Mobafire League of Legends Build Guides Mobafire League of Legends Build Guides
's Forum Avatar

Potentiall silly idea

Creator: JEFFY40HANDS August 5, 2012 7:40pm
1 2
JEFFY40HANDS
<Altruistic Artist>
JEFFY40HANDS's Forum Avatar
Show more awards
Posts:
9270
Joined:
Aug 10th, 2010
Permalink | Quote | PM | +Rep | Commend August 5, 2012 7:40pm | Report
I was just thinking; what if elo was determined by 3 game averages opposed to per game basis.

In principal, you'd have 3 games averaged to provide your elo instead of 1. This would of course mean your gains/loss would be potentially higher. Im sure it sounds like a lot more of a risk in terms of climbing the rankings, but it allows for a bit more compensation. What i mean is, assume you only won 1 out of 3 games, instead of gaining 10 elo for the win, but losing 30 (net loss of 20 elo) you gain 5 elo at the end of your 3rd game. The reward is smaller but you arent as punished now by "trolls" or "unskilled players". If you won 0 of 3 games your elo loss may be around 20 or 30 elo. But winning all 3 might equalize the losses by giving you 25 or 30 elo.

Mathmatically you'd be playing more games to aquire a near identical gain, but the idea is that youre less punished for losses as long as you aquire at least 1 win.

The benefits i see of this system are as follows:
Less punishment for losses
Encentive to play more games
Potentially less pressure
Slightly improved balance

The negative affects:
More work for slightly better or equal gain.
Potential to create laziness (meaning you got one win so why try hard in your next 2 games)
Lessens the excitement of each win.

Im sure there are more possitives/negatives that im overlooking, but then again this is just a shot in the dark suggestion.

The reason i thought of this was boredom. I wanted to discover a way to potentially make the elo system less haphazard and more player friendly. The current elo system focuses on the idea that all 5 players on either team all intend to win. Ive stated before that the system works for organized teams, who obviously intend to win as much as possible. On an individual player basis however, the goal/drive might not be there. That said this system isnt perfect, it means all players c,ould hypothetically reach 2k and up....but this is just an idea, nothing id expect to ever be implemented, just a bored idea i guess.
Arkharan
<Member>
Arkharan's Forum Avatar
Show more awards
Posts:
540
Joined:
May 4th, 2011
Permalink | Quote | PM | +Rep | Commend August 5, 2012 7:53pm | Report
That's silly because that came of you.
If I hide your name tho the idea become good! :P

ty xaiolu for the sign :3
DuskToGlory
<Member>
DuskToGlory's Forum Avatar
Posts:
329
Joined:
Mar 21st, 2012
Permalink | Quote | PM | +Rep | Commend August 5, 2012 7:54pm | Report
It certainly is an interesting idea, not sure how I feel about it though.

Thanks to DuskToGlory for the sigs. LOL :)

Working on a Pantheon guide, will finish when I get around to it, could be never.
+Rep? I help you? you help me, no?
BluAnimal
<Member>
BluAnimal's Forum Avatar
Posts:
509
Joined:
Jun 15th, 2011
Permalink | Quote | PM | +Rep | Commend August 5, 2012 8:02pm | Report
It seems like a lot more work, but in the end it would accurately represent your skill. (I am using the term skill very loosely)

+ Rep cause it's an intelligent post :D

The only problem is wouldn't you also suffer more, over-all, if you hit a bad streak or were paired with trolls 3 games in a row?
JEFFY40HANDS
<Altruistic Artist>
JEFFY40HANDS's Forum Avatar
Show more awards
Posts:
9270
Joined:
Aug 10th, 2010
Permalink | Quote | PM | +Rep | Commend August 5, 2012 8:11pm | Report
I could see players who are already at 1600 and rising not encouraging the system. The fact is that there is no "elo hell" persay, more along the lines of "elo limbo". Players hitting the high end of one bracket and the low end of another ie 1570 and 1620 the same at lower or higher brackets. These areas are typical plateus. Eventually the players will move on if theyre meant to or sink back down, my suggestion destroys these plateus thereby making it likely more players will progress...this would hypotetically flood all elos with players who might not be on "that level"....people can feel free to scold the idea, just figured general discussion needed something worthwile to...well discuss.
uPWarrior
<Member>
uPWarrior's Forum Avatar
Posts:
1
Joined:
Apr 9th, 2012
Permalink | Quote | PM | +Rep | Commend August 6, 2012 4:23am | Report
I'm surprised nobody mentioned this, but it can't work because it would lead to rating inflation.

Ratings are not supposed to go up per se, they are supposed to match players fairly. You can't lose more games than you win and still manage to gain elo, that would only make players with more games to be ranked higher.
Wayne3100
<Community Manager>
Wayne3100's Forum Avatar
Show more awards
Posts:
7148
Joined:
Aug 3rd, 2011
Permalink | Quote | PM | +Rep | Commend August 6, 2012 4:29am | Report
Why would someone gain elo if they only won 1 out of 3 games? :|

Thanks to MissMaw for the signature!
MrCuddowls
<Member>
MrCuddowls's Forum Avatar
Show more awards
Posts:
1733
Joined:
May 31st, 2012
Permalink | Quote | PM | +Rep | Commend August 6, 2012 5:29am | Report
Can you explain what exactly your Averaging
JEFFY40HANDS
<Altruistic Artist>
JEFFY40HANDS's Forum Avatar
Show more awards
Posts:
9270
Joined:
Aug 10th, 2010
Permalink | Quote | PM | +Rep | Commend August 6, 2012 9:15am | Report
uPWarrior wrote:

I'm surprised nobody mentioned this, but it can't work because it would lead to rating inflation.

Ratings are not supposed to go up per se, they are supposed to match players fairly. You can't lose more games than you win and still manage to gain elo, that would only make players with more games to be ranked higher.


I did address "inflation"

I stated this system could be problematic because the fact that it rewards a "negative" percentage, 33 percent not being over half, which by most standards is "subpar". However the way the system works now is based on a per game win or loss. But more often than not a loss is more detremental than a win. Your net gain after a one win one loss is somehwere around -5 if im not mistaken. So it requires 1.5 wins to break even. I also stated the current system is based on the idea that all players in a soloqueue match intend to win. That is not always the case (more often than not it is). But factor in dcs and ragequits and any other set of anomalys noncondusive to the "they all want to win" hypothesis and youre looking at a relatively unbalnced system. By spreading the gain or loss of elo over 3 games there is a blanket with which all players are somewhat covered. It takes into account additional external factors not currently adressed with the current system.

@wayne: the net gain for one win with my suggested system is negligible. The idea isnt perfect, but basically i figured 5 elo gain for a 33.3 w/l after a set was small enough that a player would have to repeat the same percentage hundreds of times to climb the ladder effectively. Therby creating incentive to win as often as possible to expediate the process...

I know the idea is relatively weak as a whole. This was simply my attempt to create discussion amongst players at all levels as to how they view the current system. As i stated this was by no means an idea i thought merrited an official attempt to make reality. It is simply thinking aloud.
lifebaka
<Scout>
lifebaka's Forum Avatar
Show more awards
Posts:
1126
Joined:
Dec 12th, 2011
Permalink | Quote | PM | +Rep | Commend August 6, 2012 11:45am | Report

Your net gain after a one win one loss is somehwere around -5 if im not mistaken.

If I win one game and lose one game, I don't have net ELO changes of greater than 2. If your ELO can fluctuate down by 5 after a win and a loss, you likely don't have many ranked games and the system is still trying to place you, at least somewhat. (However, in this case, if you win a game then lose a game you should have a net ELO gain, not loss.)

Once a player is well outside of his (or her) initial seeding, ELO change per match is usually 10 to 12, depending on whether or not the system thinks that player should win the game. So for one loss and two wins in a set of three matches, net ELO change should be between -8 and -14. By changing this to +5, you introduce an extra 13 to 19 ELO to the system in this scenario. Over several dozen games, this will almost undoubtedly lead to any given player's ELO rising, as it is difficult to lose three games in a row (I believe that I have only done it twice in Ranked, both part of the same losing spree).

Now, a little more generally, I don't think it's worth anyone's time to try to "fix" the ELO system. Frankly, the major thing that's wrong about the ELO system is that it is trying to model skill in an extremely complex game in a fairly simple way. Therefore, there is no simple fix to the ELO system; there are no simple fixes when the problem is that the system itself is too simple.

EDIT: I decided to model this a little more mathematically to demonstrate how it would easily inflate every player's ELO. Let's suppose that a given player has a 50% chance of winning any ranked game they play. Every 3 games, their ELO changes. If they lose all three, they lose 33 ELO; if they win one and lose two, they gain 5 ELO; if they lose one and win two, they gain 11 ELO; if they win all three, they gain 33 ELO. (Numbers chosen for exactly 11 ELO change per game except for the one win two lose case.) Their expected ELO change for every three games, then, is (.125 * -33) + (5 * .125 * 3) + (11 * .125 * 3) + (.125 * 33) = 6; therefore they should expect to gain 2 ELO, on average, for every game they play.

Now suppose this player starts at 200 ELO, because they really, really suck. They should still, after 1000 games, expect to be ranked at 2200 ELO. It'd be unreasonable to expect them to be below 2k ELO. This, really, is what's wrong with allowing ELO gain for the one-win-two-loss case.
OTGBionicArm wrote: Armored wimminz = badass.

My posts may be long. If this bothers you, don't read them.
1 2

You need to log in before commenting.