Throughout the years of League of Legends the players have been complaining about how unfair climbing the latter in ranked is. However, quite seldom does someone give a possible solution to making the ELO system "fair". In this post I am going to explain a way that it can be done that will account for the following: champion win-rate, an accurate representation of player skill, promoting champion pool diversity, and non-statistical categorizing. TL:DR is at the end of this post.
The new system will be a construct of the old in terms of the rankings' values and matching through MMR, but the key difference is that Riot would hold record of the win rates of each champion in each ELO and give +/- LP with accommodation for these win rates. To expand, if someone was to play Azir in a silver 2 game and won, their LP gain would be larger than if they had won with an Ahri due to the win-rate gap between those two champions within that silver 2 ELO. The key to this working is that it awards based off of skill and knowledge rather than playing the "OP" champions simply because they are the META.
**Disclaimer:** This change is not intended to make drastic changes to the ranking system. It would simply be a push in one direction or the other. No more than 3-5 LP per match.
Going back to the thesis, accounting for the champion win-rate is a major factor in this. By making sure that any champions with a clear advantage over most match ups get punished ultimately by the ranking system while still giving award for victory. This also works the other way too. Losing with the champions with a high win-rate will yield a larger loss if defeated. Additionally, it goes vise-versa for those with low win-rates. Keeping people's rankings at bay simply because their favorite champions become stronger/weaker would make for a much more enjoyable experience.
"An accurate representation of player skill" is what the community has been begging/complaining for since the community started. This system would achieve this by making it so that if you are better at the game than your counterparts, you will rise regardless of what role or champion you play. This is a HUGE claim and I believe it is one that is supported because this system will always average out to making a player rise/fall in ELO over enough games. Blatantly, it will not have a huge effect on those who do not play a lot of ranked games, but for those that complain about getting stuck at a certain point which they do not belong (*cough* *cough* ELO Hell), there would be a way out.... a motive!
My favorite affect that this would have is the promotion of champion pool diversity. Since there is larger penalty to losing with a "stronger" champion, players will be dissuaded from using the same champions in most situations. Instead of picking a Yasuo into a Zed because you are great with him, you could pick Annie and global him any time he tries funny business for an easier win. Expected value is going to be much greater when someone can pull out a sleeper in the right composition.
Non-statistical categorization means that you will not be overcompensated for the grading system; the grade you get from a game is never going to be accurate enough to give large sways to awarded LP (IMO obviously). This is due to the fundamental nature of grading players at all: compositions change, unaccounted participation (such as warding) exists, and the numbers are inaccurate from things like KS'ing and unnecessary shielding from a support. Instead, going off of win-rate will give a much better _average_ grading of skill.
Thank you for reading through my explanation and feel free to comment anything in response to this: I would love to hear what others think.
**TL:DR:** Handing out more LP to those who play "weaker" champions and less LP to those who play "stronger" champions would assist in making a player's ELO an accurate representation of their skill.