Sure. But I was addressing one specific issue raised.
…which is inherent in the Elo system, and not something Riot implemented. Very easy - you start at a specific amount of points. The winner gets more, the loser loses some. So, the starting point represents the (hypothetical) center of the distribution.
That being said, so far I played ca. 10 games ranked and was only matched up with guys of similar numbers. And when I started a new account in the US (while the EU server was borked) I also ended up playing with guys on my level, which only started to change after winning a few. All indicators that Riot takes the number of games into account - of course only a small number of games, so I might be wrong.
But I dispute the claim that someone with say 1500 games and an elo of 800 or so is better at this game than me with a total of 600ish and an elo of 1400. In other words - if his experience makes him so far better, he should have a higher Elo. People have very different capability regarding learning from experience, and different talent for this kind of activity. To presume that higher number of games = better player is a bit naive, IMO.
As stated above, even longish bouts of loosing at a level below your real ability can of course happen, but the more games you play, the more likely it becomes that the displayed number is a good representation of your actual play - which incidentally means that it is mainly the ‘old horses’ with 1000+ games that are rated correctly. Finally, Elo is a mirroring way to rate (mostly). So, for every guy vastly underrated there should be an accordingly vastly overrated player. Watching high Elo streams, I do see variety in ability. But not nearly to the extend that people claim happens at low Elo. The one thing that I would agree on is that as an experienced player unfairly rated below average, the effort to cross mid-range Elo would be more difficult than in a non-team game like chess.
Which is why these ‘how to get out of Elo hell’ advice posts in the official forum are kind of interesting. What do they describe ? Learn to carry, or learn to Jungle. In other words - learn to play a specific aspect a lot better…if you get better your Elo should rise. So, they are not teaching how to circumvent Elo ratings, they sneak advice to players to ‘l2p’ who think that they do not have to. Fun !
Frankly, Elo isn’t perfect (as stated above), and has flaws. And using it for a teamgame has drawbacks, sure. But many criticism raised against Riot using it are either based on a wrong understanding of the way it works, what the Elo number represents, or a flat-out denial that it might be somewhat fitting.
Well, if you show statistics I believe it. I do not see this effect, and know how treacherous subjective opinion is.
There is no indication I know of that Elo does work differently at a high level. Frankly, the top players will exhibit less movement simply because they train and play a lot, and are very consistent in their play. More likely explanation, IMO. Also, I have seen no data proving that for example the top quintile sees less movement than the second or fourth one. The mid will, by definition. Except possibly if -as I think- Riot takes new players out of that one.
Edit: Except if you talk of the utmost extreme, of course - but there the randomness should be less. Simply because there are only very few, very consistent player in there. Not good to compare a bracket containing an ‘around average’ group of maybe 1-2 million people with a group of 30 ? 50 ? consistently fantastic players, who train, who play a ton of games etc. Slight comparison problem, with numbers different by a few orders of magnitude.
TL;DR
a) I was responding to a specific claim about Elo, not claiming Elo is perfect.
b) I think Riot does weigh number of games in at the beginning, later the number should be irrelevant.
c) If experience means that you are a better player, your W/L ratio should go up, improving your Elo. If it didn’t, you aren’t.
d) Criticisms with the Elo-system as used by Riot to me mainly seem to come from a Dunning-Kruger effect, and are not based on actual data.
e) Of course Elo isn’t perfect. But it is far from useless.
Edit: To provide one criticism - the initial rating in ranked should take into account the normal game Elo to a small extend (maybe shift as much as a hundred in either direction ?). This will take a known metric about player skill into account, making the whole experience less frustrating for players new to ranked. Would also help with fast positioning.
