To Define the Abstract

( ( (xp*kills) / (xp*deaths) ) + ( ( xp*matchwin) - (xp*matchloss) )/ time


or else:

( ( (xp*kills) - (xp*deaths) ) + ( ( xp*matchwin) - ( xp*matchloss) ) )/ time

would be interesting to see

Originally Posted by thewesdude

( ( (xp*kills) / (xp*deaths) )

In this the xp cancels itself out so it's just k/d ratio

@Firewraith
We basically either have an equation based on time where it's possible for new guys to hit high up the rankings pretty soon or one not where the old timers will dominate utterly and it will be unfaiir on the new guys. Unfortunately i don't see any middle ground in this one.

My suggestion, the data is all there.

Assuming maps won/lost are recorded properly (ie if you leave a server before round ends does that count as a round played?)

3 sets of data which are available.

Total matches
Kills
Deaths

(Total kills / total matches) - (Total deaths / total matches)

Using a few stats from the top 20 players, myself and punisher first since we are quite close most of the time.. (posted numbers so someone can confirm my math if they want and see the numbers in play)

Weapon X (current rank 20)
(3258/ 442) - (1623 / 442)
= 3.699

Punishr (current rank 19)
(3660/476) - (1719/476)
= 4.0777

ForeverZero (current rank 12)
(3224/580) - (1360/580)
= 3.214

Black (current rank 1)
(12166/1090) - (2045/1090)
= 9.285

MurphysLaw (current rank 2)
(6977/1254) - (5018/1254)
= 1.562 - sorry murphy

LittleGirl (current rank 3)
(7569/990) - (3216/990)
= 4.397

Its basically a KDR over time equation and i'd say much simpler than many others proposed. Granted we currently dont have assist points it doesnt reflect gameplay other than killing and survival. Cap POI doesnt take any skill imo - if other team is pinned back to spawn u can cap at will. Flag captures dont give points. All that really matters are kills right now in the game which gives majority of points.

Later today I might do the calculations properly for the top 20 players. Just did this roughly (manually) instead of a nice spreadsheet lol to give you guys an idea of how it would look.

Edit: After giving this more thought I realised it isnt that simple. This formula doesnt work at the low end of the spectrum. There would need to be a minimum matches played criteria, perhaps to determine the top 50 players.

Also about XP - There is an artificial ceiling on this as you would not earn xp after hitting max rank (15), correct?
In time this would only serve to bolster those who play multiple classes over just 1 for instance.

you need to know "what do you want to see?" to derive such an expressions.

come up with your problem statements ("aim") and the equations will the a logical leap from their. Its how we roll in physics when a formula is not evident from results or an experiment can not be devised. This is basic dimensional homogeneity


So what we want is something based on the following relations

kill/death
experience/time played
win/loss


this gives us the units of

Kills wins and experience per death loss and time played.


the equations is fairly evident now, all that needs doing is assigning sensible static values/powers/roots to these relations so that things such as time played per experience gained is important enough to ensure that only a god like player with only 5 hours under his belt can be number one...

and that's a job for mathlab... or perhaps someone with excel and way too much time on there hands. I would probably look into using a squares and a cube to start with and experiment from there. Oh, and always test the most extream (banana ) cases to get an idea of the limitations to the formula your propose.


------------

just saw this post after posting:

Originally Posted by Weapon X

......
Assuming maps won/lost are recorded properly (ie if you leave a server before round ends does that count as a round played?)
......

I am sure (not 100%, but almost positive) that none of your stats for that round count if you leave a match before it ends.

Originally Posted by foxandthehen

just saw this post after posting:



I am sure (not 100%, but almost positive) that none of your stats for that round count if you leave a match before it ends.

lol

I keep discovering why it seems i've wasted so much play time

Was wondering when you were gonna pop your head in here fox

I ran your equation and here's the results:

#  Name        XP      Time(m)  Kills  Deaths  Won  Lost  Result
1  Black       253602  13947    12166  2045    896  194   499.61
2  Culgan      24952   1769     836    253     127  27    215.94
3  Rivalen     105628  6672     4907   1554    460  150   153.30
4  Whitetiger  38982   3989     1728   555     189  40    143.77
5  Caliaa      37810   2771     2107   729     162  60    106.48
6  Moko88      109778  7246     5876   2165    417  232   73.91
7  quack quack 108056  11781    4541   1379    550  231   71.91
8  kal1981     80158   6657     2915   1112    370  169   69.11
9  LittleGirl  175498  15925    7548   3205    656  334   50.97
10 Do0msayer   118658  9796     4172   2275    470  241   43.32
11 Kheno       100774  8690     4428   2332    381  210   39.95
12 recon911PDW 8286    637      323    121     25   24    36.17
13 ManuelFirst 23436   3149     1125   422     85   67    25.17
14 Murphyslaw  180022  17242    6964   4995    771  483   23.24
15 Dr_Jackson  158294  12706    4249   4202    569  501   14.31
16 Baconsquad  105450  8886     4212   4063    361  427   10.40

Seems to favour k/d and w/l a bit so i'll mess around with the weightings a little in the morning.

Originally Posted by Deus Viator

Was wondering when you were gonna pop your head in here fox

I ran your equation and here's the results:

#  Name        XP      Time(m)  Kills  Deaths  Won  Lost  Result
1  Black       253602  13947    12166  2045    896  194   499.61
2  Culgan      24952   1769     836    253     127  27    215.94
3  Rivalen     105628  6672     4907   1554    460  150   153.30
4  Whitetiger  38982   3989     1728   555     189  40    143.77
5  Caliaa      37810   2771     2107   729     162  60    106.48
6  Moko88      109778  7246     5876   2165    417  232   73.91
7  quack quack 108056  11781    4541   1379    550  231   71.91
8  kal1981     80158   6657     2915   1112    370  169   69.11
9  LittleGirl  175498  15925    7548   3205    656  334   50.97
10 Do0msayer   118658  9796     4172   2275    470  241   43.32
11 Kheno       100774  8690     4428   2332    381  210   39.95
12 recon911PDW 8286    637      323    121     25   24    36.17
13 ManuelFirst 23436   3149     1125   422     85   67    25.17
14 Murphyslaw  180022  17242    6964   4995    771  483   23.24
15 Dr_Jackson  158294  12706    4249   4202    569  501   14.31
16 Baconsquad  105450  8886     4212   4063    361  427   10.40

Seems to favour k/d and w/l a bit so i'll mess around with the weightings a little in the morning.

Have been away for a few days and am ill atm! Sorry for the delay! hehe

and as for your initial results; thats the ticket!

now, as you have said, all that needs doing is a little weighting. I suggest using a graph and powers to try and get an exponential curve which increases the likelihood of having the highest score with respect to time, but which is controlled via the other ratios. have fun!

Deus, my equation would allow new people to shoot up the charts, which is why i proposed a XP and time threshold before they're allowed on the charts. The threshold would be set at a point where they would have to be regular players before they would be considered for the charts.

Originally Posted by foxandthehen

you need to know "what do you want to see?" to derive such an expressions.

come up with your problem statements ("aim") and the equations will the a logical leap from their. Its how we roll in physics when a formula is not evident from results or an experiment can not be devised. This is basic dimensional homogeneity

i fully agree with this idea so lets break it down:

The problem: determine a method of ranking players based on overall skill, using multiple stats to ensure that all good (not just high K/D or high exp) players are ranked higher, keeping brand new players out of top ranks without preventing good players who join later from being able to ascend the ranks and eventually reach the top.

based on this we need to ask: What do you look for in a good player?
high experience with the game, knowing the maps, classes, and weapons, and how best to use what your class has, as well as the map terrain, to deal with an opponent.
killing ability, being able to hunt a target down and finish them off, therefore removing them from the battle for a short period of time as well as scoring your team points.
survivability, being able to evade enemies when low on health so you can fall back to get healed and rejoin the battle sooner than if you died, also prevents opposing team from gaining points for your death.
damage output, dealing a high amount of damage to opponents, making them easier for your team to kill, or forcing them to retreat allowing you to make an advance.

other factors: (not being used)
accuracy/headshots, this may heavily depend on class (commandos/ashrak are likely to have higher accuracy), also having a low accuracy does not prevent the player from being able to output high amounts of damage or gain many kills.
damage taken, often this can be unavoidable, while you have an amount of control over this by being able to dodge, however i dont feel you have enough control over all situations to make it viable for this equation

interpreting the factors:
kills, deaths, and damage output are are simple enough to have a value for, gaming experience is harder, while its not a perfect value i think exp is a good estimate for it, i have determined that if you do better in a game you get more (if we could eliminate the base value you gain for winning or losing and only use what you gained for your performance, it would be almost perfect in my opinion). then to go back to the problem itself compare all of this to time played to see a good average of the stats by using a ratio and not sum totals, which would prevent any new players from reaching the higher ranks.

the equation: (exp*kills*damage given)/(deaths*time)
this is very similar to our previous equation, i think all we really need are a few more minor tweaks, if anyone sees anything i missed do make comment, other factors can (probably should) be added to this equation. maybe knowing all the stats that can be tracked would help us work on this, so we know what we can easily use.

ORRR how about just making a bracket on the ranking for those with less than x amount of play time. and do your funny little equations on those

then


another one for people above x amount of play time. and then having special things

that say top 5 biggest movers this (week, day, month) for each

we could leave that current stats as overall stats and then have a seperate stats.
i.e

I haven't even noticed this thread due to the title

I guess, omething like this should probably work:

cos((sqrt(xp) * pow(kills,3) / integrate(time + headshots) / rand(0,1) * rand(1,10)) / sin(deaths))

Ok, enough jokes, I will wait for you to come up with something "fair"

by funny i mean i dont understand them and

after a little searching im hard pressed to find a game that doesnt initially use xp to determine ranking.

whether it be by xp gained since game start, or monthly or weekly. the problem i think we have is that xp is given out a bit "funny" in this game. your punished for attempting teamwork xp wise

on even more thought how about stats for best ashrak etc etc prolly should post this in other threadx

Sorry for not doing anything on this for a week as i've just moved back to uni so been a bit busy.

I ran all the equations in the thread and weighted the one i did last and these two seemed to be the best fit. In addition to the 16 people i had previously i added the person with the lowest k/d and a decent amount of time played and one person from half way down the table to test the paramaters.

Here are the equations and the results for them.













Imo the latter of these is better because black isn't crazily out infront and newbies don't shoot to the top but will climb pretty fast once they start playing. I also tried working out the Exp without the bonuses but was getting back negative numbers so not sure what was going on there.

Does anyone have any problems with these equations?

Im just curious about the square root of the kdr bit..

Figuring out xp without the bonus from win/loss sounds tricky.

For example: Based on my stats I worked out that after bonuses and kill xp I calculated to about 8k short of my total. Now that must be a crazy amount of poi captures & possibly Armana kills count as 4xp? opposed to other maps.
(assuming Win = 160xp : Loss = 120xp & Kill=2xp) Ntm if you switch classes you lose bonus during the round which is what I suspect many ppl do. Xp is the funny data that will muddle up your calqs.

Edit: Need to work out how much xp you get for a kill, i was using 2 points lol.

Having read through this post, to many people seem to assume you need to use all the major stats to define how the top ten should be defined.

There are major flaws to using all the stats:-

Certain classes have higher KD to other classes, so this will obscure the result.
Certain maps take much more time to complete a round, so this will also obscure results.
Taking into account overall experience gained will obscure how a player actually plays, because his team bonus will be added to his exp.

My formula for this is much more simple and is based on pure exp and games played, a lot of time has gone into experience awarded in a game, so its more than likely to be a lot more accurate coming from SGR itself.

Any way :-

Take a players total EXP (A)
Take a players total matches win (B)
Take a players total matches lost (C)

If 160 is bonus for win and 120 is bonus for loss. ( I am not sure what they are)

(B * 160) + (C *120)
---------------------
C

This formula works out how much exp you gained per round (averaged).

You could add further things on to check for KD (biased to certain classes), head shots (biased to certain classes), time (biased to certain maps), that will then obscure the data even more.

It would be interesting to actually see how the top 100 changes with this formula...

Originally Posted by yellowperil

Having read through this post, to many people seem to assume you need to use all the major stats to define how the top ten should be defined.

There are major flaws to using all the stats:-

Certain classes have higher KD to other classes, so this will obscure the result.
Certain maps take much more time to complete a round, so this will also obscure results.
Taking into account overall experience gained will obscure how a player actually plays, because his team bonus will be added to his exp.

My formula for this is much more simple and is based on pure exp and games played, a lot of time has gone into experience awarded in a game, so its more than likely to be a lot more accurate coming from SGR itself.

Any way :-

Take a players total EXP (A)
Take a players total matches win (B)
Take a players total matches lost (C)

If 160 is bonus for win and 120 is bonus for loss. ( I am not sure what they are)

(B * 160) + (C *120)
---------------------
C

This formula works out how much exp you gained per round (averaged).

You could add further things on to check for KD (biased to certain classes), head shots (biased to certain classes), time (biased to certain maps), that will then obscure the data even more.

It would be interesting to actually see how the top 100 changes with this formula...

To find out the xp without bonuses it should be
total xp - ((wins*160) + (losses*120))

or A-((B*160) + (C*120))

Then divide this by B+C to get the exp per round. I tried this though and it gave me back negative numbers so it may be theres a problem with the stats or that we have the bonusses wrong. I'll try sticking it through again and seeing if it works this time.

EDIT: Just ran it again and again its giving my negative answers. I don't think the stats are tracking properly so getting the wins without the bonuses may not be possible.

Originally Posted by Deus Viator

To find out the xp without bonuses it should be
total xp - ((wins*160) + (losses*120))

or A-((B*160) + (C*120))

Then divide this by B+C to get the exp per round. I tried this though and it gave me back negative numbers so it may be theres a problem with the stats or that we have the bonusses wrong. I'll try sticking it through again and seeing if it works this time.

EDIT: Just ran it again and again its giving my negative answers. I don't think the stats are tracking properly so getting the wins without the bonuses may not be possible.

because you not accounting for before the last patch, as before that patch when you changed class it removed a large amount of your experience for the round.

So, your formula works, but only on players who started playing after the last patch. for player before the last patch, you would have to know how many matched they changed class in and if they won or lost the match, otherwise you will get negative numbers if they change class in a lot of matches.

fox = maths god.

Originally Posted by foxandthehen

because you not accounting for before the last patch, as before that patch when you changed class it removed a large amount of your experience for the round.

So, your formula works, but only on players who started playing after the last patch. for player before the last patch, you would have to know how many matched they changed class in and if they won or lost the match, otherwise you will get negative numbers if they change class in a lot of matches.

fox = maths god.

Scratches head. I'll leave you guys to it