To Define the Abstract

Disclaimer: Any comments i make in this thread are not attacks on people personally, merely trying to objectively assess peoples abilities.


As many of you know by now the wonderful site SGR:stats(dot)com has recently been updated with a rankings section. This has however created new problems: How do you define the best player? (and for all you smart ***es black is not a valid answer)

No, we need something else, a definate equation by which we can rank the players of the SG:R. So here is you challenge community: To define the undefinable, skill.




We want something that is going to last here so it hopefully should have a time element rather than having to reasses the equation every few months.

In the sgrstats thread there was an initial disagreement over whether XP or k/d ratio should be used to rank the players initially. Both of these have advantages and disadvantages.

Ranking by XP seems to reward the time someone has availible to play the game over how good they actually are at it. If we're trying to sort out the best players then surely we can't punish them for having a life? This might also cause issue if the game persists for a long time, as i'm sure we all hope it does. Players who have been here from the start will pretty much dominate the rankings for the rest of the game if this is used for too long. This could be compensated by introducing a time factor into the equation for example XP earned per week or simply total XP over time played.
one advantage of XP though is that it is based off points and your team winning. This means that kills contribute to your XP but there is absoloutely no penalty for death. This doesn't seem right as effectively thats like saying a kamikazee pilot that takes out 4 tanks and a boat is better than one who take out 3 tanks and survives. All the kamekazee pilot did was crash his plane into the ground wheras the other needed to take out enemies and dodge their fire.

Ranking by K/D means that it is possible to have someone come in, hit a few good games right off the bat and already be at the top of the rankings. Yes that player has shown a lot of skill in those games but most agree that until they are consistantly hitting that form they should not be that high up the rankings. It also favours some classes over others, ashrak being a good example, which could skew the results towards certain players.

From these points it seems that a compromise between XP and k/d is needed.


First of all i tested simply multiplying them together. To test how this affects the rankings I took the current top 10 based on the xp rankings and the top 10 based on k/d. This left us with 16 candidates due to a couple of duplicates. I ended up dividing the result by 10000 as the results were in the hundreds of thousands.

Table for (XP*k)/d

#   Name           XP        k/d       k/d rank   Skill(/10000)  Sk. rank Change
1   Black          248502    5.9110    1          146.89          1       -/-
2   Murphyslaw     173110    1.3938    14         24.13           6       -4/+8
3   LittleGirl     166364    2.3018    11         38.29           2       +1/+9
4   Dr Jackson     154510    1.0051    16         15.53           10      -6/+6
5   Do0msayer      114458    1.8779    12         21.49           7       -2/+5
6   quack quack    108056    3.2930    3          35.58           3       +3/-
7   moko88         104166    2.6798    7          27.91           5       +2/+2
8   Bacon squad    102050    1.0445    15         10.66           13      -5/+2
9   Kheno          96944     1.8644    13         18.07           9       -4/-
10  Rivalen        95078     2.9838    5          28.37           4       +6/+1
11  Kal1981        78088     2.6409    10         20.62           8       +3/+2
12  Caliaa         37810     2.8903    6          10.93           12      -/-6
13  Whitetiger     37638     3.0668    4          11.54           11      +2/-7
14  Culgan         24264     3.3156    2          8.05            14      -/-12
15  ManuelFirst    23436     2.6659    9          6.25            15      -/-6
16  Recon911PDW    8286      2.6694    8          2.21            16      -/-8

This still seems too biased towards XP as only one of the top 10 for XP dropped out of the top 10 for the skill equation. It does however seem to be more reliable, causing high XP players with low k/d and high k/d with low XP to drop down the rankings.

Next going to try out (XP*k)/(time*d)


Couple of questions:
What is the difference in XP between having an awesome game on the loosing team and a crap game on the winning team?

XP is flaky. I think u get half te xp u normally would for losing the match.

Meaning the top player on the losing side will get less xp than the avg player on the winning side. ie 136 vs at least 100 xp for someone who made a few points on the winning side. Ntm if u switch classes at all during the round ur xp plummets, which i did alot early on hence the disparity between my play time and xp.

i have done a fair amount of research looking into how exp is calculated, for myself i average about 180-200 exp per game for a win and about 120-140 for a loss, all depending on how well i do. from what i can tell the exp for a win is 160, with roughly 4 points added per kill (you get 0 exp if you dont get at least 1 kill), not all of the games i recorded data for match this however, but it seems to be the average. for a loss it seems to be roughly 120 exp plus appropriately 4 per kill. from what i can tell, neither total points nor deaths seem to effect exp, only kills. to answer your question a game resulting in a loss where the player preforms very will (many kills, about 20) will earn said player roughly the same exp as a win where the player only has 1-2 kills. however there is a good difference between a win and a loss when the players does about the same in both games. this may not be exact but most of my data seems to match this fairly closely, likely meaning there are some other factors i didnt account for, though im unsure of what they could be, more research will me needed, hope that helps.

edit: forgot one thing, i think the bonus you get (+160 for a win, +120 for a loss) is dropped if you change class, i believe that this is because the game doesnt know which class to give it to and wont give it to both, or all 3 if you use all classes

also i feel that the best equation will be: (exp/time)*(kills/deaths) or (exp*kills)/(time*deaths) depending on how you want to look at it, both yeald the same result.

to start, you will never be able to satisfy all players, nor will you be able to 100% accurately display THE best player. So you will have to settle for less, and something that gives a fair representation..

but i think definitely the EXP/Time is good, because then it awards you for your EXP/minute, not just EXP in general, as that would kill new players. So that is one option.

Then you need to factor in K/D ratio ... But I don't know if multiplying the 2 is the best option. Because if you have someone with a EXP of 2000 and they have a K/D of 1.0, and they played for ... 60 minutes, you will get. 33.333

then if another person has 1000 EXP with a K/D of 2.0, they would have also played 60 minutes and they would still have 33.333 ...

so you have to decide which to give more weight... the K/D ratio ... or the EXP ...

to solve this. multiply one or the other by 2 ...

So XP/Time=How well you build XP
K/D=How well you do with that time.

Starthing with (XP/Time)*(K/D), turn that into. [XP*(K/D)]/Time... Because I think what you do per minute is the most important. Because that allows people who have spent a small amount of time to still be worth something because they have spent that time wisely. And allows those who have perfected the art to be represented because they did even better with the time.

So, I figured out what the current 1-5 would earn (and then me):
Black: 107.844
murph: 14.3500
littlegirl: 25.1089
Dr_Jackson!: 12.4109
Doomsayer:22.9194
zainea13(me): 7.1622

now idk how fair that is, you be the judge, try it with more people.

I threw several ideas out here...

or ask the devs what they think


also, you have to award people for XP ... as they put the time into it... not awarding them for it is just wrong. so it deserves at
least 50% of the weight

oh, and i like my idea because it divides everything by time, again showing what you do per minute, as opposed to the whole time you have played

just ran the numbers for this equation ( (exp/time)*(kills/deaths) time was in hours), the final score for each of the same 16 players is as follows (with minor with some potential rounding errors):
Name:-----Player Score:---Rank:
1 Black---------6480.6--------1
2 Murphyslaw---860.9--------14
3 LittleGirl------1506.7-------10
4 Dr Jackson----744.8--------15
5 Do0msayer----1357.5-------11
6 quack quack--1812.4--------9
7 moko88-------2379.1-------4
8 Bacon squad---741.5-------16
9 Kheno---------1293.4------12
10 Rivalen--------2932.3-------2
11 Kal1981-------1918.3-------8
12 Caliaa---------2367.5-------5
13 Whitetiger-----2163.9-------6
14 Culgan--------2798.3-------3
15 ManuelFirst----1190.1------13
16 Recon911PDW-2147.4-------7

eidt: i decided to go ahead and run my numbers as well (am 12th for exp, 15th for K/D ratio), i came up with 2264.0 which would put me at 6th place among the players noted here

ignore the equation in my first post, it sucks

what about (EXP*time)/(KDR) (then divide by 100,000 to make the numbers small)

I tried it out, comparing it to me, who has a decent amount of play time, and a decent amount of XP to someone who has about the same KDR and much less XP and much less play time...

it was zainea13 vs. X303 if you want to check it... I ended up with at 4278.98 and he got a 2.05

I think xp is flawed factor in all of this, simply due to the way it is earned. You get less xp if ur team loses. This is a reflection on the team and not individual player. Also the bonus for playing just 1 class renders xp earnt in a round unreliable.

Id rather have accurate measure of how well ppl are doing now then what they did a mth ago too.

K/D in relation to total matches would be better. Time played, well piramess games can go forever. 2 sgc maps can be won in half the time with more xp.

How about kills per match - deaths per match?

When ever I jump between classes I always end up with less exp. :S xp is a bad way to define top players.

Originally Posted by Bonnonon

When ever I jump between classes I always end up with less exp. :S xp is a bad way to define top players.

Thats why i leave matches to switch classes. Also the pursuit of an even game leads too much time wasted and no xp earnt from leaving unbalanced servers.

XP is only rewarded at the end of a round correct?

perhaps a simpler solution? base it on kd, but people aren't included in the calculation until they have 10 hours of playtime AND 5000 xp. (to prevent people from sitting in empty servers to make their high kd but low totals count)

The biggest problem (atleast for me) is that the following things aren't being counted: Hurting someone but not getting the final shot, destroying turrets, destroying dispencers, and HEALING!

Originally Posted by Weapon X

I think xp is flawed factor in all of this, simply due to the way it is earned. You get less xp if ur team loses. This is a reflection on the team and not individual player. Also the bonus for playing just 1 class renders xp earnt in a round unreliable.

Id rather have accurate measure of how well ppl are doing now then what they did a mth ago too.

K/D in relation to total matches would be better. Time played, well piramess games can go forever. 2 sgc maps can be won in half the time with more xp.

How about kills per match - deaths per match?

I agree that XP is by no means ideal however it happens to be the most reliable thing we have atm. Imo points per match/deaths per match would be the best equation but we have to work with the data we have. As far as i know we don't have the points available to work with or the games played. The switching classes thing is also a bug i think so hopefully that will get fixed.


@ zania
In your example the person with 2.0 k/d would probably have more XP unless all they did was go for k/d and forget the rest of the team.
Also if you're going to start messing with the weights of the equations then you're gonna have to start bringing powers and roots in. Multiplying simply changes the magnitude of the equation answer not the weight of the different factors.

@Lantean
Those issues are with the game and theres currently nothing we can do about them. Once they introduce these we will modify to include them.


Most people seem to be coming to the conclusion of (XP*kills)/(death*time) or however you want to write it atm. I will add another table into the first post for this soon.

EDIT: just found that you can get the total games from adding up the won/lost maps thing.

Here are the values for the 16 i was using:

black: 1087
murphy: 1221
LittleGirl: 951
Dr Jackson: 1071
Do0msayer: 714
quack quack: 775
moko88: 648
Baconsquad: 762
kheno: 590
Rivalen: 572
Culgan: 144
WhiteTiger: 225
Caliaa: 225
Recon911PDW: 50
ManuelFirst: 143
Kal1981: 528

Originally Posted by Deus Viator

Disclaimer: Any comments i make in this thread are not attacks on people personally, merely trying to objectively assess peoples abilities.


As many of you know by now the wonderful site SGR:stats(dot)com has recently been updated with a rankings section. This has however created new problems: How do you define the best player? (and for all you smart ***es black is not a valid answer)

No, we need something else, a definate equation by which we can rank the players of the SG:R. So here is you challenge community: To define the undefinable, skill.




We want something that is going to last here so it hopefully should have a time element rather than having to reasses the equation every few months.

In the sgrstats thread there was an initial disagreement over whether XP or k/d ratio should be used to rank the players initially. Both of these have advantages and disadvantages.

Ranking by XP seems to reward the time someone has availible to play the game over how good they actually are at it. If we're trying to sort out the best players then surely we can't punish them for having a life? This might also cause issue if the game persists for a long time, as i'm sure we all hope it does. Players who have been here from the start will pretty much dominate the rankings for the rest of the game if this is used for too long. This could be compensated by introducing a time factor into the equation for example XP earned per week or simply total XP over time played.
one advantage of XP though is that it is based off points and your team winning. This means that kills contribute to your XP but there is absoloutely no penalty for death. This doesn't seem right as effectively thats like saying a kamikazee pilot that takes out 4 tanks and a boat is better than one who take out 3 tanks and survives. All the kamekazee pilot did was crash his plane into the ground wheras the other needed to take out enemies and dodge their fire.

Ranking by K/D means that it is possible to have someone come in, hit a few good games right off the bat and already be at the top of the rankings. Yes that player has shown a lot of skill in those games but most agree that until they are consistantly hitting that form they should not be that high up the rankings. It also favours some classes over others, ashrak being a good example, which could skew the results towards certain players.

From these points it seems that a compromise between XP and k/d is needed.


First of all i tested simply multiplying them together. To test how this affects the rankings I took the current top 10 based on the xp rankings and the top 10 based on k/d. This left us with 16 candidates due to a couple of duplicates. I ended up dividing the result by 10000 as the results were in the hundreds of thousands.

Table for (XP*k)/d

#   Name           XP        k/d       k/d rank   Skill(/10000)  Sk. rank Change
1   Black          248502    5.9110    1          146.89          1       -/-
2   Murphyslaw     173110    1.3938    14         24.13           6       -4/+8
3   LittleGirl     166364    2.3018    11         38.29           2       +1/+9
4   Dr Jackson     154510    1.0051    16         15.53           10      -6/+6
5   Do0msayer      114458    1.8779    12         21.49           7       -2/+5
6   quack quack    108056    3.2930    3          35.58           3       +3/-
7   moko88         104166    2.6798    7          27.91           5       +2/+2
8   Bacon squad    102050    1.0445    15         10.66           13      -5/+2
9   Kheno          96944     1.8644    13         18.07           9       -4/-
10  Rivalen        95078     2.9838    5          28.37           4       +6/+1
11  Kal1981        78088     2.6409    10         20.62           8       +3/+2
12  Caliaa         37810     2.8903    6          10.93           12      -/-6
13  Whitetiger     37638     3.0668    4          11.54           11      +2/-7
14  Culgan         24264     3.3156    2          8.05            14      -/-12
15  ManuelFirst    23436     2.6659    9          6.25            15      -/-6
16  Recon911PDW    8286      2.6694    8          2.21            16      -/-8

This still seems too biased towards XP as only one of the top 10 for XP dropped out of the top 10 for the skill equation. It does however seem to be more reliable, causing high XP players with low k/d and high k/d with low XP to drop down the rankings.

Next going to try out (XP*k)/(time*d)


Couple of questions:
What is the difference in XP between having an awesome game on the loosing team and a crap game on the winning team?

oh no it's "MATH BOY"

LOL!

do not get upset that just popped in my head, i mean no disrespect.

Originally Posted by Deus Viator

Those issues are with the game and theres currently nothing we can do about them. Once they introduce these we will modify to include them.

I know its not your fault. The devs should do something about it.

Originally Posted by Azag_Thoth

oh no it's "MATH BOY"

LOL!

do not get upset that just popped in my head, i mean no disrespect.

Go Math Boy!
-Colonel Everet Young

I agree we need the Devs input on this.

How about

XP - (wins * winbonus) - (losses * lossbonus)

Wouldn't that give an indicator of actual "earned" xp? You could divide the whole thing by time if you're afraid the new players will get left out.

While there are some decent calculations for a very general top players chart... until more tracking functionality is added server side there will never be anything all that accurate.

Eg: who you killed and who killed you for your k/d.

Somewhat unrelated but a very popular mod I used to play on JA allowed you to register your nick and it would do it's rank calculation vs how well ranked the people you killed/were killed by...

Cause lets face it. You could get a high rank by simply camping noobs all match or a low rank for targeting really strong players. What if you are only playing with a bunch of very experienced players wheras someone else is doing alot of rather nubish games. It's not as big an issue now as the community is small but imagine it in a few months when (hopefully) there is a large number of players online nearly 24/7. Issues like that for ranking will become more and more important...

I'm well aware that's not really something you can do with your current system as it's not tracked server side but it's definitely something I think developers should look into in the long run. Ranking is far more effective if it is built right into the game systems.

That being said, keep on with the calculations. I'm sure you will work out something quite reasonable at least until/if developers add more functionality to their current system.

Not super experienced at the game but how about this for a sample calculation..

where x is a balancing number (so easily changeable dependent on how well exp is working at the time) and time is hours played

(((exp/time)/x) * kills) / deaths

Some sample scores (using x = 1 so exp still needs to be reduced most likely)


(using random siggys I see around here...)


Deus Viator: 1039
Weapon X: 1087
bctrainers: 645
Lord Atum: 629


Sadly you didn't list the kills/deaths as seperate in your chart and I don't know any of the high ranking players id's so I cant just look them up...

What I'd do for this though is further divide the exp by 3... 5... maybe even 10 to make it less of a factor, use the full decimal values from each calculation and in the end multiply it by like 100 or something (because obviously if I made the exp divided earlier we would end up with some small numbers and decimals become important)

Originally Posted by DeliveryKing

How about

XP - (wins * winbonus) - (losses * lossbonus)

Wouldn't that give an indicator of actual "earned" xp? You could divide the whole thing by time if you're afraid the new players will get left out.

Ooohh I like the idea of putting a win/loss calculation in with exp!


Why not just add in a win/loss ratio to boost/nerf exp to a somewhat reasonable level?


So something like...


(((Player.Exp / (Player.Wins * 0.5)) + (Player.Exp / Player.Losses)) / Player.HoursPlayed) * (Kills / Deaths)

Could work... Using Deus as an example I got ~ 12 for the first section (so factoring exp, wins, losses and time played) and that is then just multiplied by the k/d ratio... so you will end up with a small number (KEEP ALL DECIMALS IN THE EQUATION and most likely multiply it out to give a score that is easily readable)

On second thought, I'm sure you could come up with a much nicer calculation... I just wanted to incorporate more elements.

i dont think having win/loss ratio be part of this is a good idea, first winning a game can highly depend on who else is on your team, and who you are up against, if you team is terrible and you are up against several very good players, then you will likely lose regardless of how good you yourself are, the opposite is also true. second, it would be too easy for players to simply drop out of a game they are about to lose just before it ends, therefore they get no loss on their record.

The only thing I have to say to you stat studying people is that I want assist kill points. I play mainly commando, I get a shot off take a guy down by 100, then someone else jumps in. I get nada for that.

I hope we get a fix for it soon

Originally Posted by ForeverZero

i dont think having win/loss ratio be part of this is a good idea, first winning a game can highly depend on who else is on your team, and who you are up against, if you team is terrible and you are up against several very good players, then you will likely lose regardless of how good you yourself are, the opposite is also true. second, it would be too easy for players to simply drop out of a game they are about to lose just before it ends, therefore they get no loss on their record.

The game throws curve balls that make all stats unacceptable though.

K/D: Does your turret's ability to kill enemies reflect your own? Does your ability to shoot at claymores mean you are a better player? What about pucks?

Hit rate: Does it show skill to splash some grenade damage on an enemy? Or to fry their brain with a hand device? Are you a poor shot if you spray bullets into space hoping to reveal an Ashrak?

XP: does it prove anything more than how long you've played?

Zain said it:

Originally Posted by Zainea13

to start, you will never be able to satisfy all players, nor will you be able to 100% accurately display THE best player. So you will have to settle for less..

Zero, you said you were experimenting with how XP is calculated. If we can find the exact xp awarded for wins/losses then we can use that to find total XP due to wins/losses and subtract it from a player's total. Leaving only xp earned through kills, capture points, etc. Divide it by time or total matches to get "earned xp" per game/time played. Just like all stats it isn't a perfect measure but I think it would be pretty good.

I'm not sure how to make this into an equation, but i think the following should be included:
XP-- not heavily weighted, but everyone has the same disadvantages, so over time, it should even itself out

Time-- i like the idea of doing exp/min

Damage given/Damage taken-- a damage ratio of some sort, not necessarily this one

K/D-- weighted about evenly with damage ratio

quick equation, maybe:
(XP/min)*(DMG Ratio)*(K/D)

just thinking out loud here

just realized, that once healing is added, DMG ratio could be (Damage given+amount healed)/(Damage taken)

nm about XP/min, it should be (XP*time)/big number depending on how much weight we want to give xp

Originally Posted by cookiec

While there are some decent calculations for a very general top players chart... until more tracking functionality is added server side there will never be anything all that accurate.

Eg: who you killed and who killed you for your k/d.

Somewhat unrelated but a very popular mod I used to play on JA allowed you to register your nick and it would do it's rank calculation vs how well ranked the people you killed/were killed by...

Cause lets face it. You could get a high rank by simply camping noobs all match or a low rank for targeting really strong players. What if you are only playing with a bunch of very experienced players wheras someone else is doing alot of rather nubish games. It's not as big an issue now as the community is small but imagine it in a few months when (hopefully) there is a large number of players online nearly 24/7. Issues like that for ranking will become more and more important...

I'm well aware that's not really something you can do with your current system as it's not tracked server side but it's definitely something I think developers should look into in the long run. Ranking is far more effective if it is built right into the game systems.

That being said, keep on with the calculations. I'm sure you will work out something quite reasonable at least until/if developers add more functionality to their current system.

Not super experienced at the game but how about this for a sample calculation..

where x is a balancing number (so easily changeable dependent on how well exp is working at the time) and time is hours played

(((exp/time)/x) * kills) / deaths

Some sample scores (using x = 1 so exp still needs to be reduced most likely)


(using random siggys I see around here...)


Deus Viator: 1039
Weapon X: 1087
bctrainers: 645
Lord Atum: 629


Sadly you didn't list the kills/deaths as seperate in your chart and I don't know any of the high ranking players id's so I cant just look them up...

What I'd do for this though is further divide the exp by 3... 5... maybe even 10 to make it less of a factor, use the full decimal values from each calculation and in the end multiply it by like 100 or something (because obviously if I made the exp divided earlier we would end up with some small numbers and decimals become important)




Ooohh I like the idea of putting a win/loss calculation in with exp!


Why not just add in a win/loss ratio to boost/nerf exp to a somewhat reasonable level?


So something like...


(((Player.Exp / (Player.Wins * 0.5)) + (Player.Exp / Player.Losses)) / Player.HoursPlayed) * (Kills / Deaths)

Could work... Using Deus as an example I got ~ 12 for the first section (so factoring exp, wins, losses and time played) and that is then just multiplied by the k/d ratio... so you will end up with a small number (KEEP ALL DECIMALS IN THE EQUATION and most likely multiply it out to give a score that is easily readable)

On second thought, I'm sure you could come up with a much nicer calculation... I just wanted to incorporate more elements.

atm theres probably only a max of 5 games going on at once so we all play the same players with a pretty high regularity and therefore i don't think it's really needed at this moment in time.

if you want to find out any stats go to sgrstats.com and go to the rankings tab

In reference to your equation i don't get why half of the things you have are in there and they seem to be in odd places. Why is the XP being divided by the wins and losses respectively? replacing the top bracket with DeliveryKing's equation might make more sense though. This would give:

[XP- (wins*win bonus) - (losses*loss bonus)]*kills/(deaths*time)
basically the same as the equation we had previously but taking out the win/loss bonuses.

Problem is that taking out the team bonuses means that teamwork is not rewarded at all and imo thats a problem.

Originally Posted by FireWraith

I'm not sure how to make this into an equation, but i think the following should be included:
XP-- not heavily weighted, but everyone has the same disadvantages, so over time, it should even itself out

Time-- i like the idea of doing exp/min

Damage given/Damage taken-- a damage ratio of some sort, not necessarily this one

K/D-- weighted about evenly with damage ratio

quick equation, maybe:
(XP/min)*(DMG Ratio)*(K/D)

just thinking out loud here

just realized, that once healing is added, DMG ratio could be (Damage given+amount healed)/(Damage taken)

nm about XP/min, it should be (XP*time)/big number depending on how much weight we want to give xp

i like a lot of this, but you cant do XP*time, it will benefit the players who go the game sooner and play a lot far too much, anyone new to the game will have no chance at catching up

only problem with (XP/min)*(DMG Ratio)*(K/D) equation, is new players could be able to shoot to the top of the charts. Maybe that equation, but require at least 10k XP and 20 hrs game play? not sure if those numbers are in the right ball field, but you get the idea.

not sure how to get total damage given/taken, if its even possible, but if someone can let me know, i can run some numbers on excel