Current rating system can't accurately place lowest performing players

[Zojha]

The current rating system cannot accurately place ANYONE because there's too few people playing >_>
Rating only reflects your standing compared to people you played with/against.

To give an extreme example - if only a top team and a mid-class team queue up at the same time, they'll be matched against each other and the top team will gain rating and the mid-class team lose it. The mid-class team might be able to beat all bottom class teams in the same vein, but those never match against each other, because when the mid-class team plays, only the top class team is consistently available. This results in their rating being much lower than it should be. Likewise, the top team's rating will be inflated.

Similar holds for solo queue - if you keep playing with/against the same people, your rating will be less accurate than if you play with/against completely different people all the time, as you have less points of comparison. Which is a big issue with how small the playerbase is.

Honestly? I find the very idea of a rating system pointless under the given circumstances.

[Aviars]

The current rating system cannot accurately place ANYONE because there's too few people playing >_>

That isn't true. Can you honestly say that someone who has a 5% win rate can be weighted pretty much equally to someone with maybe a 35% win rate in bronze? Can you honestly say that when someone who loses so much yet doesn't drop in rating isn't being placed where they should be? In any other game most matchmaking systems try and place people in a way that nets them a 50% win rate, where their rating stabilizes with their wins and losses. In this game that isn't the case. Someone can lose 100 games in a row in unranked and the way the matchmaking places them does not change simply because they can't lose rating.

A good example of you being wrong is World of Warcraft arena pvp. Everyone there starts at 1500 rating. Their PVP community is probably 100 times larger than ours. But you will still find people managing to get lower than even 1000 or 500 rating. Why is that? If what you say is true then nobody should be lower than 1500 because the WoW community is so huge right?

Anyway it is just common sense that if someone loses that they should have matchmaking somehow compensate that, but since they have a charity window that can cover so many different skill levels, it literally makes it luck on what type of low level player you get, which isn't fair.

If I remember correctly when Feast initially came out it used to be something like multiples of 500 for each rank. As in 500 for bronze, 1000 for silver, 1500 for gold, 2000 for platinum and 2500 for diamond. People then complained about that saying that an abnormally large majority of the population was stuck at bronze/silver, so I don't see your second option being very popular nor lasting unfortunately. I can't remember clearly but I don't think it had that thresh hold preventing loss of points as unranked either initially. Speaking of unranked, these people you claim have been unranked since the season started clearly do not PVP so they don't matter in regards to formulating suggestions to matchmaking imo. It takes 8 wins to get out of unranked, even the shittiest player will get carried to 8 wins if they play enough matches so I'm skeptical that they even PVP at all.

Neither option literally does nothing to change how many people are in silver+. I don't know where you get that idea from. It only makes it possible for people who are currently unranked or sub below 38% win rate to actually drop rating. So you are just wrong there. Again I should clarify my second option has people START at 1000 and allow people to fluctuate from there with bronze being 0-1500, silver 1500-1900 and so on. No other difference in silver plus.

Also, yes there are people that literally do not have what it takes to win with where they are placed. And yes they do pvp, idk why you are making a assumption that they don't when I clearly say that they have been playing since the beginning. That means I still see them PLAYING PVP. And I just got to say, if these players can't actually go negative or w.e, then not even the best players can carry them, it might as well be a 3v4.

[Zojha]

That isn't true.

It is. Let's say two exactly equally skilled people people queue at different times, one at off hours where only hardcore PvP players (still) play and the other during peak time where there are also several newbies and moderate players around to be matched with. The latter wins more games than the prior, despite being equally skilled, simply because the matchmaker could also offer them enemies not way out of their league.

So both people are equally skilled, yet both have different rating. Accuracy much? This is actually precisely the fact why wintrading and other such practices work - if you only got matched with the trading group once in a blue moon, it wouldn't have a big influence on your rating and the rating would still be largely valid. Since you get matched with them all the time, the influence of that one group is huge, which makes the practice work- and profitable and the rating very inaccurate. And the same goes for legitimate players that are just better/worse.

Rating can never be accurate without a vast pool of players. That's just a statistical fact, just as you cannot accurately determine a ranking without a large amount of games. A person having a 100% winrate over 2 games isn't necessarily more skilled than one with a 75% winrate over 200. And a person having 0% winrate against the top 10 isn't necessarily worse than a one having a 30% winrate against the bottom 10. With a large playerbase, that's a non-issue, you just don't match people with people way outside their league up or downward. With the playerbase as small as it is, to the point you frequently recognize your enemies by name, that is a big issue that distorts all ratings in the system.

[Aviars]

It is.

Your explanation can still apply to large populations of people as well. Queing on during primetime or downtime still applies to large populations of players. I mean people have been wintrading in WoW and still do, player population doesn't change that.

Anyway, you still have failed to address how people not losing rating for when they lose isn't causing a problem. You probably haven't because you can't really say it's wrong.

[Zojha]

You probably haven't because you can't really say it's wrong.

Oh, I don't disagree on principle - I just find it a moot point when the rating system has no solid base to stand on in the first place. And if people can still viably wintrade in WoW, it just means that the population at the time of the wintraders queuing is small - so small that the trading teams are matched against each other often enough to make a difference. It makes a world of difference if the matchmaker has 1000 teams to pick from or 2, namely that you'll not be able to play against the same team very often in the prior case, which vastly improves accuracy of the rating.

That's just the population/time bias. We also have our lovely tier biases, because the tier affects how much you gain lose, whereas the matchup scarcely does. Stomping newbies first round grants/loses as much rating as a hard-earned clutch victory in the last second caused by even teams. Being in a lower tier however automatically loses you less points per loss than being in higher tiers. A perfect 50% winrate at bronze would allow someone to gain rating faster than a 50% winrate at silver or gold respectively. (And truly, a 50% winrate shouldn't even allow someone to climb in the first place).

Do I have to mention the bias based on roles and jobs? Not all roles and jobs are equally influential, which again, screws with rating accuracy.

With all these issues, yours just being one in a pile, I see no point in having a rating system in the first place. It gives people the false illusion that it's an accurate representation of skill when it fact, it suffers from loads of issues that make it more of a broad hint.

[Aviars]

Oh, I don't disagree on principle

With all these issues, yours just being one in a pile, I see no point in having a rating system in the first place. It gives people the false illusion that it's an accurate representation of skill when it fact, it suffers from loads of issues that make it more of a broad hint.

Don't know why you don't think rating can't be relateable to skill. The top 10 of both Aether and Primal have familiar faces in them. I wonder why? Is it just luck? You sound like that one guy who said your rank is all luck, delusional.

But anyway I'm not getting into issues of player population or w.e. I'm am talking about a issue that everyone has been complaining about and I give a explanation why and how to fix that one problem. You going on about scrapping the whole mode is extreme and makes no sense.

[Zojha]

Don't know why you don't think rating can't be relateable to skill.

I'm inclined to wager because you didn't read my post thoroughly, because I explicitly called it a broad hint, which indicates a relation.
If someone keeps a low rating, he's probably bad - probably because the rating is distorted through numerous factors and there is reason to doubt that the active PvP population is a statistically representative sample for the whole - chances are, good players are more likely to play regularly and there is a selection bias.
If someone rises to a high rating, he's probably good - probably because wintrading and boosting exist and you know it. It happened last season, but surely the guys in question just had a sudden skill explosion/implosion.

However, with the rating distortion factors, you just plain cannot say two people of the same rating are equally skilled - which means it is not accurate - A DRK with the same rating as a WAR is most likely higher skilled than the WAR, rather than equally. It's not accurate in bronze, it's not accurate in silver and it's not accurate anywhere else, because it cannot be accurate under the given circumstances. You could scrap it entirely and have completely arbitrary tier promotion/demotion series instead, better players are still more likely to win those on average (thus having a higher chance of getting/keeping the higher tier) while worse players would stay lower and accuracy wouldn't considerably suffer in the long run. Which is why I think the inaccuracy at bronze you point out is a non-issue and a rating system as such is pointless, seeing as the point of a rating system is to provide an accurate skill metric, rather than a shotgun approach.

And since there is precedence of you making assumptions, let me spell it out: This neither means that the rating system must be scrapped nor that the Feast must be scrapped and I suggest neither. For one, while interconnected, they are not one and the same, for another, there's no point removing pointless systems unless they do harm and the rating system only does harm if you fall for the illusion of accuracy. I do suggest however not to put any more work into it until the population issue is fixed and we can have proper, job specific Elo for our matchmaking rating.

Lastly, I would refrain from hollow ad hominem - it's not good style. But that's just me.