killed and the NW's that were killed off. the multiplier at the end is an attempt to scale it on a scale of 10 and to adjust based on who got the FS. therefore a perfectly even war in #'s, kills, NW destroyed, and restart level and ending NW, would be 1+1+1-0+1=4. the attacker would have score of 4 and defender a score of 5. a perfectly good score of an outnumbered clan getting hit by a larger clan and ending with twice the NW, and killing twice the countries, both having perfect restart rates would be 2+2+2-0+2 = 8*1.25 = 10, whereas if the smaller FS'd the bigger, and won in the best way possible as described in the last case, they would have 2+2+2-0+2 = 8*1 = 8.

no matter how you score it, a downside will be for the winning side to never want to declare a ceasefire, since the more that the other guy is pounded into the dirt, the higher their ending score will be

kills on your own tag would be ignored

for a war score to take effect, each side would have to lose at least 10% of its starting membership. therefore if a 100 member tag kills off a 25 member tag in a couple days, with the 100 member tag only losing 4 countries, a war score would not even be generated for either side. since it wasn't really a 'war' more of a genocide. or something along those lines at least, such that killing a suicider or two in another tag before they're detagged, doesnt generate a war score for either side

(networth lost of B dead countries in M / networth lost of A dead countries in M)

that term would tend to make it so if a small clan missile killed the T10 at the end of the set, they would dominate

(case when A had first kill then 1 else 1.25 end)


would preempt a co-FS

see note about needing to kill 10% of a clan before war score generated



also, the co-FS, the implication i was thinking of, was a friendly war, forgot to note that that wasn't covered. i was thinking if Clan A got first kill and clan B got first kill within 30 minutes (or hour) of each other, no multiplier would take effect, but even then, in a co-FS the incentive would be to cheat and just wait 30 minutes (or hour), though that would piss off the other side and ruin the intentions of the friendly war

as for a co-FS, clan A and clan B both hitting clan C. if clan A gets first kill, and clan B uses any special attack on C within 24 hours and gets a kill within 72 hours on C then clan A and clan B will both be penalized for taking the FS. then again, clan B could only break, and A finish for 72 hours, therefore allowing the first kill on B by C to count as C FS on B. so some sort of measure would be needed to prevent that

it would get complicated with multi clan on multi clan wars of course, which i'm not sure how to explain yet, but this was attempted to be a starting point of a formula of some sort ;)

mmm say it's a 10 man clan with the T10 of 150M+ each; if you kill them at the end, you'd end up with a higher score than the all-set-long 300 vs 300 wars =/


Yea, don't take my criticisms personally or anything :) I'm just trying to point them out for all of us to think about; so that we can iteratively make it better.

In the next couple of days my formula should be ready. When it is released I think it will be highly balanced.

:) i await it expectantly then :)

If I didn't have some big work deadlines today I think you could expect it in several hours =P

quit f-ing around here and do it !!

:P

problems with llaar's initial draft formula

1) the score should be based on a high number that can be saved across multiple setws as the record i would suggest a max of at least 100 not 10

2) initial draft does not address timing within the set of the attack

3) initial draft gives too much weight to restarts

4) draft weighs kills only

the only reason i care is that i hope the smart guys can come up with a formula which can then be tweaked to make sense for individual servers.
i think it is vital to come up with a formula for the individual games

both sides have to lose 10% i said

so just suiciding last hour of set and killing a few top countires of a small clan, still would generate no war score



1. just multiply by 100 then... that'll make it in a range of 100 lol. there would be decimals so all you have to do is multiple to get any sort of large score you want

2. correct, length of days would have to be measured somehow for a start and end period to the war. that would actually help account for CF's, as the length would be from first to last kill. how that should affect the number though is a big question for sure... a 5 day war in which one side is obliterated quickly, how does that rate to a 30 day war in which one side is obliterated quickly, yet continues to try to fight another 25 days?

3. probably true, restart rate shouldn't have same weight as kill ratio or NW killed, maybe divide the restart rate by 2 or something

4. it rates kills/NW killed/restart rate/who got the FS/ and then somehow can also rate the length. not sure what you mean by only kills, since it clearly does more than that

I'd like to see somehow accounting for spy attacks. You could do this in the background to use the spy attacks to also rate the effort.

scoring is hard, but id like to see more metrics available in war reports

so you could see the status of the tags as they went to war, things like

turns stock oil land net income missiles both total and average
and average war techs

perhaps only make some of them available after a period of time or at the end of the set if the infos too sensitive

stock should be in food and cash, perhaps a total tech as well, individual techs might make the most sense but its more important to know if one tag has warfare and the other doesnt than whether one tag had a pile of military tech stocked instead of bus/res/agri/indy in techers

would be nice to know how the resources are spent too, and how much mil is lost on offence or defence, perhaps one side picks high breaks and the other low

if you could see during a given hour/day of play what proportion of income was spent on increasing military, military losses, expenses, building it would be very interisting after the war, or on a delay

cash would probably be the easy one but if you averaged sale prices for food and tech and oil and military you could estimate incomes

the alternative is to take actual sold income then apply it to either the turns it was produced in or the turns ran with it

another interisting one would be FA, both total sent and received by a tag and internal FA, without knowledge of outside interference wars can be harder to guage

would need to track aid that restarts get before they tag up as well

then your just left with market based aid which is a lot harder than it used to be and not a very easy thing to do efficiently anyway

I like the use of HPK and civs. You can still read alot into them.

You could argue similar points for our networth indicators. You can't tell who is stockpiling better, because you can't tell when they started. Each strat would be fitting into a different niche during different stages of the NW game. War would be the same, and we'd all be making our best judgements and interpretations based on what we can see or know about about government composition, prices on the market and stage of the game.

The simpler the better for an ingame monitor.

been busy at work and forgot about this over the weekend.. soon....

well i take it mines still the best so far since its the only one :P

been wondering what you're gonna come up with detmer

Basically the more things you do the higher your score will be. I just want to balance it well rather than assigning arbitrary values which will inevitably favor some things more than others.

Hmm. Even if there isn't some amazing coefficient made to rate war performance, what's stopping EE having weekly medals of some sort for different categories related to gameplay, like most money made off market sales, best land grabber or best grab, most successful attacks, most successful defends, even the highest networth. Could even have some that were still given out weekly but were based on whole set averages like defense or attack win percentages on either land based or warbased attacks .

Calculating winners on a weekly basis would take very little resources but would give every player something to strive for other than winning top networth. These could be done on both a clan basis and individual basis.

Stuff like this will keep new players playing far more than an amazing coefficient for war performance would, as much as it pains me to say that.

Basically saying, it would be great for the developers to find a way to rank war clans as they do netters, but why stop there?

true heh... definitely have to think about some stuff like that...

this is my attempt at brainstorming for a war performance formula.

This suggestion has been made by me and a few others, many times since EE opened, that clans should be able to declare war(or peace) with other clans in the clan type servers. This would give EE a very clear dataset to work with when determining war performance, and could possibly open the door to war related changes in the game like how dec war on a country changes taxes/gains but on a different scale, but for now, i'd be quite happy with an option to declare war or peace.

Since Networth has always been the deciding factor in who wins the game, Networth should be at the very heart of any war performance formula. I think that this will be the only way any EE ranked war performance would be accepted by the greater community.

We could also be looking a this the wrong way... Maybe we should stray away from the whole networth ranking style and actually just make some formula that shows the clear winner and loser of specific wars and maybe EE could keep track of war related scores from reset to reset and show a "war win/loss percentage" like old earth used to show for individual countries, tho this would probably stray too far from earth's "reset" arcitechture where the beginning of a new set is always a clean slate for everyone and would possibly be hard to code for(tags don't always stay the same from set to set and even if they did, there are always tag stealers, but could be rectified by registering the alliance NAME(Imaginary Numbers) once and then making a new tag from that each set). Something like this would give a new player, or someone looking for a new clan, a clear, accurate picture of the clan's past/present war performance.

I also think that whatever equation we come up with should give each attack or defend X amount of points, and a clan's war performance is all their clan's members points added together, then you could even rate single players(both on solo servers and clan servers). A forseeable problem with this method would be a clan that wars the full 2 months of the set is very likely to get far more points than a clan who only wars the last 2 weeks(or whatever) but that might not be a problem if we really are trying to rate general war performance. max(or base) points for an attack/defend could be 100 and use different factors to decrease that number by

maybe an easy way to keep networth at the heart of the equation for EE to keep a snapshot of each clan's NW and members at the beginning of a war and use that info as factors in each points related equation.

In relation to clan members, could give the side with the fewest members an advantage like Defender'sSide'sMembers/Attackers'sSide'sMembers = factor. so, if side A had 50 members and B had 49. if a country from side A hit a country from side B, one of the limiting factors on the attack would be (49/50) = .98 and if a country from side B hit a country from side A, it'd be (50/49) = ~1.02. might be too much of an advantage, but i'm sure some mathematical genius could work out a better equation for that with max() or min() involved to make it so you couldn't have more than 1.

In relation to NW, it could give points based both on the original NW difference of the two sides and how much the networth of your side had strayed from that original number(up or down) in relation to the other side of the war. so, say Side A's starting NW was 10,000,000 and it's 9,000,000 at the time of the attack and Side B's starting NW was 12,000,000 and is curently 8,500,000. Side A's NW relative to its own starting NW would be 9,000,000/10,000,000 = .9 and Side B's NW relative to it's own starting NW would be 8,500,000/12,000,000 = ~ .71 and then compare those two numbers together. something like Max(1,(AttRelativeNW/DefRelativeNW)). in this case, it'd give sideA a 26% bonus because it had destroyed more of sideB's NW. could even make it so the advantage couldn't go more than 25% to Min(1.25,Max(1,(AttRelativeNW/DefRelativeNW))) and in this case SideA would get the max with a 1.25% bonus in points. if total NW didn't seem feasable or gave bad results, ANW could be used instead.

another possible factor which relates a lot to NW and normal gameplay is land, and as most wars are won or lost in relation to land, we could also use relative acres in the same fashion that NW could be used as a determiner, tho if NW has dropped, it can be assumed that land has dropped drastically aswell.

an equation relating to warfare should also include civs killed, but in a manner tha a 500civ per hit in week 2 should be comparable to a 22000civ per hit in the last week(or whatever the numbers really are). I was thinking that an easy way to accomplish this is to use relativity again. basically, the attack's result compared to the max result achieved SO FAR. so.. if the GS killed 500 civs and the max civs killed so far in the reset from a GS was 600 it would be 500/600=.83 could even include a minimum result if you didn't want that number to get too low on finishing hits, but most people would probably agree that civs per hit points should really be aggregated by the breakers anyway. this method would also mean that chem's killing thousands and thousands of civs wouldn't be too terribly different than results from a GS or BR because it would only relate to attacks of the same nature. It might actually cause warring clans to use chems for start or a KR rather than for finishing(chem rushing) effect would be a lot of KR's lasting longer.

will just say that I'm only brainstorming here. there's likely to be lots of holes in what I put, but just trying to put something out there other than what llaar has to give us something to compare.

and, if you've read this post and are asking yourself "has he actually even mentioned a full equation?" the answer is no, and I probably won't. Whatever equation that EE Admins/Developers come up with, I don't really want to know the full extent of it, as long as it's a balanced equation that works in tie with the rest of the game mechanics.

just thought that it might be good to give each one of the determining factors a base of 100 points and then either total or average the points out by total number of factors.

and i've just remembered that I failed to mention anything along the lines of defender points. I think that, if the attack is successful, the attacker should almost always get more points from an attack than the defender, but the defender should at least get some and there are times where the defender should possibly get more.

A def held should give the attacker -X points and the defender +X points, where X is a static number of points. I thought about making the defender's points be dynamic based on how many hits they've taken in the last X hours but that would give a lot of points to someone stonewalling while giving nearly none to a country that had to be lemming'd down(which should get points IMO).

defender points should also use the same kind of factors mentioned in the above post, but likely be cut in half.

but qzjul also mentioned something about possibly using oversend as a factor and this could possibly be something that gave defenders more points than the attacker if the attacker was sending far too much military when not needed. some would argue that sending far too much military should give you better results b/c 500 guys kill 2 guys a lot easier than 3 killing 2, but i think in reality, 3 guys killing 2 would be deemed more a "hero" and given praise than the 500 who slaughtered the 2. *shrug* just my opinion.

was also thinking that this would put too much emphasis on total amount of attacks, so maybe give each attack points, but to get the clan's actual performance score, take an average of all the points gained. *shrug* might be problems with that as well.

Making formalized declarations of war is definitely something we want to do, though we're still debating the consequences etc....

I used to use a battle-rating for each member in a couple wars, which I had a lot of fun with. Although it required fixed start and end points in the war, and relied on old GT formatting. Don't ask me about the numbers in the equation. I used excel and made em up where it felt right. This was just some inhouse fun.

Battle rating=[NLG]x[NWmod]x[Battlemod]x[Attmod]x[Defmod]x[sizechange]x[survival]x[govtmod]

[NLG]= =ABS(STANDARDIZE(prewarNW,clanAVERAGE,stdDEV))+1

[NWmod]= std deviations away from the average of membership.
So if you're much smaller or larger, this mod could be 2-3x.

[Battlemod]=(LOG((civs/hit),300))^4
Those pesky spartans need to die with one hit.

[Attmod]==EXP(std dev of members hit to clan hits to kill)

[Defmod]=EXP(std dev of members hit to enemy hits to kill)

[sizechange]==ABS(STANDARDIZE(NW change between war,Avg NW change of members,std dev))+1

[survivalindex]= Cant remember but something to do with number of members on both sides and NW sizes stdised somehow...

[govtmod]=Govt mod (fixed arbitrary)
0.95
0.95
1.05
1.1
0.95
0.9
0.95
1.3


I have the spreadsheet if ya want heheh!

The simple solution would be to count everything as the networth it is worth and have a separate war networth. Technically you could do it without any extra information which could somehow be reverse engineered. Networth before the attack, then after. The database then just adds it up and you could have it tracking for individual players even, so there could be individual and alliance ranking for it. Too simplistic?

It would be funny to see which alliance has the highest War ANW, aka literally the most destructive alliance. Though depending on land gains, some of the big farmers might still be pretty high up there.

I dont see why we need a war scoring formula and as usual some of the suggestions are rather silly

just add an earthgraphs section to earth and yer set :P

ivanfluff

take total net worth

add a million for each kill
add a million for each lg over 1ooo acres
add a million for each cash/food/tech grab over 25 million dollarss

My other wasn't a suggestion.
General top-10 stuff is what I'm using as a guide anyway.

fwiw, still working on this =P

HPK and civs killed are probably the most relevant publicly available info - though making that too formalized would no doubt lead to a lot more kill stealing.

How bout we all just score wars in terms of funny AT posts? :P

lol detmer.. still wondering what you're going to come up with

Every day I come here to check this. For weeks now I've been waiting on testing missiles.

I had been doing so much Earth I was on the verge of losing my job so I have cut back and thus haven't done the missiles yet.

I am hoping soon. Monday maybe I will test them then I will just need a short period of time to work something out =P I know the basic idea behind it and it will just be balancing the numbers.

holy crap. Detmer is like WTK. we have a post from him saying "i will update this tomorrow" since like 2 years :)

heh, I haven't forgotten =P It is just a matter of prioritizing. I will get to this when I have the time to record some stats =P

its been nearly half a year now buddy... :P

I hope you can come up with something here qz. My two-bob:

1. You cannot come up with a viable 'ranking' unless you formalise alliance vs alliance warfare. This needs to be broken down into two options for Tag controllers (an alliance's leaders). To initiate a war with another alliance you may (A) Challenge Alliance or (B) Declare War. If a challenge is issued, the challenged alliance has 48 hours to accept or decline - this would be used for mutually agreed wars - no first strike penalty. In the Declare War option, the aggressor is instantly at war with the alliance and will suffer in the points system via a "first strike penalty".

2. Work out a "War Score" that is similar to how the Windows Experience Index rating is calculated. Output and efficiency are the only real factor if you want to rank a war alliance on a Power Ranking. Average Hits/Day, Kills per Total Countries/Day, Average Defends/Death are factors which would weigh into a Skill Ranking (as well as maybe tech levels, SPAL etc..). Power and Skill Reankings woiuld equate similarly to Total Networh and Average Networth.

3. Make the ranking a rolling ranking over say 5 'wars' (tracked by the ingame 'Declare War' option), with the 6th last war dropping out of the calculations and the previous 5 war numbers declining in value by a percentage.

Ah fluff... ok... for reals... not too long...

fwiw I am in the same place I was when I first posted... its not like it is taking me longer to actually do it than I expected =P

lol at detmer :)

llaars basic idea sounds good, with some tweeking

mm i like dagga's ideas

My idea (would require a formal war dec option in game):

(Enemy Clan NW / Your Clan NW) / (1 - % of enemy networth destoyed while at war).

So if enemy clan starts at 50m NW and you are at 100m NW and you knock them down to 10m NW your score would be: .5 / (1 - .8) = 2.5

If the enemy clan had started at 100m and you are at 100m, and you knock them down to 20m NW: 1 / (1 - .8) = 5

Reason this is good:

1) NW Destroyed is probably the best estimator of war performance.
2) Doesn't benefit you much to destroy somebody way smaller than you.
3) Benefits you to kill bigger alliances

Perhaps could add HPM in there too. Maybe multiply it all by HPM.

(% of NW of target clan destroyed) x (% of NW of server destroyed) x (% of set elapsed) X 1000000 = WARSCORE

nice stats page you got so far. Collecting everything?

Is:

"Total Defends
This leaderboard ranks users by the total number of defends they have taken across all rounds. All countries, except those played in FFA, count towards this leaderboard."

different to this:

"Successful Defense Percentage: 4%" (ingame)

I think it is so should be redefined?

This stats stuff, makes each set all the more precious...

There is alot more at play here than what has been suggested so far.


In order to get an accurate "War Ranking" the following MUST be involved in the Forumula.


Avg HPK
Avg HPD
Avg Attacks per day per country
Avg NW destroyed per day per country
Avg Kills per day per country
Avg NW per country Start of War
Avg NW per country End of War
Avg Land per country Start of War
Avg Land per country End of War
Total restarts made



The top war clans are going to kill more countrys than there enemy, kill with less turns than there enemy, destory more NW than there enemy, restart there dead countrys 5-10x's if they have too, make there enemy use more turns to kill them.



The simpliest way to make the formula is to add modifiers to certain peices of the forumla, depending on what week it is in the set. Week 3 for example would give a higher reward for "HPK" than Week 4. Week 4 may have a higher reward for "AVG NW Destroyed per country per day".

I say that because in Week 3, country will have higher population, but maybe not the higher NW quite yet. So getting lower HPK's is more effecient than killing off NW. But in Week 4, countrys should have some decent NW. So killing off more NW may be more effecient than the HPK.


Anyways, thats just my 2 cents on the subject.

<qzjul> well i've been asking on the B&S for a War Scoring formula for like 8 months now
<toma> maybe 5k nw to scores added for each hit and 1m nw for each kill
<toma> maybe even give points for well played market manipulation etc etc
<toma> so would you guys be willing to test it for 1 or 2 sets if i start thinking of some game scoring changes?
<qzjul> well we'd probably test it on an alpha server first
<qzjul> if you put it on the war scoring thread, we can debate as well
<qzjul> keep in mind i'm also thinking for clan-style warring
<toma> nw*sqrt(1+kills)*ln(e+damagedonetoNW) = score
<qzjul> mm
<qzjul> but in a clan style server you can just cherry-pick kills
<qzjul> and you can get low kills really easy
<toma> yeah that is for solo only really
<qzjul> i just get a buddy to keep restarting, getting OOP and killing in 3 hits
<qzjul> or 10
<qzjul> then i win
<qzjul> seems easy to game
<toma> nw*sqrt(1+( 1 - e^(-hitsToKill * someConstant) ) * kills)*ln(e+damagedonetoNW)
<qzjul> so if you're like the best mid-breaker ever in a war, would that show it?
<qzjul> or mid-hitter even
<toma> ln(e+damagedonetoNW) this part would grow
<qzjul> mm true
<qzjul> well
<qzjul> maybe not tho
<qzjul> the breakers would drop the nw most
<qzjul> the mid-hitters just hit fast
<toma> well include 3 factors 1 that favours cherry picking, 1 that favours breakers and 1 that favours mid breakers
<toma> # kills , hits/second*damage, damage
<qzjul> :) good line of thinking anyway :) post on B&S on that thread though :) it's topped heh
<qzjul> or stickied rather
<toma> then balance them out with constants etc
<toma> can i just copy pate parts of the chat?
<qzjul> sure
<toma> too lazy to write something new :P
<qzjul> lol

seriously... should be happening soon now for reals