Hey all - I wrote this post earlier and after seeing another post about rating hit the number 1 spot on the subreddit I figured I'd make a followup post.

The current debate is over the ranking system and how it feels like there is not enough spread between ranks. Part of this comes from the natural way that most people who have gone through the Western education system view grades. 50 feels bad, and 56 (though MUCH better - about 34 percentile points better!) doesn't feel much better than 50. However, Blizzard definitely wants to make the rating ladder matter, and if players are hitting ratings in the high 80s and 90s that represents a REAL achievement worth praise and admiration. In order for this to happen, there must be a serious barrier to achieve that level of play.

The way I see it, there are a few main questions that need to be answered.

1. Do we want a symmetrical or skewed system of ranking? ex: Chess Elo and percentile rankings are symmetrical, League of Legends tiers are VERY left-skewed (most people are in bronze and silver)
2. Is there a minimum ranking below which players should not be able to drop?
3. What amount of difference in rank should represent a REAL achievement?
4. What amount of statistical change (I'd recommend using standard deviations assuming normal distribution) should be considered a REAL achievement?

I'm assuming Blizzard has thought about most of these questions, but it's worth hashing these out as well since there seems to be quite a bit of frustration with the current system. Here are my thoughts, though I'm sure that I don't have all the answers and that I haven't considered every relevant factor.

1) Symmetrical vs. Asymmetrical vs. Hybrid - this is, in my opinion, the most critical question to answer. Riot Games as a competitor has chosen Asymmetrical - the vast, vast majority of Ranked LoL players are in Bronze or Silver, and Platinum/Diamond/Masters/Challenger are separated by slivers of percentage points. Riot seems to intentionally hide percentile ranking and it seems like standard deviations are not considered at all - this is simply a ladder system with more spread at the top. The logic behind this system seems to make people at the bottom of the scale feel less discouraged. After all, if your rank is obscured somehow and literally millions of players are in Bronze 5 with you, you can't be THAT bad, right? There's always someone worse than you. Conversely, the system has very fine differences between rankings towards the top of the scale, and that is where this system shines. There are a lot of benefits to this system, but it leads to serious frustration for people trying to climb. Moving from Bronze V to Bronze I is a huge jump in percentile ranking, but at the end of the day you're still in Bronze. How successful this system is depends on the perception of the system itself.

Some people might be inclined towards a Symmetrical system (and Overwatch currently seems to be purely symmetrical!) and I don't blame them - but there are advantages and some major disadvantages to this system as well. The advantage is primarily in transparency and ability to compare skill levels between players. With a symmetrical system, as long as the mean is known you can know exactly where you fall and how much you have to improve before you rank up. It is also a great way to show improvement! Moving from rank 45 to 51 in Overwatch is a BIG DEAL.

Generally speaking, however, systems that are symmetrical are much more transparent than Overwatch MMR. Chess Elo is brutally transparent - if you are at 1068, you are average. If you are below that number, you are below average, and if you are above that number, you are above average. It is worth noting that Chess Elo more directly correlates to skill than the Overwatch MMR does, but this kind of completely transparent system does have the tendency to discourage those at the bottom of the scale. Blizzard has attempted to address this by limiting the scale to roughly ~15-85, but that has its own issues. As for why this is the case, I'll refer you back to my first post on this.

2, 3, 4) What amount of rating should represent a truly meaningful change? What amount of statistical change is meaningful? Is there a rating below which players should not be able to drop?

I like somewhere between 10 and 15 for one standard deviation. Have the Average rating be 50, like it currently is. Set 10 as the floor. My proposed system would be something like:

A good system feels rewarding at every meaningful level of improvement. If you improve from 10-20 it feels good, and that is a meaningful statistical change in your rank. If you improve from 30-40 it is still a very meaningful change, and it still FEELS like you are going up quite a bit in rank. This continues up the scale. From 81-90 you're starting to get into elite territory - that would be like improving from Platinum to Diamond in LoL. From 90-100 that's what separates great pubstompers from the pros.

This is a hybrid ranking system - it is mostly symmetrical, like the current system, but it adds a floor and is easier to move at the top end. Rank 100 is attainable in this system (as is, 100 is completely unattainable in our current system). Ranks below 10 are not allowed due to feel bads. Though that could be manipulated in a variety of ways.

Let me know what you think. This is one possible system, and there are many, many others. What do you want out of a ranking system? Would you prefer a tier system like LoL?