GG Rank System

[ZERO]

You can see example stats here: http://www.ibisgaming.com/gg_stats/

This is just a fee hours worth of data. I am still building the actual parts that let you see wins ect so keep your pants on.

Please note you can see the current win system in work. Notice how the 2 best pacers have the same numbers of wins and then are mostly separated by the gglevel up events.

Now currently we can not count for gg_lose events. However losses starting today will count. So for example if we pretend that this is rank data from today twurk would be in second place which makes more sense as the current data shows that that is where he would belong if we are trying to clearly evaluate the player skill.

[ZERO]

Lets compare how the ranks will differ from those before today and after today: (this does not take into account the higher number of level down events also we do not independently track leader or last level data but you are awarded for it in both of these models)


----------------BEFORE JULY 1st---------------------
Formula: WINS(5000) - Losses(0) + Levelup(5) - Leveldown(5) + knivesteal(10) + leader + lastlevel

twurk = 1st
twurk=42032skill
twurk=8(5000) - 1(0(not counted)) + 345(5) - 34(5) +43(10)= 40000+1725-170+430=41985 [42032-41985=47 times user triggered gg_leader or gg_last_level]

Mutnaggewinu =2nd = 41976skill

----------------AFTER JULY 1st---------------------
Formula: WINS(5000) - Losses(2500) + Levelup(5) - Leveldown(5) + knivesteal(10) + leader + lastlevel

Mutnaggewinu =1st = 41976skill (currently he has no losses so skill does not change)

twurk = 2nd
twurk=39532skill
twurk=8(5000)- 1(2500) + 345(5) - 34(5) +43(10)= 40000-2500+1725-170+430=39485 [39532-39485=47 times user triggered gg_leader or gg_last_level]

So as you see from the data I believe these values to be quite logical. It is based mostly off win# but from there is able to then separate out players that have close to the same number of wins and define who is better.


Please post comments about how you all feel about this ranking.

[Carmichal]

IDK if it is too late for this, but can you please change GG_lose to maybe -500; also GG_win to +1000. IMO it seems a bit much to get +5000 for a win, and a loss is -2500 when the rest of the ways to gain points are so far from that.

IMO. There should be no GG_loss. I can see a lot of regulars, who may not be the best getting frustrated trying to dig out of that hole since you need to kill 500 people to just get rid of that. 5000 thousand point for a win in over kill! Yes you should get a hefty reward for winning, but 5000? I can see many more people get frustrated. I can see people leaving, and that is not good for the server.

If I had my way all the time you would only get 500 points for a gg win, and no points lost for having been naded.

[ZERO]

In the previous stats you got 1 to <1point per kill and 500 points for winning and -500 for losing. You also lost points for level down and got no points for being a leader or having the last level.

Previous rank was determined basically only by wins. This rank is also intended to be win base but also account for who is better for players with close to the same stats.

Here is the rank shown again and then the explanation:
WINS(5000) - Losses(2500) + Levelup(5) - Leveldown(5) + knivesteal(10) + leader + lastlevel

Winning a gungame is 1000 times harder than leveling up and 100 times harder than stealing a level. IT is also 5000 times harder than being the leader or on the lastlevel.

Being the person killed at the end of the gungame is 500 times harder than losing a level.

Stealing a level is twice as hard as leveling up (you also get the level up so when you steal your effectively getting +15)

A better player would logically be one that has more wins relative to losses. Has more levelups than leveldowns has more steals than another player, is more often the gg leader and is on the last level more often.

The current formula makes it so that two players with the same wins makes it so if one has 1 more win give they effectively have .5 as many.

Who is better a player with 500 wins and 20 wins given or a player with 495 wins and 10 wins given to other players. Oh wait these are the same... the amount of losses you need to go down is more than you think. A player with 505 wins for example would need 30 losses to be ranked the same!

Lets take a look at how this formula effects the rank of the past stats:

(remember old stats would drop the skill points if it got too big)
Here is the top 3 winners from 2007 to 2011 and their rank followed by what their rank would be now:
Xerenix | ibis.a RANK175 1,148wins 121losses
Nissan GT-R R35 <ibis.a> PEW™ Rank5808 1,041wins 123losses
Fluffy Frufflebottoms RANK8 809wins 32losses

It is clear given the massive difference in wins that these players should be in 1st 2nd and 3rd. However how does our new rank better deal with people that are close together on wins who deserves the higher rank? Lets take a look:

	wins	losses	L+	L-	skill
RellikDloc ™♠	756	54	155,884	5,991	4394465
Penis シ	744	194	123,239	4,787	3827260
yed	719	93	55,603	1,302	3634005

So as you see even with a lot of losses it takes a huge amount before you start dropping below someone else. The amount of levelup events builds up over time as well thus giving the rank of a player that plays a lot a slight buff over someone who just started. This makes ranks stick more which is also logical in a ranking system.

But we can compare more from old data. Here we see how the number of losses still does not cover the difference that simply having many gglevel up events does:

Bink of the North Village	533	53	39,109	770	2724195
کHÅקY V1яU5 ibis.a	512	272	232,259	10,751	2987540
Envy | Ibis.a	501	166	86,933	3,785	2505740

In that example HÅקY V1яU5 ibis.a still had a higher rank than Bink of the North Village despite having less wins. In this case blink would need to keep playing longer in order to close the gap. So we see here that there is already a system that takes into account how much a player is actually playing and rewards them. However we do not want that reward to be too great. Now that score may be a lot closer if we also accounted for the other parts of the model like steals last level and leader but that data does not exist for these players.

So do we still think that the difference between wins and levels is too much? Or is it not enough now? Do we see how taking points for causing someone else to win effects the rank but now critically?

Thoughts?

---------- Post added at 07:49 PM ---------- Previous post was at 07:35 PM ----------

Here is another example:

	wins	losses	L+	L-	skill
Rated EC-10	462	274	269,387	8,524	2929315
[???.?]<?????¡??>	456	88	39,995	837	2255790
Vagina ツ	455	162	144,434	6,950	2557420
lietuvis	450	71	30,066	556	2220050

In this case we once again see how having a lot of levels can equate to gg wins on their own over time. Even though ?? has more wins to losses Vagina has way more level gains and thus enough to still be ranked above.

Lets see for some people that have the same gg wins and if the data is looking logical:

	wins	losses	L+	L-	skill
Maladjusted	278	119	138,864	7,358	1750030
Kr@zy Simon	278	75	45,018	1,652	1419330
Toffee	277	63	59,300	2,254	1512730
Tyrael	277	73	71,268	1,639	1550645

Now we see again that longer time player comes out ahead even when the gg losses are stacked against them. Maladjusted clearly the worst player out of this group but still has a higher rank right now. This will change though in a few more weeks as the other 3 are gaining wins and other stats at a faster rate.

From what I can see this system allows ranks of long time players to be protected but not at the cost of preventing new players from moving up in the ranks at a reasonable pace. Can we get any comments or suggestions as to other ideas.

If someone else has a more fair way of dealing with the ranks and can show the math I would be happy to use it.

[brett friggin favre]

here's my concern: should it be based on "difficulty" or on probability? especially for giving up the gg win, saying it's 500 times harder than losing a level...well look at it probability-wise. let's say it takes 15 rounds and there are 20 players. the average number of deaths would be 15, multiplied by 20 players = 300. that means that about 1 in 300 deaths gives up a gg win. similarly the average number of kills should be 15, again 1 in 300 kills grants a gg win. but the other question is whether or not we should punish the one who gives up the win so severely. probability dictates that you stand (on average) a 1 in 19 chance when the server is full of giving up that win. the fewer deaths you have, the lower your chances, and vice versa. but since it's just a matter of probability, should it really cost 500 kills if you just happen to be in the wrong place at the wrong time?

another point i hinted at in there is that the good players will get their wins, indeed, but they will also have excellent k/d ratios. so should being the lucky one to get a win be so much more advantageous? to where it's worth 1000 kills? ALSO if some people, say, fluffy, penis, mut, and myself, all get on at the same time...everyone else in the server will suffer rank-wise, because they can't get that win worth 1000 kills, and those of the top players in there (competing for high ranks) would also suffer as we stand much less chance of getting a win as compared to when it's just one of us and a bunch of scrubs. it is my personal belief that the rank system shouldnt be so dependent on how many hours you've put into the server. it should prioritize these three things (after a certain point with wins): win%, kdr, knife kills. to me, those things are the truer display of skill than putting in 900 hours and winning the probability game. after all, the determining factor in this rank is called "skill" and not "determination". i think we need to think in terms of what gets you wins, not just that you have wins. kdr and win% will get you wins, so being the base of what makes you a successful player, it should be the base of the stat system imo.

[ZERO]

The issue is that the only time death is counted against you is when you actually lose levels or lose the game.

---------- Post added at 08:52 PM ---------- Previous post was at 08:48 PM ----------

Also I was using those numbers to demenstrate the value of one thing over the other. You notice that in the stats that the best players have much less gg_lose events for the number of wins. In the existing stat system I have aldrady shown in mutiple examples above that the new strucutre allows for those with less wins to have a better score than someone with more by that plaer having more level up events.

I would like to see someone actually do the math and see if they can come up with something better. You can see the example data mutiple times in the previous posts.

[ZERO]

Lets Compare some various systems in these we will ignore stats from stealing levels, leader and last level b/c I do not have real world data for those on these ranks.

This is data built from 4 YEARS worth of gg stats. These are real numbers from actual players over that time. Lets see how my system ranks them and how other ideas rank them. Then let us decide which of these methods is the best way to rank players. These ranks come in 3 groups.

• Top 5 level rank
• Top 100 level and close rank
• Low level rank

So here is some variants:

--------------------------------ZERO Ranking-----------------------------

	wins	losses	L+	L-	skill	RANK	online time
RellikDloc ™♠	756	54	155,884	5,991	4394465	1	523:01:32
Penis シ	744	194	123,239	4,787	3827260	2	768:24:52
yed	719	93	55,603	1,302	3634005	3	476:20:22
Maladjusted	278	119	138,864	7,358	1750030	4	587:32:53
Kr@zy Simon	278	75	45,018	1,652	1419330	7	269:46:44
Toffee	277	63	59,300	2,254	1512730	6	277:05:53
Tyrael	277	73	71,268	1,639	1550645	5	259:53:38
ying	17	11	51,828	2,180	305740	8	97:17:13
Killer	17	9	4,814	271	85215	9	33:26:10
kamizzza	17	2	913	11	84510	10	9:31:33
traiblazer	17	48	12,097	379	23590	11	142:07:20

-------------------------------------END ZERO RANK-------------------------------------------

-------------------------------------1000ggwin -500gg loss------------------------------------

	wins	losses	L+	L-	skill	RANK	online time
RellikDloc ™♠	756	54	155,884	5,991	1478465	1	523:01:32
Penis シ	744	194	123,239	4,787	1239260	2	768:24:52
yed	719	93	55,603	1,302	944005	3	476:20:22
Maladjusted	278	119	138,864	7,358	876030	4	587:32:53
Kr@zy Simon	278	75	45,018	1,652	457330	7	269:46:44
Toffee	277	63	59,300	2,254	530730	6	277:05:53
Tyrael	277	73	71,268	1,639	588645	5	259:53:38
ying	17	11	51,828	2,180	259740	8	97:17:13
Killer	17	9	4,814	271	35215	9	33:26:10
kamizzza	17	2	913	11	20510	11	9:31:33
traiblazer	17	48	12,097	379	51590	10	142:07:20

This method produces the exact same results however we see that the difference in skill between Maladjusted and yed has greatly reduced. With this method lets say all else equal Maladjusted can get the same score as yed if he just had 67 more wins. So do you think below is a good result?

yed	719	93	55,603	1,302	944005	SECOND	476:20:22
Maladjusted	346	119	138,864	7,358	944030	FIRST	587:32:53

I just do not see how Maladjusted should have a higher rank than yed sure he has played 100 more hours but he still have hundreds of less actual wins and more actual losses. The point of gg is to win is it not? This is why I reject the idea of setting the difference between wins and level gains so low. Also please remember that level steals have a 10x multiplier and if that was in here it would be even messier for yed.

-------------------------------------END 1000ggwin -500gg loss------------------------------------

-------------------------------------ZERO with no losses-------------------------------------------

	wins	losses	L+	L-	skill	RANK	online time
RellikDloc ™♠	756	54	155,884	5,991	1505465	1	523:01:32
Penis シ	744	194	123,239	4,787	1336260	2	768:24:52
yed	719	93	55,603	1,302	990505	3	476:20:22
Maladjusted	278	119	138,864	7,358	935530	4	587:32:53
Kr@zy Simon	278	75	45,018	1,652	494830	7	269:46:44
Toffee	277	63	59,300	2,254	562230	6	277:05:53
Tyrael	277	73	71,268	1,639	625145	5	259:53:38
ying	17	11	51,828	2,180	265240	8	97:17:13
Killer	17	9	4,814	271	39715	10	33:26:10
kamizzza	17	2	913	11	21510	11	9:31:33
traiblazer	17	48	12,097	379	75590	9	142:07:20

With adding a few wins to Maladjusted

yed	719	93	55,603	1,302	990505	2	476:20:22
Maladjusted	346	119	138,864	7,358	1003530	1	587:32:53

Here we have created a real clusterfuck. Basically wins have no real value. This may be surprising b/c what we did was remove the cost of losing which one would think would increase the value of wins. Instead it has a reverse effect.

The reason for this is simple as stated earlier in the thread stats should be based off probability and that is correct. Good stats should take into account the chance of a bad player winning vs a good player wining. yed here managed to get hundreds more wins in 100 less hours than Maladjusted however is ranked lower. It is clear from the stats that most of Maladjusted wins are likely him getting lucky and having played a lot relative to yed that is. This is clear in both the time played and the wins he has gotten. Random chance says the players have == likely hood of winning and losing when these players are on the last level. However we reward winning 2x as much as losing. There is yes a higher chance of losing b/c you can lose on any level. However the data shows that most players with decent wins have more wins that losses. This we would expect. Just as we do not want to count only level up counting only wins and not losses makes it virtually impossible for new players to ever rank up. There needs to be balance, balance that a reletive scale by subtracting some points for losses takes care of. As you see in the original model there is no such problem.

Therefore any method that does not include subtracting losses can not be accepted. ALSO PLEASE NOTE THAT LOSSES PRIOR TO JULY 1ST CAN NOT AND THUS WILL NOT BE COUNTED SO ALL PLAYERS PRIOR TO THAT DATE BASICALLY GET A TEMP BUFF THIS SHOULD GO AWAY OVER TIME THOUGH SO NO WORRIES LONG TERM.

-------------------------------------ZERO with no losses-------------------------------------------

So here I have shown mine and 2 popular other ideas and explained and showed with real rank data why my method retains more logical results. If someone can show my model creating illogical results or another model wich creates better results I want to see it.

[XX0wnsXY]

Like I said previously, I don't mind anything but the losing team aspect, which seems to have been removed anyway.