Loading...
  OR  Zero-K Name:    Password:   

elo on big big user page

37 posts, 918 views
Post comment
Filter:    Player:  
Page of 2 (37 records)
sort


DErankChesti
9 months ago
hey secret-devs
You hide elo again on frontpage. Can we have it back please? 1 vs 1 elo and Casual elo, visible for everyone?
+4 / -0



AUrankAdminGoogleFrog
9 months ago
Idk. First figure out who wanted it to happen.
+0 / -0



CZrankAdminLicho
9 months ago
CHrankAdminDeinFreund
+1 / -0


AUrank4hundred
9 months ago
Arbitrarily removing shit; shits people.
+1 / -0

NLrankmalric
9 months ago
DErankChesti you are not the first to notice: http://zero-k.info/Forum/Thread/24740
+0 / -0


DErankChesti
9 months ago
malric I know...
but secret-devs ignored it- ;-)
+0 / -0


CHrankAdminDeinFreund
9 months ago
I've added a whole new website section dedicated to nothing but displaying ratings. The background image on your profile is exclusively used to depict your elo. There are new ranks and progress within ranks to show elo progression.

But you ask me to show you the number you're used to seeing again. The number that drifts over time (look at chess ratings). The number that doesn't show your progression. The number that has a misleading amount of decimal points. Yet you have been seeing these ratings for years, and thus are an expert in interpreting these numbers and can extract more information than any statistician could ever fathom.

You ask me to separate casual and competitive ratings. You think it rates you too high, you'd rather have easy games and have the newbies tank for it. You'd rather add a pseudo-random offset to your ratings so it has more inconsequential correlations. You don't care about the statistics, you want to cook lobsters.

Why not just revert to Elo? Elo has:
  • No display of statistical significance
  • A lot of random variations that can be freely interpreted
  • Convergence times so long that smurfs can create 10 new accounts before being balanced properly
  • and many more cool features!

Why not just go all the way and look at the most successful F2P games. They often don't have any real rating system at all, meaning you can play random matchups all day!

Go ahead and revert all of my changes, I'm sure you'd like it more. I'll find another use for my free time.
+3 / -0

NLrankmalric
9 months ago
(edited 9 months ago)

Do you read what we write?

So, here we go again, quoting myself:
quote:
Many people like rankings, and some idea to assess "how far" are you. For me colors are "too generic" (but please don't come with the idea to remove those as well :-p, they are good in game to know what you can except). The percentile and the ranking considering only active players was perfect considering what can be available. And the fact that it can jump it is only normal as rankings change always...

Also:
quote:
I feel there is a discrimination between first 50 players and the rest (not being able to know how good they are except an icon color).If you really believe that numeric ranks are not very informative, then remove the ladders as well.

And:
quote:
And in the end what is the "official objective" regarding showing ranking/ladders? It is frustrating to get these back and forth changes without understanding the reason behind (and lets keep separate the underlying algorithm and information used for balancing, I think in that area the objective is "make the most balanced games")

Reference: http://zero-k.info/Forum/Thread/24740

So, not sure why you comment on which system to use (you say "why not revert to elo?") when many people complain of not showing a number/percentile, while I feel most people get it will change/wobble, get it is not perfectly informative, get it can be used internally in whatever way is needed.

And I find it quite hypocritical to show the number in the ladder, but not in the user profile...
+2 / -0

LUrankAnir
9 months ago
CHrankAdminDeinFreund, you are taking this way to personal.

First of, we didn't knew it was your idea to remove the exact WHR number from the frontpage, so how could we try to work against you.
Secend, what the hell does a number on your userpage has to do with beeing a smurf, or to like killing noobs?
Third, i dont see where chesti asked for elo to get used again.

So as this is cleared (And dont come up with that again), we can go to the real topic.

What is the problem in showing us a number on our frontpage? For real. The statistic page is like usefull for nothing at the moment, because the reaction time of the chart is like slow, and i rly mean slow. If i would play a year, it would maybe change by 10 points in total, while my real WHR jumpes up and down. Also you are favoring the top 50 players, because of the Top-Lists.
(Btw when i use the statistics to add other peoples charts, half of the time i land on their frontpage, because idk, missclicking of whatever. That alone makes it already a pain to use.)

And since you mentioned the Casual and Competitive Rating connection, here is some more on that point.
Making those two dependend from another is like the most stupid thing that could have been done for the rankings. I know it is like that for a long time already. I mean, there is 0 reason to do that, because if you have one competitive player, having like 2100 elo on that ranking, and he now suddenly plays teamgames, then his WHR will jump to his teamgame elo in such a short time that it doesn't mather, and it will be closer to his actual whr then atm.
+2 / -0


CHrankConnetable
9 months ago
(edited 9 months ago)

LUrankAnir , you cannot say "there is 0 reason to do that," because actually CHrankAdminDeinFreund explained that including casual elo into competitive elo was IMPROVING the predictive power of WHR.

So you cannot say there is no reason. Now, it has what some players feel is a downside, which SErankGodde mentioned. The difficulty is to make a judgment call between the up and the downside.

Maybe CHrankAdminDeinFreund reacted a little to strongly but I can understand that after having spent that much time bringing up a better rating system, the fact that some guy comes up and say in two lines "secret devlobs revert things pls" might feel annoying.

(pro-tip: you can see your cherished number DErankChesti by hovering over your name)
+0 / -0

NLrankmalric
9 months ago
(edited 9 months ago)

Can we please focus on an issue at the time? This thread (and at least another one) is about showing (respectively not showing) number(s) on the user page, while they are shown in the ladder/top 50 - which is something inconsistent.

I could have seen an argument for hiding all the numbers: to avoid people commenting about how it should be computed, people should see just the final ranking (ex: Godde's thread and the fashion of creating "CasualX").

The strange thing is that now we have a situation that annoys a lot of people and seems bad for the game (people not in top 50 because it is not obvious for them how they compare with top50, top50 players because when they play 1v1 they see the rating changed in the other ones which makes them create smurf, developers because people complain and they feel not appreciated).

An example answer to the questions about the number missing could have been: if you do mouse hover over the bar of the rank there will be a "progress", which is the new "number" you can check. But what exactly are those ranks and progress is again not explained at all (tried to look in the wiki for them, I could look in the code but...). Edit: for reference, I liked best the percentile number, and guess that "transformed" to the progress, but seems people prefer to be offended rather than give a simple explanation and we need to discover/imagine things...


+0 / -0


CHrankAdminDeinFreund
9 months ago
It's not just this thread, there's a dozen other people with other wishes for the ratings. I understand and appreciate that zk development requires constant discussion with players and other devs, yet I neither want to continue working on this nor think it'll improve the game. If you need help with implementation ask me on github. I'd appreciate if you considered the statistics beforehand and I don't recommend crowd sourced web design.

A lot of thought has gone into developing rating systems leading up to whr and every modification we make for zk is butchering those algorithms. This hurts.

I am looking forward to a successful steam release.
+6 / -0

NLrankmalric
9 months ago
For me the problem is more about communication - the best work will be ignored/changed/not appreciated if not properly communicated (and already seen that in real life many times).

My feeling is that the math/theoretical part of WHR/rating is great (better than ELO etc.), but the presentation/interaction/communication is inappropriate.

What do you want to achieve? Shouldn't the target be "keep most players interested/happy/engaged"? If you want to "prove WHR is better algorithm than ELO" then I think you are talking with the wrong crowd and will get frustrated.

In my opinion the internal computations used to determine rank and balance should be separate for what it is displayed, because the objectives are different (internal computation has objective best statistical result, display has giving a feeling of ranking to players). And by separate I mean, not exactly the same, it can be "based on", "derived from", etc.

As I do not contribute actively here an opinion is not very relevant, but judging by the results (many forum threads, various people unhappy, you included), maybe the devs can have an internal discussion and think what the problem is or at least what they want to achieve and try to be consistent...


And regarding "crowd sourced web design (of ranks)", my feeling is that it is "random web design (of ranks)" (by random meaning not backed up by a clear consistent goal, and considering the users which will get confused).
+0 / -0

LUrankAnir
9 months ago
Oh dont get me wrong, i agree that WHR is bether then elo.

But just as NLrankmalric sad, the communication is lagging.



quote:
Anir , you cannot say "there is 0 reason to do that," because actually DeinFreund explained that including casual elo into competitive elo was IMPROVING the predictive power of WHR.


Of course i can, and i still dont see an real argumentation against it.

You can't ignore the fact that seperated WHR Ratings will always be more precise then merged Ratings.
+0 / -0


ILrankAdminhokomoko
9 months ago
quote:
You can't ignore the fact that seperated WHR Ratings will always be more precise then merged Ratings.

You actually can, as it's false.
Your accuracy depends on two things:
1) The number of data points
2) Their quality

By separating the ratings, you reduce the number of data points in the casual rating and you increase their quality.
For some players that may have positive effect, such as players who play a lot of MM and a lot of teams games with significantly different skill.
For others it may have a severe negative effect, such as new players who played 10 MM games and then drop into their first team game. Separate ratings will have no way to rank them, while merged ratings will have a good clue.


These are the two extremes, I suspect that most regular players probably fall somewhere in the middle, where separating the ratings won't change anything, but the statistics do show that in general (which contains lots of users playing very few games) this has a positive effect on accuracy.
+4 / -0

NLrankmalric
9 months ago
Statistics are a lot about assumptions. Are we making the assumption that a player plays the same way in matchmaking and casual?

I think you can prove that using combined ranking gives better results than separated, because on average people try to play as good as they can in both. But some players will play differently (ex: computer lags in big games, or they like to troll in big games and not in matchmaking) then they might "suffer" and choose not to play big games because they will need to play at their best. Then you can claim based on numbers that the system is "great" ignoring the actual real life effects...

PS: damn, you made me reply on something that I do not think we should discuss here, which is the theory of rankings, versus what you should display. Well done ;-)
+0 / -0

LUrankAnir
9 months ago
for most players, the MM games and casual games have a ratio of like 1:15? or 1:30?
For most players i guess 1:100-1:1000

So those few games should make the difference? Shure it makes the overall Rating more acurate, but then you dont need two rankings, because one makes the other one irellevant.

But NLrankmalric is right, this is another topic and only a repetation of what has been sad already.
+0 / -0


PLrankAdminSprung
9 months ago
quote:
You don't care about the statistics, you want to cook lobsters.

dont we all
+1 / -0

{redacted}
9 months ago
(edited 9 months ago)

Actually I always only ever cared about the most accurate number and full information as far as skill ratings go.
+0 / -0

NLrankmalric
9 months ago
If you cared that much about statistics you can understand why people not in top 50 want to see more than a colour - which is why all this thread started....
+0 / -0
Page of 2 (37 records)