Alright, so going back to the spec idea instead, it then leaves these issues unresolved:
Well, the last paragraph would be solved with it being optional. Obviously. But the MMR manipulation is a real concern.
The artificial inflation scheme would just need some tuning, so that is also possible for them to solve, although I doubt they’d be able to fully predict its impact so they’d need to tune it constantly until a clear pattern would emerge.
Good point.
I think more reasonable solution to problem with playing with alt specs, that you don’t really play well yet, would be to implement simply rating checkpoints.…
“If you reach … rating, then you cannot fall under … rating”
Then you could learn at higher rating without actually losing too much and thus being disincentivised to try new things. This also works on multiple levels, not just for specs, but also for gear / covenants / comps.
It might be even done in a assymetric way.
For example: If you reach 2100 rating, you cannot fall bellow 1800 rating this season (basically as PvP ranks are, it would mean new rule: “you cannot fall two ranks below max rank you earned this season”).
That idea, I think sounds good. Although I’m hoping they get rid of upgrades on gear so that it won’t become a concern for a system change like that. 
Yes. It’d be a huge concern and bad idea now as well.
First they really need to fix gearing. Remove gear upgrades. It’s unhealthy now as it is and such great implementations would make it even more unhealthy.
Original idea. However, wouldn’t it favorise rating inflation? Also, for the first few games, your mmr is gonna be extremely high for an alt spec no?
No, it wouldn’t favor rating inflation. As long as the MMR isn’t prevented from falling then it would have no impact on rating inflation in any meaningful way whatsoever.
A potential issue would however be boosted players. Depending on where those checkpoints would be placed, they might be able to basically act like a “poison” on the group finding system even more than now, to trick other players into getting invites they shouldn’t be getting, since their rating wouldn’t fall.
The only rating inflation that impacts the matchmaking system, is the MMR inflation. So “current rating” is irrelevant for the matchmaking. And the events that inflates the MMR is increased both naturally and artificially (by Blizzard) the more games that’s played.
Personal rating matters a lot for the title cutoffs that are based on percentages, but since they were removed and replaced with static cutoffs instead, it’ll have no impact on that at all. Only r1 cutoffs could be impacted, if they’d place checkpoints that high up in the ladder, although that’d be very unlikely.
And by “that high up in the ladder” I mean literally at the r1 cutoffs, which would be the only way r1 cutoffs would be affected.
The reason why it wouldn’t affect the static cutoffs is because as long as the MMR moves freely, then rating systems with a separate MMR are fundamentally designed to move the rating held ever closer to the separated MMR in various ways. So dumping the MMR while keeping the personal rating higher, will never lead to an increase in personal rating in a way that would let them reach the next checkpoint without first getting the MMR to that rating as well.
It’d need some recalibration in the artificial inflation scheme though.
That would be solved if the rating was symmetric instead. (If you reach 1800 you cannot fall below 1800)
But let’s be honest, this might be issue even now, just maybe less often.
That’d go against the core purpose of a rating system. But a way to deal with the problem of boosted players in such a hypothetical, would be to carefully place checkpoints and how the checkpoints would function.
For example, like you mentioned here:
That’s why I wrote it only as a potential issue, because it only depends on where the checkpoints would be placed, and how far it’d still be able to drop the rating.
So asymmetrical checkpoints > symmetrical checkpoints, since asymmetrical ones still allows for a rating system to do what it needs to, which is to quantify player ability relative to every other player in the system, at that moment in time. Because a player’s ability isn’t static, and rating systems are basically mathematical proofs to quantify it with the most accuracy theoretically possible.
Symmetrical checkpoints wouldn’t really be a rating system anymore, it’d be more like a simple point competition with no purpose to quantify ability behind it. Or maybe I should say, it’d trade away a big chunk of that accuracy if it would be symmetrical. There’d still be some comparison possible, but not as intuitively.
Well not necessarily. Depends how you perceive a purpose of rating system.
I percieve it more like high score game. And in such, there are often checkpoints too.
I’d place it as tiers are. If you reach duelist (2100), you cannot fall bellow 1800, etc. If I were to program it in code, it’d translate to condition sounding like that
If you reach a new PvP rank, you cannot fall under the minimum rating treshold of previous rank. And since previous rank is 1800-2099, then if you reach duelist, you cannot fall under 1800.
Possibly, yes. But that again depends how you perceive rating. Whether a simple race to high score or whether an establishment of player’s skill.
Because I’d wager while developers might see it as latter, players see it (when they themselve play) as former.
I wouldn’t really say that, we rely on that core functionality of a rating system to get a feel for levels of ability. So we don’t see that as a “race to high score”, but rather as a quantification of a player’s ability. It’s already common practice, and have been for a long time.
Also, it’s not a matter of perspective, the rating systems used in games are all based on chess rating systems, and they’re all designed with the core function to quantify that. Some do it better than others. It’s the reason why you can lose rating, instead of only being able to gain rating.
The MMR in WoW also got a built-in volatility, based on Glickman’s rating deviation systems, to move you quicker to the rating where you’d belong, while the Elo system is a lot more basic and slow.
But anyway, what they both have in common, and so does WoW, is that where your MMR and personal rating stay close to the same with a 50% win rate, that’s when the system has quantified your ability to say that that’s where your ability to play belongs at that moment in time.
That’s why it matters that you can lose rating. If you could never lose rating with symmetrical checkpoints, then it’d lose a big chunk of that accuracy in its ability to quantify ability, and it’d make it harder to get a read of what people are capable of.
At least with asymmetrical checkpoints, like your example, it’d still keep more of that accuracy since you’d then be someone who got 2.1k but would then currently have 1.8k so you’ve lost a lot. It does lose the ability to show people just how far you’ve dropped, but -300 tells a bit already. People would be going more blindly in the group finder if it’d be symmetrical and the person would never drop below 2100 after having reached it once.
Do we? I certainly don’t. In a week, everybody’s rating is going to be zero, but I don’t think that means their skill is the same as well.
That’s true, but their downside is, they don’t really incentivize to try new things if you need to drop too much rating in order to do so.
So you don’t think of a 3k player (of course, assuming it’s legit), as being good? You think it’s possible they’re really bad? What about a player who gets 3k every season for years in a row? Are they also maybe really bad, in your eyes?
That’s kind of irrelevant, it goes without saying that it takes a while for a fresh ladder to stabilise.
This isn’t an example where we are in conflict however.
The example where we are in conflict is when you drop rating.
And yes, I don’t think multiR1 player that does have 0 cr in this season is worse than 1500 cr and max xp player this season.
That’s where we are in a conflict.
Because I consider it more like a high score (how much you could reach). So a 3k player reached 3k, so that’s how I assess him.
You refused my metric and told me that it’s more like
And having 0 cr but 3000 (high score) max rating doesn’t in my eyes mean they’re worst player.
And that’s EXACTLY why I said that people perceive it as a high score and not according to CR necessarily. You disagreed. But your example is very bad, because it doesn’t reflect any conflicting situation.
You seem to have misunderstood a part of what I’ve been saying. A rating system in games quantifies ability, that’s essentially what they do. So on a theoretical level, a rating system without any played games whatsoever, can theoretically have anyone at the top and anyone at the bottom once games have been played. Because a rating system doesn’t take into account past performance that has been reset because of a season change, a season change is essentially a different system that just ends up resetting the rating system.
Naturally, past performance can still be used as an indication, but it’s irrelevant to the rating system’s mathematical functions to quantify ability.
But ok, so you had a problem with the example I gave you, so I’ll give you a different one that might shine a light on the problem better:
What about a player with 1800 rating and 50% win rate with 200 games played? No past record either, except for peaking at 2k that season. Would that be a player, without any coaching or help from anyone, be able to play on 2500 rating and play equally as well as all the other players on that rating, in a season that has been ongoing for let’s say, 2 months, according to you?
Or should I change the question a bit, and ask you if you think he’s a 2k player or an 1800 player, based on his win rate at 1800 rating after an excess amount of games played?
The essential point I’m trying to make, is that we use the rating system to infer ability, not based on high scores alone, but rather on performance using the numbers and functions the rating system provides, including the loss of rating.
Rating system doesn’t and that’s exactly why I said you’re wrong.
Because people don’t perceive their current rating as assessment of their skill, but rather they see it as high score (how much they’re able to earn).
Probably not. But again, you chose an example, that says his high score is 2000 and you ask me if his high score can be easily 2500.
Which again shows me to be right and people perceiving it rather as high score.
Well and I simply disagree. Majority of people ask for highest earned achievments and thus perceive rating as high score while dealing with other players, but also why playing themselves (because they want to raise their high score rather than simply play without caring about current rating as it’d simply reflect their skill so in your worldview, they shouldn’t really care about that).
I am talking about perception, not what it is, nor what it should be.
And people TRY to raise their rating as high as they can, and judge others mostly based on how high they were. Thus they perceive it as high score.
^ Then answer that.
i believe most people would assess him as “2k player”
Which would grossly confuse peak and average performance.
That would only matter if there was high difference. For example somebody would play regularly between 1800 and 2200, while other person would play regularly o 1900-2100.
But I don’t think that’s really the case.
Both of these players would’ve had average of 2k, but one of them would’ve had higher peak value. But then again, I don’t think there are many players that have such a huge deviation regularly (one session -400 rating, other session +400, next session -300 and differences like that while having same average).
And even if I did, again, I talk about perception.
That’s a misrepresentation. Which you even covered yourself with your edit:
A 2k peak but staying at 1800 which is where the volatility levels becomes the most stable, thus staying at 1800 which is where the win rate averages out at 50% constantly, thus being “stuck” at 1800 but having peaked at 2k. That’s what it means to average out at 1800. Not to average out at 1800 by fluctuating between 2500 and 1100 rating, or other silly examples like you pointed out yourself in your edit.