Elim 2.0 Discussion

[Lofty]

Before reading this, do take note of the stuff I mentioned in my previous post.


Anyway, in an attempt to bring back the most accurate ranking possible to what it used to be as we do not have access to Robo Ref's script, I will provide what data I do have that I've found.

There have been arguments in the past concerning legit/non-legit possibilities of ranking in elim, and I will use references from those to benefit our explorations, and also to debunk Spacely's ridiculous claims from yesterday in elim spec about how 'we just are mad we can no longer abuse spec for rec with the new potential system and need to win for a change now'.

Here is a complete statistics table for my Sirius name, which demonstrates ranking from scratch 0-0 of everything, to #1st place on the elim ladder. Milosh, you may try to use it to derive what formula was being used to calculate this.

http://i90.photobucket.com/albums/k2.../elimstats.jpg

And the final result on ladder -

http://i90.photobucket.com/albums/k2...elimladder.jpg

Now, a few things should be noted here. On the statistics page, the 'Rating' column only shows the rating you had BEFORE that particular games' rating adjustment. The next game up shows the adjustment AFTER the previous game. As you can see, my first game got me 354 rating, but it shows 0 because that is what my rating was before that game.

Now concerning Spacely's argument, note that I had not once 'specced for rec', won about 75 % of my games, and did not just 'chase newbies and run from better players' as I had the highest Ave. out of the top 25 players. Hopefully those stats will help round down or confirm the formula used.

Also, the way Robo Ref responded with information when told !rank (if you want to use it at all) was

Player> !rank
Robo Ref> No.# Rating #(+-#) Wins:# Games:# Kills:# Deaths:# Ave:#

[milosh]

I'm fine with increasing the void ratio.

[Zeebu]

i agree with lofty a bit.

maybe instead of making it just positive, make it not count if its closer to 1.5:1 ratio instead

9-5 wouldnt count, but a 9-6 would

[Lofty]

I agree with ALL those points MK brought up, including the speccing a hider 3 times...but maybe make it 5...because sometimes if the arena is big and there's like 5 people left there are times you keep going to ALL the wrong places to find the others and it would suck to have it pwn you out just because of crappy circumstance.

However, I dont agree with the take out greens thing. It used to be somewhat strategic to have greens around, even if they are full charge. It requires extra care from opponent, especially to take note of when someone ate a green.

I think a simple void will do if you are positive. It will count against you if you leave and you have a negative win ratio though. - milosh

I think this could be abused. If you are trying to rank by the old formula, and you have say....40-12, your rating will be around 900-ish +/- whatever ratings of enemies you've been killing effect, (and yes it WONT be ranked yet because you havent reached the 500 kill/death total to be ranked), but suppose that is your rank, and you are working on it.

Now say in the next game, you fuck up big time and go 11-9 and are in game atm..11-9 if you spec would make your 40-12 turn into 51-21, and you would lose like 200 rating to 720-ish or something. However, if you can just spec and void that mess-up because you are positive, that would be horrible.

Also, don't forget, make it so that the stats and stuff reset every 5 months or so. I don't know how Robo Ref did it exactly, but he basically 'upkept' the ratings and kill deaths, and deleted the oldest stats past a certain point..I just don't know what point it deleted them from. So it didnt just wipe all of it after 5 months...it like wiped a certain portion of the oldest stuff in your rankings from many months ago, but kept the newer stuff from like the last couple months. If you hadnt played for a few months, then there would be a good chance your whole ranking would get wiped, which can be good to start anew as well.

[milosh]

- If you spec, your kills won't count (but your deaths do).

I think a simple void will do if you are positive. It will count against you if you leave and you have a negative win ratio though.

- Allow up to 3 lagouts, but you must return in under a minute.

Lagouts are already allowed and I may add a time limit to them.

- If the bot determines you are hiding 3 times, it will spec you and treat it the same as speccing for rec? Tho, we would need to determine what constitutes as hiding because sometimes you just can't find anyone.

This is vague and (I feel) unnecessary.

- Winning an elim should bestow some sort of reward to your rating. Maybe add a hardset amount? Like 5 rating, maybe?

I disagree with this as winning an elim against 40 players should give more points than winning one again 3.

- It would be nice if deaths don't count if you die in less than 1 second after spawning... Spawning on bullets only takes away from skill. Though, if you manage to kill someone in less than a second after you spawn, it shouldn't count either.

I thought I had already implemented this... I believe it works; let me know if I'm wrong.

- You may want to take the greens out of that arena, too. People like me (runners) tend to abuse the Full Charges when given the chance.

There are no full charge greens in the arena.

[MetalKid]

So, that would be:

Wins / Loses * Average Rating of kills

Everyone starts out at 300 rating.

You have to have at least 500 total wins + deaths in order to rank.

I don't know how the +/- rating part works in the formula, however.

Additions I think will improve elim:

- If you spec, your kills won't count (but your deaths do).
- Allow up to 3 lagouts, but you must return in under a minute.
- If the bot determines you are hiding 3 times, it will spec you and treat it the same as speccing for rec? Tho, we would need to determine what constitutes as hiding because sometimes you just can't find anyone.
- Winning an elim should bestow some sort of reward to your rating. Maybe add a hardset amount? Like 5 rating, maybe?
- It would be nice if deaths don't count if you die in less than 1 second after spawning... Spawning on bullets only takes away from skill. Though, if you manage to kill someone in less than a second after you spawn, it shouldn't count either.
- You may want to take the greens out of that arena, too. People like me (runners) tend to abuse the Full Charges when given the chance.

[milosh]

After the first day of testing I've had a lot of negative feedback about the new rating system. I'm trying to hear both sides of the argument, so I feel that perhaps the best solution is using the old rating calculation but adding a few extra features to it, such as a bonus for winning and a voided rating change for speccing before the game is over. Please give some feedback on this... I'd really rather hear from you before I code a bunch and test it in-game.

[MetalKid]

Tho rewarding people for playing a lot... I can see that, but for people who can't play all the time, there is no way they'll get to the top. This will reward people who just suicide and get like 12-10 every time.

[MetalKid]

Also, there should be a minimum number of games played before you rank. Otherwise, that could be abused? Or won't it matter because ppl who play more are going to get extremly high ratings?

[MetalKid]

You don't plan on showing aim % on the elim webpage?

[milosh]

I changed the zero mark to even(100% win ratio). Another week of elim beta testing has begun. Stats have been wiped to show the new rating system. Go to it my munchkins... go to it.

Here's the formula:

W = win ratio (kills/losses * 100%)
P = probability of winning(1/(1 + 10^((Avg Rating - Player Rating)/400)))
MAX_BONUS = currently 25

Rating change = (3W/players)(1-P)
Win Bonus = (MAX_BONUS)(players/30)(1-P)

[milosh]

Why would the zero mark be at 7-10 instead of 10-10? And is the objective of this new rating system to reward people for playing more since they will eventually get higher ratings by just playing? (for example it seems like if i go 10-10 200 times ill have a higher rating than someone that goes 20-10 20 times)

Well if the zero mark was changed to 10-10 this wouldn't be true. You would have to go positive those 200 times. But yes my intention was to reward players that play more often as that shows better consistency and activity.

[Izor]

I dont think you're ever going to prevent people from abusing the rating system no matter how much effort you put into it.

Why would the zero mark be at 7-10 instead of 10-10? And is the objective of this new rating system to reward people for playing more since they will eventually get higher ratings by just playing? (for example it seems like if i go 10-10 200 times ill have a higher rating than someone that goes 20-10 20 times)

You can get pretty complex and in-depth with this, but something simple will probably do since its only getting used for e-penis wagging competitions

[Lofty]

Yeah I don't quite get how it works exactly either

Here's the results of the formula I am proposing:

...

As you can see for this formula the zero mark is at a 70% win ratio. I can take that up to 80, 90, or even 100 if desirable though. I think 80 would make a happy median.

I don't understand the formula. I was expecting to see mathematics.