Not sure if this is by design or it's broke but I only seem to be able to gain 1 elo per win, regardless of who the enemy is. If they're above me, they lose a lot of elo but I will only gain 1. Does this mean that climbing up the ladder is a consequence of knocking other people down the now?
+0 / -0
|
This happens when you're inactive for longer periods. WHR no longer updates your ladder rating, but once you get back in, your gains/losses will be biased until you're back to your true rating. The other way for this to happen are imbalanced battles where WHR now gives you fake ladder rating although you deserve none. The third factor maybe be playing too many games in a short time period. Rating changes then decrease as they converge to your actual skill unless there is a major change in skill.
+0 / -0
|
So the solution to people losing rank when they win is to instead only give them 1 point when they win?
+0 / -0
|
Losses are smoothed out over multiple matches, but if that's not enough the one point gain is the last resort. Unfortunately that can even be too much sometimes, since one point is still a great reward when you're winning games with 95%+ win chance.
+1 / -0
|
It's unusual to imagine losing rank on win for any reason. Right now, I'm a little mind boggled by the system because I gain 1 elo whether it's someone 30 places beneath me or a higher player (been toasting the rusty drone / randy). I'm guessing they don't offer much reward because they've been inactive and the only way to gain elo now would be to win against Godde or Kingstad?
+0 / -0
|
I've given my rating time to "settle" based on what Dein told me about how it was working but I'm still left with the following conclusions: My uncertainty value is ~100 despite playing a lot. I'm still only gaining 1-2 rating per win, even vs UG it was 2 points. It feels as though the system has reached an inflated point where the gulf between players ratings are really vast and you get put into a bubble where you either gain super slow OR get beaten by someone not far beneath you in skill, but very far in rating because of the system, and lose a load of rating. Yet, even when you beat someone much higher in rating than you, you're not rewarded much. Is that because the system believes you just got lucky? Another thing of note: If Godde never plays, it would be impossible to get #1 because you just can't climb this ladder, you have to tear people down. Essentially, it's better to be matched against someone higher rated than you but fighting people lower down is high risk with little to gain. Honestly, even if the system IS working properly, this feels stagnant. I'd suggest calling in a poll to see if people would be interested in a ladder reset so that we can start fresh after all those changes got made.
+0 / -0
|
The system is working astonishingly well. Your recent WHR estimated rating was all over the place, yet your ladder rating is stable. Meanwhile for newbies and others with large skill changes the rating changes quickly. So now we have the pros of the highly variable rating while still having a stable ladder. Who was it again who told me a few months ago that the ladder is way too easy to climb/drop and that all the time he invested climbing it was now worthless. Now we have the opposite and here's somebody asking me to circle around.. I'd love to implement seasons, if there is a concept that makes good use of it and a plan to raise MM activity to a point where we don't see repeated 98% win chance matches.
+2 / -0
|
Seasons shouldn't wipe actual record of what's used for team balancing, so we don't have to deal with seasonal semi-random team games.
+0 / -0
|
I get how it looks like I'm saying two things at once Dein, I do. What I was saying before was relevant at the time, but it's been some time since the system changes and I think it's effects are only just becoming apparent to me. I'm going to call it the "Expanding ladder theory". Back when the changes were made, the newer system was fresh and people were taking fast chunks out of eachother's rating, but now peoples ratings look more like galaxies drifting apart than close competitors. Whilst it appears on the surface that it's working well in my case because my rating appears stable despite heavily variant WHR, I think that's overlooking the heart of my issue - which is why the WHR is so variant in the first place. In this heavily inflated phase of the ladder that we are in, someone with 2.5K rating may be far off acheiving 3K rating statistically, but they could certainly be grinding enough to be close to a 3K player in terms of skill. That makes them a present danger to a high rating player's WHR variability, especially because it will take months for the ladder to actually reflect their growing skill. From my point of view: I WILL lose to someone in the top 30 occasionally, especially to someone like I have described above. And that's ok. What isn't ok is that position-wise we're maybe 20 apart but the rating gap could be 600 points apart. If it isn't obvious how badly these large gaps between ratings affect whr variability and the actual effectiveness of the system, I don't know how else to explain the issue. Edit: Feel free to tell me if I just don't know what I'm talking about.
+0 / -0
|
Could it be that the range at which reward/loss amount is calculated just isn't in keeping with the much larger range of ratings now? That it was suitable for a system where people were capping at 2k?
+0 / -0
|
@Sparkles: (might be just me) it is not very clear what you do not like. quote:
heart of my issue - which is why the WHR is so variant in the first place. |
What do you consider more relevant/important/informative: WHR or ladder position? In my view WHR is a number used to compute a skill presented as an order of players (the ladder position). So, I think it is of low interest on how that number varies, the interesting thing is the result of the computation: the ladder position. Is it a problem that we have the graph of WHR instead of a graph of ladder position for example? quote:
effectiveness of the system, |
What is your idea of an effective system? For me giving a percent change of winning that is similar to reality is an "effective" system. With 80% chance of winning means that you will win 4 out of 5 games. That can be measured, but seems to be rather correct for 1v1 MM. And it includes the case for you loosing games occasionally.
+0 / -0
|
In super short, I don't like how my experience with the ladder has gone from Very slow, to very fast, to very slow again changes.
+0 / -0
|
quote:
In super short, I don't like how my experience with the ladder has gone from Very slow, to very fast, to very slow again changes. |
The past can't be changed so yes, would have been better to have less trial and error but that's life. Do you think anything should be done now considering also the player can be the one that "evolves" slower or faster and he might not like the speed he evolves... ? (also, think that another change might make people unhappy with even more changes)
+0 / -0
|
The answer to that is real simple - I don't know and you can't please everyone no matter what you do. I do know that if the rating system stays the way that it is now, metaphorically speaking, the top 10 is going to be a grindy marshlike terrain to traverse whilst sub top 10 players periodically dampen your progress. You could just say "Yes, that's how it's meant to be now" and the conversation will be over. If there's value or truth in my "Expanding ladder" theory, it at least explains how we got here. Then Dein and development can decide whether or not they would like to change things or not.
+0 / -0
|
quote:
the top 10 is going to be a grindy marshlike terrain to traverse |
What do you consider grinding? Winning at the rate predicted, you should not progress at all! You can't "traverse top 10" just by playing more, you need to play "better". If you suddenly always beat the number 1, I think you will get fast to number 1. Or, if suddenly the number 1 looses constantly, he will go down fast.
+0 / -0
|
That's the thing malric, the top 4 don't play often so there is very little opportunity to rise. The people who I personally deem a threat to me end at the current #11 player, ATOSTIC. Beneath this, the chances of loss curve downwards with occasional exceptions. Keep in mind that #11 and below all grant me 1 point per win. I only got a few for beating Anarchid 3 times tonight. 1-2 points for beating UG. So that's my definition of a grind.
+0 / -0
|
imo the ladder is perfectly fine as-is. I'd recommend you stop caring about your ladder ELO if at all possible, just care about the outcome of individual games (or your MatchMaking Current Rating if you really need to care about some number to measure yourself).
+1 / -0
|
quote:
the top 4 don't play often so there is very little opportunity to rise. |
In fact you would like to be able "challenge" players with higher ranking than you, in order to be to prove you are better. The problem being now that if a player plays very rarely (hence, with low chance of playing against someone close in rank, but still able to keep the active status), someone below will have a hard time taking them over. A challenge system would need to be well thought (limit the number, limit how much higher you can challenge, find suitable time, forfeit if not present, etc.), but that would make you feel you have a chance. quote:
I only got a few for beating Anarchid 3 times tonight. 1-2 points for beating UG. |
That being said, system does not seem too bad currently. Godde, in the last month, MM, Singularity level, has 8 victories and 1 loss UltraGodzilla, in the last month, MM, Singularity level, has 15 victories and 2 loses. Anarchid, in the last month, MM, Singularity level, has 11 victories and 5 loses. Sparkles, in the last month, MM, Singularity level, has 62 victories and 37 loses. Now, without any WHR, and very approximate, your ranks seems correct (you have same rank as Anarchid currently). Regarding game number: 9 games (for Godde) seems a bit on the low side, but 17 games (for UltraGodzilla) sounds reasonable for me. Edit: and in fact Godde played 40 games (if not restricting to singularity), which again seems reasonable.
+0 / -0
|
@Sparkles it used to be the case that your rating could go down upon winning a game, but we convinced     DeinFreund that such a system is ridiculous. The underlying number that goes down when you win seems to still exist in some form, and it seems like the system is trying to drag your visible rating towards that number. To do so it takes a lot more points from you when you lose than you gain when you win. If I am correct, then the ladder is only technically implementing the request "don't make player's ratings go down when they win", while not implementing the spirit of the request. Only gaining a single point when you win (no matter who you play), while losing many points when you lose, also feels bad. quote:
The system is working astonishingly well. Your recent WHR estimated rating was all over the place, yet your ladder rating is stable. Meanwhile for newbies and others with large skill changes the rating changes quickly. So now we have the pros of the highly variable rating while still having a stable ladder.
|
I think the system should be evaluated based on whether it engenders a fun and active ladder.
+1 / -0
|
Based on what you're saying malric, I'm in a bubble of my own making. Hammering my way out of retirement rust during a new patch and a little bit of experimenting (consequently throwing games) against <T10 players has thrown my variance into full swing, seemingly causing the low reward factor. Based on finally understanding how it works enough to come to that conclusion, I'm going to hazard a guess and say that the only way to escape the low reward bubble is to win consistently against higher ranked players / make my other games more consistent. Kinda leads me back to half a problem; high ranked players aren't active and are a protected species on the ladder because of that.
+0 / -0
|
