Thoughts on ELO being given based on a "good enough" metric?

Comment below rating threshold, click here to show it.

Zer0blizzard

Senior Member

02-19-2013

For example, what if you could still get LP in a game that you lost simply because you performed well enough for that game?

The most important idea that I'm suggesting in this thread is that ELO should be given based on a good enough metric; I don't know what that metric should be, considering I don't have meaningful data; however, ELO should be given to players on an individual basis so that if every player did as well as they did, they would have had a 51+% of victory.


Comment below rating threshold, click here to show it.

Zer0blizzard

Senior Member

03-01-2013

Waiting for responses.


Comment below rating threshold, click here to show it.

extreme person

Senior Member

03-01-2013

It just wouldn't work, if you're good enough to win more games than you lose, then you move up, if you aren't you don't. It's simple and it works, so many people get caught up in their score, but that doesn't tell anyone that you played cocky/missed objectives/threw the game. In other words, if kda meant everything why even try to win? and league inflation would be too high because it would be an uneven distribution. The fun thing about lp gains and losses is that 50% will gain and 50% will lose, why mess with those numbers?


Comment below rating threshold, click here to show it.

Zer0blizzard

Senior Member

03-06-2013

Quote:
Originally Posted by extreme person View Post
In other words, if kda meant everything why even try to win?
Way to ignore the fact that most games are won by teams with more kills+assists than the other team.


Comment below rating threshold, click here to show it.

Arcticfury

Senior Member

03-07-2013

Quote:
Originally Posted by Zer0blizzard View Post
Way to ignore the fact that most games are won by teams with more kills+assists than the other team.
Last night I lost a game where we were up 3 kills but got aced and lost due to minions when we respawned. KDA does not equal wins. I have gone 1/2/20 as Soraka and we blew them up, I've gone 12/1/10 as Sejuani Jungle and lost. Funny how my KDA didn't equal team win or loss in either. If you are truly better you'll win more than lose and move up. The ADC and Jungle shouldn't have an unfair advantage of gaining LP if they lose while the support is doomed to lose LP on every time. Why play support if you are doomed? As support why not take the kills so that you keep some LP even if it isn't best for the team?

This is a bad idea. Very bad.


Comment below rating threshold, click here to show it.

Zer0blizzard

Senior Member

03-07-2013

Quote:
Originally Posted by Arcticfury View Post
Last night I lost a game where we were up 3 kills but got aced and lost due to minions when we respawned. KDA does not equal wins. I have gone 1/2/20 as Soraka and we blew them up, I've gone 12/1/10 as Sejuani Jungle and lost. Funny how my KDA didn't equal team win or loss in either. If you are truly better you'll win more than lose and move up. The ADC and Jungle shouldn't have an unfair advantage of gaining LP if they lose while the support is doomed to lose LP on every time. Why play support if you are doomed? As support why not take the kills so that you keep some LP even if it isn't best for the team?

This is a bad idea. Very bad.
KDA = KILLS+DEATHS+ASSISTS

Quote:
Originally Posted by Arcticfury View Post
I have gone 1/2/20 as Soraka and we blew them up, I've gone 12/1/10 as Sejuani Jungle and lost. Funny how my KDA didn't equal team win or loss in either.

How about the kills and assists total FOR BOTH TEAMS? Oh wait, you wrote that point, but still said I was wrong. Either way, MOST games are decided by TEAM kills+assist totals.

Also, just b/c you posted ONE example does not refute the idea that MOST games follow that. While we're basing things blatantly off of the experience of single games, everyone should go Khazix, Skarner, Caitlyn, Gangplank, Swain, Akali, Katarina, Taric, Singed, Nunu. I've never lost the only games that I've played with those champions.

Also, never buy wards, for any lane, because I've won a game without buying wards as support. Never kill dragon, b/c I've won a game when we never killed it. Never even try Baron, b/c I've won a game without that. LOL, don't kill towers or get kills, b/c I've won a game just by farming, not getting the last hit on turrets and only getting assists.

Srsly?


Comment below rating threshold, click here to show it.

Arcticfury

Senior Member

03-07-2013

Quote:
Originally Posted by Zer0blizzard View Post
KDA = KILLS+DEATHS+ASSISTS




How about the kills and assists total FOR BOTH TEAMS? Oh wait, you wrote that point, but still said I was wrong. Either way, MOST games are decided by TEAM kills+assist totals.

Also, just b/c you posted ONE example does not refute the idea that MOST games follow that. While we're basing things blatantly off of the experience of single games, everyone should go Khazix, Skarner, Caitlyn, Gangplank, Swain, Akali, Katarina, Taric, Singed, Nunu. I've never lost the only games that I've played with those champions.

Also, never buy wards, for any lane, because I've won a game without buying wards as support. Never kill dragon, b/c I've won a game when we never killed it. Never even try Baron, b/c I've won a game without that. LOL, don't kill towers or get kills, b/c I've won a game just by farming, not getting the last hit on turrets and only getting assists.

Srsly?
Your original post is on the logic you just attempted to refute. Basically, if this one experience you just had ended in a loss, but you have a good score you should gain LP anyhow. If you end in a loss you lost and lose LP, very straight forward. If you had a bad KDA and received a win do you lose 2 LP because you didn't really contribute? What if the client crashes, that isn't your fault so do we take that into consideration? If you suicide 3 times in a row to help over turn team fights, giving your team 3 aces but now your KDA is a bit low, because we can't give Assists the same rating as kills since 5 v 1 ends in 4 assists and 1 kill, does that mean we have to tone down your LP on a win because our new metric says you're bad? What if in the same position you sacrifice yourself in a team fight and your team fails to engage over and over, now you're playing the same way as before but your KDA is horrible but it was your team's follow through that lost and not your contribution.

How does the program rate your contribution? Surely it won't be able to tell if you are playing the rabbit for your team to initiate a 5 on 3 versus you being a troll and feeding because your team won't surrender. You give me the list of logic you believe it should count and I can show experiences or current games in match history or in playoffs, finals, or anywhere else that would still result in someone getting the shaft because there isn't a human judge.

Anything you give won't work without a human doing the calculations. I've won many games with our team having lower kills than the opposing team but got that ACE at the right time. Does that mean that 3 of the other team's players gain 2 LP because they played well, perhaps better than all 5 of my team's? You can state it is all hypothetical, but I said give me the metrics, the logic for your system and I'll start showing you matches that it won't work in. It is frustrating that you lose when you outplayed everyone else, but sometimes it happens, a lot in team sports. I don't believe you can have this without a human judge or a premade 5 player team. If you could have this I'd love it, I'm not saying I don't want this, I'm saying your solution can't/won't happen because it doesn't answer the problem.


Comment below rating threshold, click here to show it.

Zer0blizzard

Senior Member

03-12-2013

The most important idea that I'm suggesting in this thread is that ELO should be given based on a good enough metric; I don't know what that metric should be, considering I don't have meaningful data; however, ELO should be given to players on an individual basis so that if every player did as well as they did, they would have had a 51+% of victory.

Human judges have issues:

  1. You need a lot of them if you want to a person to judge every game.
  2. They make errors; a lot of them if they are given a lot of freedom to judge and not enough accountability. Deciding on these rules is a giant fustercluck.
  3. The more money to be made from outcomes, the more likely they are to be prone to corruption without said accountability.
  4. AT THE VERY LEAST, THEY NEED A SET OF GUIDELINES OF WHAT IS CONSIDERED GOOD ENOUGH.
  5. See the above point; the guidelines/rules need to be established.

IMO, the list of rules/guidelines for what makes a good player is the entire point of ranked. However, there is no such official ruleset/guidelines. There are things that developers have said, strategies the pros use, etc., but it is never quantified with statistics (which is another thread that I've made; getting Riot to release user-searchable statistics of meaningful amounts; CertainlyT has said that LOLKing does not have good enough stats).

Here is the principle problem; if you looked at a list of infinite statistics for every game played within a season (including small things like total distance traveled in a game), you could basically establish a set of indicators that would guarantee a 51+% chance of victory in 99+% of games.

IMO, most games (51+%, dunno about 99+%) are decided by teams that have the most total gold AND the most total damage to champions at any time past 30 minutes.

I've also suggested that teams that have more kills+assists (specifically, at any time past 30 minutes) win 51+% of games.

Responses to your posts:

Quote:
Originally Posted by Arcticfury View Post
Basically, if this one experience you just had ended in a loss, but you have a good score you should gain LP anyhow.
I've had a ranked game where we accidentally forced a 5v5 at mid turret from 1 min to 13 min. Just because something happened once doesn't mean it's likely. People should make decisions on likely occurrences or things that have absolutely drastic consequences if they happen even once.


Quote:
Originally Posted by Arcticfury View Post
If you had a bad KDA and received a win do you lose 2 LP because you didn't really contribute?
The most likely scenarios for people that have bad KDAs are that they were AFK or they suck. Those people do not contribute enough. They should be punished.

Quote:
Originally Posted by Arcticfury View Post
What if the client crashes, that isn't your fault so do we take that into consideration?
Depends on the reason for the crash. If your ping sucks and you time out, that's your fault. You should not be playing this game. If your computer randomly fries, that's not your fault. However, I'm willing to bet most AFKs are user created; they could play the game but they won't.

Quote:
Originally Posted by Arcticfury View Post
If you suicide 3 times in a row to help over turn team fights, giving your team 3 aces but now your KDA is a bit low,
What you're describing is a situation where you die, do a lot of damage and then your team ACES the other team AFTER the countdown for assists has expired. That's so unlikely it's ridiculous. Even if you don't do a lot of damage, aces are quite rare when your team is down a person (ie, 4v5).

Quote:
Originally Posted by Arcticfury View Post
because we can't give Assists the same rating as kills since 5 v 1 ends in 4 assists and 1 kill, does that mean we have to tone down your LP on a win because our new metric says you're bad?
Again, games are not won this way. 51+% of games are won with 3-5v5 teamfights that result in 6+ deaths and 8+ assists per deathfight.

Quote:
Originally Posted by Arcticfury View Post
What if in the same position you sacrifice yourself in a team fight and your team fails to engage over and over, now you're playing the same way as before but your KDA is horrible but it was your team's follow through that lost and not your contribution.
Then my idea doesn't get in the way of anything. Under the current system, you'd most likely lose. Under my system, you'd most likely lose because you sucked and your team sucked.

Quote:
Originally Posted by Arcticfury View Post
Anything you give won't work without a human doing the calculations.
I'm assuming you don't mean mathematical calculations involving actual numbers; good luck finding a person that can add up all the sources of damage one champion has taken from all of his enemy champions over the course of an entire game with 99% accuracy IN REAL TIME. That's exactly why we play this game on the computer as compared to on a tabletop board game or something; it's too complicated for humans to comprehend at the same speed.

I'm assuming you meant something more akin to making a judgement call; something computers have extreme difficulty with without very good algorithms. LOL, at its core, is quite simple to win. Destroy the Nexus or the other team surrenders. The problem is getting to either of those paths likely.

Quote:
Originally Posted by Arcticfury View Post
I've won many games with our team having lower kills than the opposing team but got that ACE at the right time.
Many games don't mean squat when you're talking about millions of hours of playtime. Trends matter more than luck. For example, there was a Riot vs Pros game that's on YouTube. Riot made the pros go 5 supports; the pros won the game with consistency. That doesn't mean that the meta should be everyone go support.

Quote:
Originally Posted by Arcticfury View Post
You can state it is all hypothetical, but I said give me the metrics, the logic for your system and I'll start showing you matches that it won't work in. It is frustrating that you lose when you outplayed everyone else, but sometimes it happens, a lot in team sports.
Correlation does not imply causation. Stat 101 lesson done.