Random260's Forum
Sign up for greater features. Remember to donate to us to keep everything running! This is the link: http://random260.forumotion.com/h1-donate-here
Please donate xD

This is the official site of Random260
HomePortalCalendarFAQMemberlistUsergroupsRegisterLog inSearch

Share | 

 Solving the ELO Conundrum, Killing Alts, and Fixing Chiff

View previous topic View next topic Go down 

Posts : 60
Points : 84
Join date : 2013-05-11

PostSubject: Solving the ELO Conundrum, Killing Alts, and Fixing Chiff   Mon Jan 13, 2014 11:35 am

Note: this post is written by Cyray but includes ideas collected from F00ty_R3b0rn, Soundwaves, and Acccreate.
Acccreate and Soundwaves approved of my ideas but I don't know about F00ty_R3b0rn.

The impetus for this post was the discussion of ELO on the chf.ucoz.com forums (here), and I am writing in response to Bond009:

Bond009 wrote:
"Stacking is actually bad for ELO, it is based on a teams over all elo score. so if your team has someone with 1200 and 1400 and 1300 and you vs teams with less than 1000 you will gain very little elo when you win becuz your team is stacked, as where if you lose the low elo team will gain more elo from you. That concept is pretty balanced.. but ppl playing for wins and losses % ratio, thats where the incentive for stacking comes into play. Its really hard to think of how to make it better or 'fix' it."

This poorly constructed argument is full of inaccuracies and represents a crucial misunderstanding of the ELO system. Allow me to break down the ELO system and illustrate its true failings.

Problems with ELO

1. Those who are at the top of the ELO ladder, the ones who really care about their ELO have made alt accounts with very poor win/loss records and low ELO. On these accounts they intentionally lose games to bring their ELO down. They then give this account to one of their friends, go on their main account, and play together so that their high ELO account appears to be partnered with a complete beginner (someone of low elo). Then, they pick black arrow, rush pubs, mass wins, and their ELO shoots up. Examples of players who abuse using this method are [APF]-Walter, [APF]-Cesar, Adam_CZ, [FTL]-Roystom and others. While I have never abused in this way, I will admit that I often have had xMoNsteRRR, Cheesy-Gordia, and J.ST4X play on fresh 1,000 ELO alt accounts with me and it makes my ELO increase significantly quicker since the bot thinks I am soloing with beginners on my team.
Also, it is very difficult to increase ELO now, in general. When the bot first came out it was easy to get your ELO up since most opponents had around 1,000 ELO (since ELO was new). However, now, most players have very low ELOs below 500, so when I beat them my ELO only increases (>1 point per game). You will notice that most of the top player's ELOs have stagnated recently; mine is stuck around 1800, Walter's shot up to 2400 but now it won't move any more, etc. You might respond by telling me that I should play against higher ELO players to improve my ranking. However this method is extremely slow still and is highly risky, so most players won't do it. For example: if I'm in the lobby and Cou, [APF]-Walter, and [FTL]-Roystom join, chances are I would leave the lobby. This is because if I manage to get lucky and solo them, my ELO might go up 3-5 points. But if I lose to them it will drop 20-40 points. It's not worth the risk. Players like F00ty_R3b0rn, despite being at the pro level, cannot increase their ELO. F00ty_R3b0rn often plays difficult games and solos against full stacked teams, but his ELO doesn't increase because he simply has too many losses (and for the aforementioned reasons as well). And even when you do win games, e.g. by soloing high elo players on the front page, your ELO will only go up ~2-5 points. When you beat pubs it goes up >1 point. This leaves players with only one option to increase their ELO: make smurf accounts, drop their ELO, and use them to boost one's main account by massing games a la tech rush.
(Evidence that losing a game to pubs overly-punishes punishes players and winning against pros under-rewards them here)

2. ELO doesn't work properly in CHF because CHF is an FFA gametype. Teams are either made randomly or by stacking. If you stack you will have a better rank even if you earn less ELO per game, as evidenced by the fact that the top ten players are APF, TAS, TDT, and FTL players known to stack. All the top ELO players stack to some degree. And as mentioned previously, in the long run your ELO will benefit more if you mass games and keep beating pubs as opposed to playing more difficult games. So if you want a high ELO the best strategy is to just keep beating pubs even if your ELO goes up slightly slower, because in the long run its better than risking games against good players.

3. Another fact is that for a pub player, each game his or her teammates will change. Admins have a significant advantage because they can stack their team, ban players, abuse swap, etc. The playing field is unlevel because players are playing under different conditions. There are too many random factors (In lobby) and too much room for luck in CHF. ELO works better in games like DOTA because there are only 2 teams and very few admins in proportion to the overall player base; therefore, even if a few admins abused swap or other commands, it wouldn't affect a large proportion of the games.
Admin stacking is also a problem because it discourages newer players from continuing to play CHF. New players have to deal with 3v1s constantly, and as we all know, getting rushed by a full stacked team is enough to make any man want to smash their keyboard and rage quit off battle.net. New player's heroes get sniped, they get wrecked in mass, they lose their bases too quickly and can't explore new builds or strategies. This happens perpetually on ClanHPMM bot because this bot has so many admins.

The Solution

F00ty proposed to !sp in the lobby of every game but that won't work because people want to play with their friends. However, if you do introduce the !sp command it should make the teams as even as possible according to players' ELOs. Otherwise these teams wouldn't necessarily be balanced; they would just be random.

This is the tried and true method that DotaCash uses and so it's been proven to work.
(example here http://www.dotacash.com/home/ )

You need to have ClanHPMM run two bots: one for beginner players (like the current bot), and one that requires at least 25 games played and 1000+ ELO (so 0-0 alt accounts can't join) for more skilled players.
These two bots would each have their own ladder. They would be called ClanHPMM Tier 1 Bot and Tier 2 Bot. Tier 2 bot will be just for pubs to relax and play a beginner level game. Tier 1 bot will be for higher skilled players to fight it out to determine who is truly the best CHF player/team. New players would start out playing on the Tier 2 bot until their ELO reached at least 1000 with 25 games under their belt. Then they would be granted access to the Tier 1 bot, but they could still play on the Tier 2 bot as well if they wanted to. The Tier 2 bot would also have an ELO cap of 1300 (the Tier 1 bot would have no ELO cap).

To get into Tier 1, a player would have to get their ELO above 1000. This would be a fun challenge for players because it would give a certain level of prestige to people in the top tier. Players would play more and fight hard to get into the more prestigious league. They could go to Playchf.tk and check their tier 2 stats to see how close they were to 1000 and see which other players were about to move to the higher tier.
Also, by putting an ELO cap on the Tier 2 bot at 1300, players would have little incentive to keep playing on it and they would be pressured to come to the new bot. This will help the new bot fill faster, which is good because getting the bot to fill might be a challenge. But I think it's a challenge worth tackling.
This is not a perfect solution, but it would solve many of major problems currently associated with your game and ELO system, for example:

1. The bot fills too fast right now and its impossible to join. If you have two bots it will take games perhaps 3-5 minutes to fill which is ideal (this is about the same amount of time that takes to find a game in league of legends, dota 2, etc). Right now ClanHPMM bot games fill in under 30 seconds which makes it impossible to play with friends or even alone unless you have admin. Even though the new bot will have a smaller player base (~750 players according to calculations on 1-13-2014), since these are more experienced players I think they would be willing to wait a bit longer in order to play a higher skilled, inhouse-esque match. I know I would, and I've heard the same sentiment expressed by many of my pro level friends.

2. Pros are discouraging new players from continuing to play the game because the new players just get stomped in 5 minutes. With a two tiered system, this problem would solve itself.

3. Games are too easy for pro players, or even players above pub level (like people in Clan HPMM). I can't sign on battle.net and have an enjoyable game because the competition level is so low. Us pros have to organize our own games these days and its hard since lots of players play on private servers, US East, West, Europe, and Asia, and we can't easily communicate with them in chat. Having two bots would be a fantastic, automated solution to this issue.

4. ELO doesn't show who the best players are. Right now [APF]-Walter, Vetealamierda, [APF]-Cesar, Terminator, and Nexsus are at the top of the ELO ladder. But I think we can all agree they are not the best players; all they do is make smurf accounts with negative ELO to boost their main accounts, mass games with tech rush, and abuse the broken ELO system. The best players are Soundwaves, Cou, Keep3r, F00ty_R3b0rn, and J.ST4X, but these players don't have the highest ELO because they don't mass games or use cheap tactics to boost. Keep3r isn't even on the front page because he started playing again after ELO was already around for a few months, so he couldn't enjoy getting easy ELO when all the pubs were still around 1,000.

5. This will solve the problem of alts since only accounts with 25 games and over 1000+ ELO can join. Cou and Random260 have spent a lot of time having to track down alts trying to snipe people, and so getting rid of most alts would give them one less thing to worry about. It would also improve the ELO system as a whole.

Thanks for taking the time to read my thoughts.
Back to top Go down
View user profile

Posts : 267
Points : 300
Join date : 2013-07-01
Age : 103
Location : Croatia

PostSubject: Re: Solving the ELO Conundrum, Killing Alts, and Fixing Chiff   Tue Jan 14, 2014 2:40 am

Random260 said when he came CHF 4.1 comes the elo evolution 2.0 elo...But 4.1 did not come ...
Thank you for the hard work,i think random260 and cou something used by your ideas here.  Very Happy 

If you need me,send friend request Garena my name : ZiZi-____-ZiZi
Back to top Go down
View user profile http://clan.intl.garena.com/clan/profile/Footies/

Posts : 20
Points : 36
Join date : 2013-10-06
Age : 48
Location : Washington State

PostSubject: TY   Tue Jan 14, 2014 1:57 pm

TY for the input. I have played a few games with some of the "good" players and could not understand why the "newbies" on their team were so good. I was hoping to boost my ELO, yet i lost due to all of them being good. I don't care that much about my record, but do not like the stacking. I also enjoy a good game of a well balanced match. I don't like a slaughter, it is no fun. It will be interesting what comes of all this.
Back to top Go down
View user profile
Sponsored content

PostSubject: Re: Solving the ELO Conundrum, Killing Alts, and Fixing Chiff   

Back to top Go down
Solving the ELO Conundrum, Killing Alts, and Fixing Chiff
View previous topic View next topic Back to top 
Page 1 of 1
 Similar topics
» Check out this new Band
» Lyrics for the Weirdness?
» Quotes on humor and truth
» SCANDAL - Kill The Virgin

Permissions in this forum:You cannot reply to topics in this forum
Random260's Forum :: Footmen Hero Devolpment :: Idea Factory-
Jump to: