Elim 2.0 Discussion

This sounds good.



This is no good.

Player A: plays 50 games, wins 25 with a 2:1 ratio and a 300 ave. they'll have a 600 rating plus the rating they get from wins - the rating they lose in loses. So that's minus 25 for loses and plus say 350 (if the average is 15 players per game.) So they'll have 950 rating.

Player B: player hides for wins a lot, has a 1:1 ratio with a 300 average but wins 50 out of 100 games. With an average of 15 players per game they'll have 300 - 50 + 700 = 950.

So a player with a 2:1 ratio has the same rating as one with a 1:1 ratio simply because the 1:1 ratio player hid for wins and played more. Basically it just rewards players who play the most, and actual skill means very little (but hiding for wins gets rewarded big time.)

That argument is flawed. Sorta doesn't make sense.
Can you factor in the rating change of the games he did not lose cause the player doesn't only lose 1 rating point but it changes +/-. And you can't say player B hides a lot if he [averages] a 1:1 ratio. Also forgetting the fact that player B has played more than Player A, you can figure if they both played the same amount of games, Player A will be on top tremendously. It doesn't sound right to compare where their ratings meet because someone playing one game, gets 20-5 and can have a rating of somewhere near the top 10.

Hiding IS a problem and I believe if this problem is solved, then everything else will fall in place.

Players who win should get the rewards based on their ratio along with the amount of players in arena. There are certain things you should expect from a winner, especially if there are 20 people participating (With 20 people playing, there are a possible 200 kills to obtain.) Maybe you can say that if there is anyone who wins an elim with 9-9 or less (or if the game is to 5, then it should be 5-5 or less), should be penalized for winning instead. You can also say that the more kills you get, the better the reward. The following is going to be complicated, especially when it comes to the code work, so that's why I'm only presenting this as a draft or a launch pad for something better.

Say 10-9 is the zero factor. You get no reward for winning with such a rec. The example below is brought at a 20 player game with a possible of 200 kills. For every gain in player should be a point increase and for every player lost should be point decrease.

0-0 > -10
8-5 > -3
8-9 > -3
9-0 > -2
9-8 > -2
9-9 > -1
10-9 > 0
11-9 > +1
11-0 > +10
20-9 > +10
20-8 > +11
21-9 > +11
22-9 > +12
22-8 > +13
23-7 > +15

If 11-0 brings +10 in a 20 player game. It'll bring +20 in a 30 player game.
If 11-0 brings +10 in a 20 player game. It'll bring +9 in a 19 player game.
If 11-0 brings +10 in a 20 player game. It'll bring 0 in a 3 player game.
If 9-0 brings -2 in a 20 player game. It'll bring +8 in a 30 player game.

So basically, if there is lesser than 10 players in a game, there will be no rewards for winning due to the fact that it'll be easier to hide, even when you're not trying to. If there are a large amount of players in a game, you become more likely to achieve a reward for getting lesser than 10-9 due to the fact that there was more players in the game.

Could also cover for people who have horrible aim, but get lucky enough to win with 8-9. There's no need to reward luck.
There are certain aspects in an elim that you can also factor in. All those factors should be looked at. This might help you guys think it out a bit.

[milosh]

It said the exact same thing that this page does: http://web.archive.org/web/200210071...m/erating.html

[RiiStar]

What I mean is, if it's your first game (you have no stats/rating to your name) and someone kills you, you count as a 300 rating when it comes to their average.

Before Milosh changed the Elim pages on tw.org, we had a page with a
full explaination of the rating system etc...

Is it still there somewhere milo?

[MetalKid]

You didn't read the end:

I dunno, but if you want wins to count, then you need to do something to address hiding.

If hiding gets dealt with, there is no issue.

[Lofty]

I agree with MK's idea on if it adds to the rating then negate, if lowers then dont.

I don't agree with making the wins count towards anything like that proposition, as Burnt underlined. However MK, people won't want to spec anymore as now winning WILL do something..it will give you a good w/l ratio rank bonus instead of voiding it for speccing.

Since going out with 10 deaths means you'd need to go out around 25-10 or higher to not lose rating at the higher intervals of rank.

[Static Burn]

The spec for rec thing... If it increases or equals their rating, it should be thrown out. If it lowers it, it should be kept.

This sounds good.

It is true that winning usually means that you are better off k:d wise, but people spec 4 rec a lot because winning doesn't really do anything for you. I would like to see some sort of system in place where winning the elim adds to your rating. How about this... If you lose, you will automatiaclly lose 1 rating. Then, whoever wins will get the cumulative 1 pt from everyone else.

This is no good.

Player A: plays 50 games, wins 25 with a 2:1 ratio and a 300 ave. they'll have a 600 rating plus the rating they get from wins - the rating they lose in loses. So that's minus 25 for loses and plus say 350 (if the average is 15 players per game.) So they'll have 950 rating.

Player B: player hides for wins a lot, has a 1:1 ratio with a 300 average but wins 50 out of 100 games. With an average of 15 players per game they'll have 300 - 50 + 700 = 950.

So a player with a 2:1 ratio has the same rating as one with a 1:1 ratio simply because the 1:1 ratio player hid for wins and played more. Basically it just rewards players who play the most, and actual skill means very little (but hiding for wins gets rewarded big time.)

[MetalKid]

The spec for rec thing... If it increases or equals their rating, it should be thrown out. If it lowers it, it should be kept. I wouldn't make it a hard-coded value.

It is true that winning usually means that you are better off k:d wise, but people spec 4 rec a lot because winning doesn't really do anything for you. I would like to see some sort of system in place where winning the elim adds to your rating. How about this... If you lose, you will automatiaclly lose 1 rating. Then, whoever wins will get the cumulative 1 pt from everyone else. If someone comes in and wins in a 40 man match, they'd get 40 pts and everyone else would lose 1 in addition to whatever their K ended up doing to their rating. I doubt elim will get 40 players often. Hell, I doubt it will get 20 players often.

But of course, people will hide to try and steal those points in the end. How do we fix this? You can't expect people playing elim to take the time to type out a ?cheater call because they'd die a lot. You can't expect the ppl in spec to do anything because why would they care? Could you base it on people who are 2 std deviations away from the total number of shots fired than the average for that round? I dunno, but if you want wins to count, then you need to do something to address hiding.

[Static Burn]

Ah cool, Burnt has a working link to that. Rating starts at 300? Then I wonder why mine said 0 in the link I showed with the record history after first game.

What I mean is, if it's your first game (you have no stats/rating to your name) and someone kills you, you count as a 300 rating when it comes to their average.

[Lofty]

Ah cool, Burnt has a working link to that. Rating starts at 300? Then I wonder why mine said 0 in the link I showed with the record history after first game.

And you are right Milosh, it is Elimination..you work on eliminating players, and the more you work on not dying, the less chances of you getting eliminated, the more you kill, the more chances of them being eliminated. Hence, good kill/death ratios.

I just think you aren't looking at it from quite the right perspective; yeah it can be about last man standing, but it can focus on the aspects of what it takes to be last man standing that count the most.

"Associated with each rating is a rating deviation which starts at ±100% and and goes down as you play more games. Rating deviation is displayed in brackets after rating and is omitted if it is 0. Your real rating is rating + or - deviation, - is assumed, when sorting top 100."

I wonder how this is calculated.

And to be honest, I don't think it was ever Robo Ref's intention to reward people 'for winning elim'. Like Burnt and I both said in the last couple of days, winning is it's own reward and usually for the good player k/d ratio is it's reward for winning; I think Priitk already understood that when he scripted Robo Ref, so instead he made Robo Ref maintain comparison of skill between players, since that is not as easily 'recorded' or 'rewarded' otherwise as winning is (feeling good or accomplished about it, having it stated at the end of the elim, etc). So I think Robo Ref's motives were just slightly different than what you are thinking.

I've thought about this in the past, but instead of your propositions of making it count for rank, another reward for winning could be that the 'next elimination game' global can have a little message like 'Last elim's winner was <name>! Have what it takes to win the next one? ?go elim to play' or something of the sort. You could even add a streak counter if the same person wins more than one elim in a row that would change up the message such as "Name has so far won the last # elims! Come beat them in ?go elim" or something.

[milosh]

Well, the formulation for winners could be different then the average player since they will have less deaths. I just feel that the game is called 'elimination' and thus the purpose should be, even if in a very minor way, to be the last man standing. If the win bonus somehow depends on your k:d ratio then it wouldn't promote hiding as players with a low k:d ratio would get a worthless win bonus.

[DankNuggets]

This makes sense. If you're going to negate peoples records when they try to spec4rec than winning alone would be enough reward because you don't get 10 deaths.

i can see this being enough but..

The goal of elim should be to be the last person eliminated. That person should also be the best, which should give them more kills.

But..

There are always games where someone goes 15-10 or 20-10. A lot of those games the winner might have only gone 12-6 or worse for him, 12-9.

I agree, it's a better measure of skill to have a high W:L or get the most kills in a game. However, that's not exactly what an 'elimination' match entails. While hiding is illegal and avoiding enemies is lame, there shouldn't be a penalty if someone isn't going balls-to-the-wall to win an elim.

So if someone goes 12-9, and wins by eliminating someone who went 20-10, who gets the better rating change? I'm a bit confused by the formulas right now, but i think it'll be the person who went 20-10. That may be fine if the game was 2v2 with 11-9 playing against 20-9 and he won with one kill, but if it was say 8-9 playing against 20-6, then the person who put him out (4x in a row) should IMO get a much better rating change.

Basically what I'm saying is that the whole point is to win the game, if someone goes on a suicide mission and still goes 20-10 but dies early on, they aren't really playing an 'elimination' match but more of a 'race' to XX kills (which i hope is still a game type). Ranking should be secondary to winning, but it isn't, so I'm not so sure anyone will agree with this...

And if someone can go 12-9 in a game with 40 ppl, they aren't hiding, they're either shitty with aim, or great with dodging, or both.

[Vehicle]

(winners have less deaths than those who are knocked out.)

This makes sense. If you're going to negate peoples records when they try to spec4rec than winning alone would be enough reward because you don't get 10 deaths.

[Static Burn]

http://web.archive.org/web/200210071...m/erating.html

Old elim rating: Wins/losses*Ave, everyone starts at 300 rating.

I say no full charge greens, it's a lame tactic imo. Also I don't see why there should be a bonus for winning, this will just promote hiding and winning is it's own reward (winners have less deaths than those who are knocked out.)

[Lofty]

As you can see, after a certain point (around 600's rating), even going 16-7, which is higher than 2:1 ratio, gave me a little negative rating.

That is how strict elim has to be, in order to really be a good method of competitive fun. After having to get used to such strict calculation for a while, I can now easily go into elim and guarantee a win with a good ratio on command. To set a goal to rank into 1st in elim back in the day used to be one of the best way to factor lots of things into training and practicing in a wb - dodging, shooting, watching for vulches, vulching, handling outnumbered situations...scouting for vulchers on radar etc etc.

I myself have been curious about how the formula works exactly and 'The Rating System' link above the elim ladder picture I showed, but that link hadn't been working for years before Robo Ref went down so I never really had a chance to read it.

It's supposed to be http://www.dcee.net/elim/erating.html

And like I said, those rankings only appeared on the ladder once a combined total of kills and deaths equaled 500 or more. It is really not that much to ask of a person, considering the whole ranking time up there was 3-4 days. I did do a lot of games per some of those days, but for someone going at a slower pace it would take a week, maybe a week and a half. It's not that bad.

[milosh]

Thank you, I'm glad someone has at least tried to give me some resources to figure out the previous formula. I really do appreciate that, Lofty. As for the previous arguments... I'm fine with adding fc greens if the majority desires it and I'm fine with increasing the void ratio as I've previously stated, but I have to disagree with making the win bonus a hard set value. Winning an elim versus 40 players should give a bigger bonus than winning one against 3. As for hiding... yes it should be illegal and monitored, but not by the bot. Teaming in elim is also illegal but it us unreasonable to expect that the bot should monitor that. That is a moderator issue... if you see a player consistently hiding use the ?cheater feature to report it. Making the bot evaluate hiding would be difficult and unreliable. I would have to make the bot constantly track the position of every player in the arena and compare them, and even if I did get that to a functioning and working order, sometimes you just can't find anyone to fight.