Weighted K/D values, went AI-factor

Gen. Smit wrote:

Ok, could there be weighted K/D values this would reflect much better the current ability of a player than the current global K/D value.

A sum up of last games K/D's could also be interesting.

Also casualty rates would be interesting. It allows better assessment of players in current game and whether to take them in a coalition, or to join up with them.

Also went (weighted) went AI-factor could be of interest

These figures do not invade privacy (as time spent online would), but do enhance the decision making concerning who to ally with.

last 1 month

last 6 month

last 1 year

Overall

however I have doubt it still gives all info about a player

For example

Comintern doctrine is bad for K/D ratio

Dawn of Patriot map is bad for K/D ratio , you can find yourself between 5 pan asian countries at the beginning of game

Historical Map is bad for K/D ratio , mostly you start to game with low level units

even some locations are bad for K/D ratio

maps , locations , doctrines all of them important factors

so I can't say 30 k/d ratio player is harder player than 2 k/d ratio player

you can cheat to show high your K/D

service record only shows mostly played map for example it shows 65 games 100p map it is ok

so you can join 64 times to historical map and choose Germany to increase your K/D ratio

who will knows you played 64 times Germany at historical map ???

They could have some kind of power ranking, that takes into account K/D, solo wins, coalition wins, map type, game length, power ranking of other opponents in the game, etc and more recent games would have more weight. Older games would eventually not contribute to power ranking.

DxC wrote:

They could have some kind of power ranking, that takes into account K/D, solo wins, coalition wins, map type, game length, power ranking of other opponents in the game, etc and more recent games would have more weight. Older games would eventually not contribute to power ranking.

example of power ranking

when you completed to game as winner for every 1 k/d ratio at 22p nations map +1 point

when you completed to game as winner for every 1 k/d ratio at world war map +2 points

when you won solo game at 22p nations map +10 points

when you won solo game at world war map +40 points

when you won as coalition at 22p nations map +5 points

when you won as coalition at world war map + 20 points

so game length , map types , type of win and k/d ratio at map will effect to points

in this situation if you won a world war map as solo with 23 k/d ratio you will get 46+40 = 86 points

note : to get points from k/d ratio at this ranking system , you should win to map as solo or coalition , other players won't get points from k/d ratio , only winners will get points

I'm not sure how it would be fair so I wrote all numbers randomly and I didn't add other maps to show it basically

what do you think about it ?

can it be improved ?

I'd say that you shouldn't necessarily have to win. For example, you could be solo vs a coalition and have the most kills of any player and still not win. So overall placement combined with total damage dealt should be included if you are trying to evaluate the skill of a player.

DxC wrote:

Older games would eventually not contribute to power ranking.

@All

I didnt propose any other factors as it is far more complicated to implement and to create a truly good "power ranking" yet I find the arguments brought up convincing.

Sure there will be cheats/manipulations, and one must do more to investigate potential coalition candidates (if you even have a choice at points).

I do not fully agree that certain doctrines are bad for KD, that also depends on the playstyle.

Therefore casualtie ratio may be interesting too to implement, not dying is one thing, yet casualty value is more pure. But again also this is a global factor.

But I encourage strongly a simple yet powerful proxy for player capacity

Hm, some sort of k/d ELO calculation would be interesting.

What if each unit kill you got gave you points based on the opponent's raw k/d ratio, and each unit death lost you points based on the opponent's raw d/k ratio, and this would become some sort of power ratio.

I kill 18 units of a 1.5 k/d player, 18 * 1.5 = 27

I lose 9 units to a 1.5 k/d player, 9 * 1/1.5 = 6

Total k/d would be 2.0 but my power ratio is 27/6 or 4.5 because it was against a stronger player.

I kill 18 units of a 0.5 k/d player, 18 * 0.5 = 9

I lose 9 units to a 1.0 k/d player, 9 * 1/1 = 9

Here my k/d is also 2.0 but my power ratio is 9/9 or 1.0 because I got most of my kills against a weaker player.

I don't know if the math actually works out in combined scenarios but it's a start. Of course, I think it's unlikely that Bytro implements something like this but it's fun to brainstorm.

Trying again with combined scenarios, I don't think it works. Oh well, posting my math anyways:

1)

An account with a 2.0 k/d ratio that has 1000 kills and 500 deaths.

One war against a 5.0 k/d player went poorly, 200 kills and 300 deaths.

One war against a 1.0 k/d player went pretty well, 200 kills and 100 deaths.

Several wars against 0.5 k/d players went really well, 600 kills and 100 deaths.

Numerator: 200*5 + 200*1 + 600*0.5 = 1500

Denominator: 300/5 + 100/1 + 100/0.5 = 210

Ratio 1500/210 = 7.1

2)

An account with 2.0 k/d ratio that has 1000 kills and 500 deaths

Against 1.0 k/d players: 100 kills, 200 deaths

Against 0.5 k/d players: 900 kills, 300 deaths

Numerator: 100*1 + 900*0.5 = 550

Denominator: 200/1 + 300/0.5 = 350

Ratio = 1.6

3)

An account with 5.0 k/d ratio that has 1500 kills and 300 deaths

Against 2.0 k/d players: 300 kills, 200 deaths

Against 1.0 k/d players: 400 kills, 100 deaths

Against 0.5 k/d players: 800 kills, 0 deaths

Numerator: 300*2 + 400*1 + 800*0.5 = 1400

Denominator: 200/2 + 100/1 + 0/0.5 = 200

Ratio = 7

I think we'd all say Player 3 is the best of the bunch, but somehow Player 1 gets a higher score with this methodology. Back to the drawing board.

jubjub bird wrote:

Trying again with combined scenarios, I don't think it works. Oh well, posting my math anyways:

1)

An account with a 2.0 k/d ratio that has 1000 kills and 500 deaths.

One war against a 5.0 k/d player went poorly, 200 kills and 300 deaths.

One war against a 1.0 k/d player went pretty well, 200 kills and 100 deaths.

Several wars against 0.5 k/d players went really well, 600 kills and 100 deaths.

Numerator: 200*5 + 200*1 + 600*0.5 = 1500

Denominator: 300/5 + 100/1 + 100/0.5 = 210

Ratio 1500/210 = 7.1

2)

An account with 2.0 k/d ratio that has 1000 kills and 500 deaths

Against 1.0 k/d players: 100 kills, 200 deaths

Against 0.5 k/d players: 900 kills, 300 deaths

Numerator: 100*1 + 900*0.5 = 550

Denominator: 200/1 + 300/0.5 = 350

Ratio = 1.6

3)

An account with 5.0 k/d ratio that has 1500 kills and 300 deaths

Against 2.0 k/d players: 300 kills, 200 deaths

Against 1.0 k/d players: 400 kills, 100 deaths

Against 0.5 k/d players: 800 kills, 0 deaths

Numerator: 300*2 + 400*1 + 800*0.5 = 1400

Denominator: 200/2 + 100/1 + 0/0.5 = 200

Ratio = 7

I think we'd all say Player 3 is the best of the bunch, but somehow Player 1 gets a higher score with this methodology. Back to the drawing board.

I have to play a bit with this one/ think about it/ be inspired by

@jubjub bird, i think in the latest post you introduced a relativity, the first player is maybe on paper according his KD not the best, but he had to fight a 5.0 KD player, and he did not perform bad with 200 kills, his level is not used for relativity, but his victims level is.

His performance is thus not so bad at all, the 5 over 2 weighs in really heavily.

(200*5)/(300/5)=1000/60= 16.67 is a strong result despite being on the losing side/ the thing is that it is a RELATIVELY strong performance

the average player (KD=1) is expected to have

(100*5)/(500/5)=500/100=5

Maybe it should go in the direction of casualties as well, this is independent of Kills and Deaths itself, it is a raw metric.

So casualties/kill are a reflection of skill in the terms of that you are economical in warfare and know how to deploy your troops effectively to eliminate opponents. High is good low is bad

1000/1 is a KD of about 1, 100/1 is a K/D of about 10, not considering army usage and regeneration

Casualties/death are a reflection of robustness of your armies how effective you keep them alive, how you deploy them and how you regenerate them

1000/1 equals a KD of about 1, 100/1 of about 10 (low is good)

they are almost logarithmic because 1 player 10 times as good as the "average" player should have a value of 10, a player 100 times as good as the average player a value of 100 and is still 10 times better than the "10 x average"-player, a logarhitmic scale is good for discerning skill of players under the average

this is already far from the weighted idea but for purposes of skill it may be effective. Yet I like a Unifying power score as you proposed.

I don't think the ranking should be derived from the k/d directly. The ranking should be some sort of elo score that takes into account many factors, not just k/d. For example, at the end of a round your performance against all players you engaged with would be measured as some ratio based on all relevant factors, so unlike chess it wouldn't be win, lose, draw, but a continuous function. The value of this function combined with the current elo of yourself and your opponent would be used to update your elo.

DxC,that would be the best for sure!. But since it starts from scrap this is the most difficult option, while weighted KD offers a simple start. I agree, one can have lower KD's and still be very effective in achieving the goal, winning.