Ranked Ladder?!

This is not true.

Can we ask if you use Elo or Trueskill?

Actual information on it might kill rumors indeed.

It's a Glicko-derived rating system.

Is it derived from Glicko 1 or 2?

One big problem I have with the top heavy distribution is that it gives players no incentive to ladder. Why play when you are already in the top league? NO ONE should be placed directly into Uber league similar to Master league in SC. It should also have a 50 player cap to make it a goal for people to achieve, which encourages continual play.

My other problem with the ladder is that it shows no statistics. Give us win/loss, match history, or show us the hidden rating behind the system. The matchmaking is working great otherwise, but give people incentive to ladder!

They just told us the system.

Glicko.

We agree that "Uber" rating should be more special, and we're currently thinking about various ways to improve it. I've previously explained why the specific player counts are extra skewed right now (it is because not all players are included in the leaderboard), but even with all players included we'd like to change the rank splits.

This is one of those things that can be hard to experiment with without having real player behavior -- I hadn't considered the impact of players who didn't play enough matches to show up on the leaderboard.

I am saying show us the rating that they are using to rank us. Show us the actual Glicko rating. I guarantee there is a huge discrepancy in skill among the top 10 players and we currently have no way of gaging that. You have no way of knowing how close you are to passing the guy above you. Once again, StarCraft 2 has pts that are visible so you know "if I win my next two games against an equal-skill opponent I will pass the dude above me".

Consider splitting up the ability of placement, between only people who do place? Basically, like what it is now, except instead of basing placement on the score, base it on what percentile of the total number of placed people that they fall in. Top 10% of all placed people = Uber. Next 20% = Platinum. Next 20% = Gold. Next 30% = Silver. Next 20% = Bronze. Or something. Maybe 10%/10%/20%/30%/30%?

Just, don't rank people based off a score that also includes non-placent people. Take the people that do place, and then divide them up by the percent of other placed people, and give them a tier based on that.

So for a person to get a tier, they are literally in the top 6% or top 28% or top 43% or are they like the bottom 18%? You know?

As long as people who don't place, stop affecting that percentage. That they are not any part of the division of placed people against each other.

Actually the *really* interessting MMR (the rating points that show the skill of a player as a number) is hidden in sc2, at least it was when I stopped playing 1 or 2 years ago.
basically all points and divisions and stuff you can see there are for show only and not directly related to your MMR

This is still true. However there is still a strong correlation between the amount of pts you have and your true MMR rating. Blizzard realized that they want to encourage continuous play so they added Bonus pool which accumulates over time. The bonus pool matches any pts you win and reduces the amounts of pts you lose. This system ensures that only active players can reach the top ranks because the number of pts a player has increases over the course of a season. As is, I am rank 1 and I could not play for the next two months and I could still be rank 1. The current system does not encourage active play. I would love to see the actual Glicko rating though regardless because it is the best indicator of a player's true skill.

Reason I didn't mention bonus pool was because I think that would be too much effort for Uber and I am worried they will not make it better until they iron out a lot of other stuff first. Simple fixes to the leaderboards first before they lose interest!

Doesn't glicko feature deterioration of the rating of inactive players?
Same with ELO sort of, as it features an inflation of points.

But yeah I agree the system from SC2 is pretty well thought ought and it's a good example to look at for a well made system that keeps people playing by using a lot of psychological tricks.

I think you are right actually that there is deterioration over time in a Glicko rating system. Ignore me, although I might still be able to play 1 game every month to stop the deterioration because I'm not sure how strong it is. But yeah, the StarCraft ranking system is one of the best, if not the best (I think LoL is similar). I can see my match history, my ranking compared to 100 ppl in my division as well as my global ranking, my map win rate statistics, my winrates against the various races, and other information. Keep in mind, it took almost 3 years for the system to become that in-depth and its a AAA big-budget title.

If anybody is interested in Glicko rating systems go here.

There is deterioration over time in a Glicko rating system if you don't play in a giving rating period. This depends on what version of it is being used which is why I asked. The Glicko system introduces the concept of rating deviation which tells you how accurate rating is. Glicko2 introduces rating volatility which shows the expected fluctuation of the player's rating.

Starcraft 2 rating system was designed to make people play longer by demonstrating progression and being friendly to non-hardcores. Nobody likes having their arse handed to them. Everybody likes to see improvement over time. The bonus system was born. It most likely uses actual rank for matchmaking but not really sure. I'm pretty sure Blizzard uses a modified Glicko system for its rankings from what I have read.

Stuff I read:
http://www.glicko.net/glicko.html
http://sirlingames.squarespace.com/blog/2010/7/24/analyzing-starcraft-2s-ranking-system.html
http://www.teamliquid.net/forum/starcraft-2/118212-sc2-ladder-analysis-what-you-need-to-know
http://www.teamliquid.net/forum/starcraft-2/142211-sc2-ladder-analysis-part-2
http://research.microsoft.com/en-us/projects/trueskill/faq.aspx

Edit: http://www.chess.com/article/view/chess-ratings---how-they-work
http://www.chess.com/blog/kurtgodden/elo-to-glicko-your-rating-explained

Absolutely agree! Uber needs to keep in mind that people will only persistently play ranked games if the leaderboards or some new "Your Profile" menu show clearly your progress over time (recent history of actual position in ranks, statistics like ranked win/loss ratio or average game time, recent games with replays...). I really think a feature like this, allowing players to see their own and other player profiles, should be implemented at some point.

Agree, which is why SC II has the bonus pool system because it demonstrates constant improvement to the player. They don't prominently show the actual rating of the player because eventually it plateaus when the algorithm figures out where they should be.

There is still the increase in rating because the player becomes better at the game. (But this improvement might also converge over time, not sure about that.)

League is only similar in the names of the ranked groups : bronze, silver, platinum, diamond, master, and challenger/grandmaster. For me one of the best games in ranked systems is dota, it calculates a number of things (gold per minute, xpm per minute, dmg to structures and units etc. etc. etc.) and then based of that it places you. League has the same problem as pa, you win all games and get place way higher than you should be.