Boards›Gameplay

Ranking System fixes

Chainman3·9/18/2019, 5:12:30 PM·1 votes·1,458 views

I wrote this in a reply to a post on the boards but i thought i might as well post it as a discussion to see what others think.

My solution was to use a rank system based on merit not based on win. Because your rank represents YOUR skill level not the skill level of your team. Now multiple stats would be taken into consideration in respect to that game. For example, top 5 dmg dealers in the game earn +5 LP. top 5 vision score +5 LP. And to offset this worse 5 dmg dealers -5LP. Now players would argue, but then i can never have highest dmg as a full tank. But as a full tank they should always have highest dmg taken and that would be a stat too. Basically the stats would look something like this: 1.Dmg 2.Vision Score 3.Dmg taken 4.Dmg to objectives 5.shielded/healed amount 6.KDA 7. Win/Loss (not regarded as a category)

How the system would work. Since current we get +100 LP (20 to each member for 5 members im going to assume on average) we can divide those LP over the 6 categories evenly 100/6 = for 17 LP per category. Now since the best 5 members get LP for each category we can divide that further 17/5 = 3.4 LP so LP would be rewarded as followed: Top 5LP 2nd 4LP 3rd 3LP 4th 2LP 5th 1LP 6th -1 LP 7th -2LP 8th -3LP 9th -4LP 10th -5LP

For a win the winning team gets +5 LP each and the losing team gets -5LP each, making the outcome of the game still a factor.

Which means if you can essentially lead all cateogires you can possibly gain +30 LP. Now this most likely will never happen because certain stats are better suited to certain roles. Like vision score and support or dmg dealt and carries. This would do 2 things for the ranked system, 1 it would push players into filling roles for their team (not like top bot mid roles but if the team lacks a tank role because it would guarantee a player +3LP assuming they play their tank correctly and tank more than all characters who arent tanks in the game). 2. It would force players to not only compete to win but to compete to do better than their own teammates. Basically rewarding players who play consistently well or consistently carry, while also discouraging trolling and afk farming trolls.

So lets calculate a game where everyone does their part. Carries: dmg dealt, dmg to objectives Supports: Heal/shield, vision score. Tanks: Dmg taken, Healed, Dmg to objectives. KDA and Win can be applied to any role.

Overall that leaves you with a possible 10 to each role exclusively when played correctly and another possible 10 for the win and KDA of the player. Meaning a player who carries successfully can get +20 LP still while a player who carries but their team still loses will get +10 LP. Now when you look at something like brand support it can get the dmg spot but it opens up the heal and shield spot.

The point of the system would be to reward consistent play while shielding the players LP from soft inters. Basically making it if someone is going to troll they have to do it obviously in which case Riot can then implement a system where you can do player reviews with a high accuracy on reports or change our current report system in a different way to detect hard inting.

Notes: It doesnt need to follow this +5 LP to -5LP system you could have it range to the top 3 only where 4th and 5th get -1LP each and instead the first 3 roles get +7LP, +5LP, +4LP.

32 Comments

Saezio9/18/2019, 5:17:06 PM4 votes

some champions can't deal damage even if they hard carried their team.

Other champions deal tons of damage because they are poke champions.

This is a team game.

If you implemented personal scores then everyone plays for the scores and not for the win.

What I would be willing to consider

Winning team gets 100 LP to allocate across 5 players which is allocated based on their grades

Losing team loses 100 LP and it's decreased again based on grades.

Posui Gart9/18/2019, 7:59:28 PM2 votes

1.Dmg 2.Vision Score 3.Dmg taken 4.Dmg to objectives 5.shielded/healed amount 6.KDA

Dmg: Everyone builds as much damage as they can on everyone. You are playing jungle Ivern? Well, go grab that runic echoes and spam your damage or -10 lp for you Vision score: Lets put a pinkward in some whatever place nobody ever goes to. What, putting wards fro an actual vision? Nah, enemy will clear it, getting +visionscore while you get nothing, enjoy your -lp Damage taken: Time to go tank enemy towers while nobody is around, or, again, -lp. Also this rewards 0-20 running-it-down-mid feeders while punishing those who actually played safe, dodged shit and didn't take much damage Damage to objectives: Ok lets pick something ranged and spam skillshots at dragon/nashor in the meantime, or -lp Healed/shielded amount: ok so im support leona, i guess -lp for me KDA: oh im 3-0, should i participate in teamfights? Nah, my kda can be ruined, better to afk farm and not risk losing my lp

Every stat can be abused in some way. The only fair way to calculate the contribution would be to have a group of pro players and analytics to analyze everything that happened in the game, and nobody is going to do it for all ranked games. Even in Worlds, the wictory goes to the team as a whole.

Unker1399/18/2019, 5:34:08 PM1 votes

{quoted}

I wrote this in a reply to a post on the boards but i thought i might as well post it as a discussion to see what others think.

My solution was to use a rank system based on merit not based on win. Because your rank represents YOUR skill level not the skill level of your team. Now multiple stats would be taken into consideration in respect to that game. For example, top 5 dmg dealers in the game earn +5 LP. top 5 vision score +5 LP. And to offset this worse 5 dmg dealers -5LP. Now players would argue, but then i can never have highest dmg as a full tank. But as a full tank they should always have highest dmg taken and that would be a stat too. Basically the stats would look something like this: 1.Dmg 2.Vision Score 3.Dmg taken 4.Dmg to objectives 5.shielded/healed amount 6.KDA 7. Win/Loss (not regarded as a category)

How the system would work. Since current we get +100 LP (20 to each member for 5 members im going to assume on average) we can divide those LP over the 6 categories evenly 100/6 = for 17 LP per category. Now since the best 5 members get LP for each category we can divide that further 17/5 = 3.4 LP so LP would be rewarded as followed: Top 5LP 2nd 4LP 3rd 3LP 4th 2LP 5th 1LP 6th -1 LP 7th -2LP 8th -3LP 9th -4LP 10th -5LP

For a win the winning team gets +5 LP each and the losing team gets -5LP each, making the outcome of the game still a factor.

Which means if you can essentially lead all cateogires you can possibly gain +30 LP. Now this most likely will never happen because certain stats are better suited to certain roles. Like vision score and support or dmg dealt and carries. This would do 2 things for the ranked system, 1 it would push players into filling roles for their team (not like top bot mid roles but if the team lacks a tank role because it would guarantee a player +3LP assuming they play their tank correctly and tank more than all characters who arent tanks in the game). 2. It would force players to not only compete to win but to compete to do better than their own teammates. Basically rewarding players who play consistently well or consistently carry, while also discouraging trolling and afk farming trolls.

So lets calculate a game where everyone does their part. Carries: dmg dealt, dmg to objectives Supports: Heal/shield, vision score. Tanks: Dmg taken, Healed, Dmg to objectives. KDA and Win can be applied to any role.

Overall that leaves you with a possible 10 to each role exclusively when played correctly and another possible 10 for the win and KDA of the player. Meaning a player who carries successfully can get +20 LP still while a player who carries but their team still loses will get +10 LP. Now when you look at something like brand support it can get the dmg spot but it opens up the heal and shield spot.

The point of the system would be to reward consistent play while shielding the players LP from soft inters. Basically making it if someone is going to troll they have to do it obviously in which case Riot can then implement a system where you can do player reviews with a high accuracy on reports or change our current report system in a different way to detect hard inting.

Notes: It doesnt need to follow this +5 LP to -5LP system you could have it range to the top 3 only where 4th and 5th get -1LP each and instead the first 3 roles get +7LP, +5LP, +4LP.

Here is the problem. As a support, it is expected that I will have high vision and low damage. If I have the best vision and the worst damage, it's a wash and that's fine....

However if I simply focus more damage and less cc, i might have highest vision and 2nd lowest damage so i benefit...and if I can steal a kill, my kda benefits at the expense of my teammates. In fact, hang back in team fight, last hit anyone that gets low and run becomes a viable strategy for climbing. Is that the definition of a good player?

Kai Guy9/18/2019, 5:47:55 PM1 votes

I have read some publications on trueskill 2, which uses personal stats as a relevant metric to a MMR system.

The issue is that skill gaps generate skewed stats and the folks who work on MMR systems have found that basically every stat you use that can be controlled by players becomes basically useless the moment its made public. The more player control over it, the less value it has as a metric of skill. For example if healing = better skill it encourages folks to take damage then self heal.

Overall the constant needs of a game vary on a game to game. Objective statistical systems are difficult with out tacking them onto a MMR system. Subjective Stat systems that go game to game are very difficult to build and automate at any quality.

Its a bit ironic but you'd actually have a better system if you don't mention that you added stats/ are vague on what your system values.

For a Moba, it would be really important to account for champions. I would say that Role is irrelevant, but its not. Issue is multiple classes fill the roles in the game. A comp can overlap classes, but that lowers the values of each. Class interactions also skew values as do matchups.

The # of Variables to generate stats with any significance is insane to attempt to caculate.

Issues I see your system has to address.

A. If you use class benchmarks, its not fair to all champions in a class due to kits. B. If you instead use individual champions, you have to pick between using a Curve built from players, or Arbitrary values. Curve will change literally daily, popularity impacts easy of stats. Arbitrary values are datamined pretty fast and players can play to nail those benchmarks. C. People will write off games and stat pad to protect their raitings at the cost of teammates. Some People literally do this to protect their KDA with it providing 0 gains, they just want to keep good stats to brag about their champion skill and then blame teamamtes for negative WR impact. D. Uncertainty on accounts, you don't want to treat a 2000 game or a 2 game player at the same uncertainty. E. Matchup's and Synergy. Needs to be accounted for cuz it will heavily impact stats. F. Skill generates stat gaps, but stat gaps are not necessary a great measurement of skill. MMR and Elo using averages that naturally occur in a population have equal and identical rules for all players.

usul12029/18/2019, 6:39:54 PM1 votes

If you do a billion damage but it doesn't help you win games, why should you be rewarded for it?

Shahamut9/18/2019, 10:03:57 PM1 votes

We already have a performance rating which is a comparison of your play to others with the same champion.

You could go a long way by just adding a bonus +LP based on your rating that game. Wouldn't need to be much, but it would go a long way if instead of losing 17 LP on my S+ games I only lost 12.

S+ - +5 S - +4 S- - +3 A+ - +2 A - +1 A- - +0 B+ +0 B. +0 B- +0 C+ -1 C. -2 C- -3 D+ -4 D. -5