Because here I am PVPing solo. I queued randomly with three other very well geared players (all of us were in the top 10 pages on the Leaderboard, I believe). Our fifth player was...
Seriously? We're all level 60!
I agree with others, most likely a bot, or maybe someone that is rebelling and knows that is going to really urk someone.
I think we have discovered that the PVP player base is not up to the standards of an ELO system. I hear it works but its set up to allow for fast queues, ergo if not enough people of the appropriate rating are on then you get whatever is available. At times i'm ok with this, since well its like opening a dungeon chest, sometimes I get what I want, sometimes I don't.
Thanks to degraafination, futuun, Zeb, and everyone for pointing this out. However, this is not actually a bug. When a player with a high certainty rank plays in a match where the rank spread is high and happens to win, the uncertainty of that player’s rank can actually increase. An increase in uncertainty actually reduces rank. This occurs because our algorithm mixes in a slight amount of uncertainty of a player’s rank every time they play a match. This is what allows players to move quickly up the ranks when they go on a winning streak.
There are solutions to this that are mostly cosmetic and other solutions that will also affect the algorithm and affect the ability for players to “move the needle” when they go on a winning streak. We are considering some of those changes at this time.
However, the quality of the matchmaking is not likely to be affected by this phenomenon. Matchmaking quality is highly dependent on the queue population at the time of the match.
I'm a mathematician, and the problem tickled my mind, I wondering if you are allowed to tell me (I know information about proprietary code is very hard to clear with the bosses for even generalized release of information about it) if the tracking system for the matchmaking code has a lowish hard cap on the number data points it can hold per character before passing the information to the algorithm (software capped), or is the variable cap one of the practicality of processing power (hardware capped)?
I have no trouble with pvp matches having both 8kgs and 15kgs people, the problem is that the system tends to put the 15k gs on one side and the 8kgs on the other.
If they would be placed evenly then it would be much better.
Basically, if you win - you gain "some" ELO points depending on the ELO rating of your competition -- if you are way above them, you only gain a tiny bit -- if you are below them you gain a great deal.
This is False, The ELO system in NW, has no real value, Believe me.... Queing with 5/5 18K GS players (top 10 pages) and we go against 5/5 8-10K players.,.... No im sorry but there is no value or math or method to how the Leaderboard/ELO works in Neverwinter, Along with many other problems in PvP. HR's are OP obviously.... Best class in the game Currently, Heavily Broken, and WAY more powerful then GWF's.
But ya ELO system has no math or method behind it, If u get lucky and u get matched with equal players, sure....you might have a chance to move up a few spots.....
The Problem is...You never get matched into a Feasible match...... lmao what a joke.... Mod 3 was the biggest Fail in Neverwinter's history imo.
0
Options
degraafinationMember, Neverwinter Beta UsersPosts: 0Arc User
I've noticed in the past few days unless you win against someone higher rank then you, your probably not going to advance in rank
In addition, when you win against someone significantly lower than you, you'll actually drop in rank (sometimes significantly).
The entire Leaderboard concept is a terrible idea. I say get rid of it. PVP was so much more enjoyable when everyone wasn't so concerned about their ranking.
I have no trouble with pvp matches having both 8kgs and 15kgs people, the problem is that the system tends to put the 15k gs on one side and the 8kgs on the other.
If they would be placed evenly then it would be much better.
That is actually pretty sensible and I agree. was in a match this morning with my CW , where I was 12 and 0 and I was fighting myself 2-3 at a time.. again on my CW.. flip a switch next match, my 15k with a 16k gf and then nothing, 7 and 8 kers for the other two.. where the other team completely trounced them over and over. So instead of putting one 7k gs on each team, I do not know why they seem to stack one team with the majority of them vs the other.
It's all a matter of supply and demand. The ELO has nothing to do with that. It would match you up properly if there was such an opponent available in the pool. In short, by being 20k BiS, you made yourself an elite. And elites will more often face the weak than someone who's equal to them.
That's just an excuse. It seriously does not work. It may loosely be how it works, but the system is terrible and anyone who can't acknowledge that doesn't know what they're talking about or is delusional. Right now the ranked queue is only slightly more efficient than a normal queue but still worse than pretty much every other decent games implementation of ranked play.
The entire Leaderboard concept is a terrible idea. I say get rid of it. PVP was so much more enjoyable when everyone wasn't so concerned about their ranking.
Yup, I would like to see it gone as well.
0
Options
cbrowne0329Member, NW M9 PlaytestPosts: 293Arc User
edited July 2014
ELO probably is working. There might be an easy explanation for why your ranking may go down when winning. Check the people above and below you and screenshot it (1 or two pages both directions). That way you can keep track of wins / kills etc. Play a few matches, and check it again. Compare numbers.
If someone below you got a lot of kills / wins in the same time frame you did, against tougher opponents because of DC or whatever, they are likely to rise above you.
I have a couple of suggestions.
1) Create another PVP queue that allows people to do choice match-ups... No AD rewards gained from this queue.
2) For this new queue, make it so you can choose ranked or non-ranked matches. Ranked matches have to be against opposing guilds/pugs - non-ranked against anyone (so you can practice and stomp guildies for fun).
Don't have to worry about adding tons of code for ELO, or gear score. Let the true PVP players decide who they will face, GS be damned. This will bring more people to the game for PVP. At least I strongly believe that.
Based on previous posts it looks like ELO ratings range from about 22 - roughly 50. So I will use these as a metric for example. This is NOT exact science as I am simplifying WHAT happens and not HOW it happens.
Team 1 consists of (ELO ratings)
Player 1: 46.243
Player 2: 45.67
Player 3: 31.89
Player 4: 30.34
Player 5: 32.56
Team 2:
Player 1: 36.243
Player 2: 35.67
Player 3: 37.89
Player 4: 39.34
Player 5: 38.56
Now on paper this may look "fair". Ok team 1 has 2 obviously very high players but has 3 "lower" players. Where team 2 looks to be more "average"
The issue is it doesnt factor into account ALOT of things. How any one of those players could be SUPER crappy GS. How one of those players could be horrendous and lose the match for them. Ontop of that, how apparently the "confidence" interval can INCREASE due to the imbalance of the higher ELO players causing them to LOSE ELO rating.
That 39.34 player on Team 2 could be an 18k GS GWF who may not be very good - but can still hit the 30.34 ELO CW in full PVE gear with no tenacity and 25k HP....
So this GWF almsot 1 shots this guy who cries foul on the forums. Or the two high ELO guys on Team 1 carry the team (after a painful match of their PUGS going 0-34 K/D and not standing on nodes) to vitory and LOSING ELO because of their confidence interval.
DEVS:
What you NEED to do is "band" ELO like you do with levels. So if your over say ELO of 40, you ONLY have a "pool" of players to que with that are over 40. I dont know the total spread, but maybe its as simple as:
20-30
30-40
40+
Essentially you create "bands" that separate the "elite (40+)" from the average (30-40) from the casual "20-30".
"What if an ELO rating 25 pairs with a 40+ though?"
Simple!
It will be based on the HIGHEST person in the group. So it "shops" in the 40+ range for players.
This MAY create long wait times for players on the "elite" end. For the majority of players this wont make a huge difference, but what this WILL do is create more "fair" matches.
The entire Leaderboard concept is a terrible idea. I say get rid of it. PVP was so much more enjoyable when everyone wasn't so concerned about their ranking.
You have a point. Every single person says they don't "care" about the leaderboard because it is broken, yet they cry everytime something happens to their ranking.
I think cryptic wanted to implement a leaderboard to make PVPing seem to have an end goal.
Well at this point, what the dev explained why players can drop rank for winning matches, actually matches at least what I have seen. It means it could be true.
However, having an uncertainty variable that increases on the lower end, means the entire system is a less accurate system than if winning did not increase uncertainty on the lower end. It doesn't make sense to increase the uncertainty on the lower end for winning or on the upper end for losing.
It also means that for players to move down in rank on the leaderboards, their rank is represented on the leaderboards based in some fashion on their uncertainty range. My guess would be either the median or lower end of the range is a players elo and how the are represented in the ranks, but it's only my guess. But it would explain how a player's displayed rank can drop after a win if their uncertainty range increases on the lower end.
Basically the system is no good. It creates poor estimates of players' rankings and at any given time it is almost certain that a good number of players' ranking will not accurately reflect their true ranking, regardless of how many matches happen, even if no new players ever enter the system.
Now, it's nice to finally see a dev post about their elo system that actually matches what players are experiencing in game. However, because of the small player pool and despite what the dev said, it actually does either have a significant effect on the leaderboards because it means that player can drop rank much faster than they should for losing. For example, if a player wins match A and his uncertainty increases on the lower end and then he loses match B, that player will lose more ranking than if they only played match B. Again, it means the system is no good. The system poorly estimates players true skill level.
Add to this that it's quite discouraging for players to lose rank for winning and again, it means the system is no good.
They should have never introduced a leaderboard system without good reason and to compound the issue their current elo system doesn't work that well as it doesn't accurately reflect many ranks.
With a combined lack of competitive community or competitive ranked PVP their is no reason to have leaderboards because their negatives outweigh the positives. The system we have now is only with a poorly designed ranked system with individual elo ranking that doesn't accurately reflect players' true skill level and no casual PVP and still no competitive ranked system or community.
It makes be wonder if somebody thought it would be a good idea to invest money, time, and hours into developing a ranked leaderboard system based on the elo system. If makes be wonder if the people designing this game are competent
If there's nothing better to oppose them I'm pretty sure (semi)premade goes first, based on personal experience.
I'd guess that would depend on waiting time and availability. If both premades and 6 low GS people all queued together at the same time, I'm pretty sure the solo players will be split between the premades at random and then put against eachother. The same would probably happen if the solo people were all 19k or 12 and 19. And it's the randomness that gets to people. Again, personal experience. I can't prove a thing as I don't have the mathematics they use.
One thing is for certain, while I'm confident the ELO itself does what it can, the system lacks intelligence, specifically about party makeup. In the example above, the 19k solo player has a 50% chance, minus or plus his uncertainty score, to join either the weak or the strong party. When dealing with extreme differences like this the system should override the uncertainty score and force the 19k on the weak team and the 12k on the strong team. I don't think it does that right now.
If what you said is right then this ELO system is certainly not working. I did an experiment with my brother and another friend. We queued individually at the SAME time. Each of us got paired up with 3 17k+ gs premade teams who apparently hit with perfect accuracy. It is quite obvious that if premades are matched with premades first then at least 2 of the 3 premade teams should have faced another premade in the queue instead of picking up 5 individual randomly solo queued players.
On our second trials, we queued solo the same way. This time my brother and his friend were put in the same group with 3 other random noobs. I was put in another teams of randoms. Each of our teams got 2 of the 3 premades teams in our first trial. Again, the 2 premades did not see each other. Instead, they picked up individual noobs in the queue as their opponents.
Suggestion: if there is no brackets seperating low and high elo players implement a GS minimum like 10k+ for pvp to weed out the bots and AD farmers. That way the people who actually que for pvp get what they came for. Majority of the matches are spent spawn camping because people are just there to farm AD and have really bad gear.
Instead of fixing it, now you hide the only way that we could prove your system is broken?!?
WHY?
It is insanely counterproductive what you guys are doing. Please stop hiding the real mechanics.
The system isn't broken. It's not working the way you or others would like it to, but it's working the way it was intended to work. if you don't like it, stop looking at the leaderboard. It's completely meaningless anyway.
Personally, I'd like to see a rating much more than a ranking. And they apparently just removed our access to seeing that. The leaderboard itself is fairly meaningless as it doesn't tell anything about how far apart the players are in terms of ability to win against each other. A rating does do that. Just looking at the leaderboard, you can't tell if a player from page 1 is a little better or a lot better than someone from page 10, page 100, or page 1,000, just that they are *some* degree of better. A rating would tell you more.
It doesn't matter if you see the rating or the ranking. Both are completely meaningless because they are an INDIVIDUAL score based on the result of a TEAM game.
Comments
CRAZY MIKE - 14.6k PvE CW
Backbone - 16.7k PvP HR
[SIGPIC]http://i59.tinypic.com/s3hts7.png[/SIGPIC]
I agree with others, most likely a bot, or maybe someone that is rebelling and knows that is going to really urk someone.
I think we have discovered that the PVP player base is not up to the standards of an ELO system. I hear it works but its set up to allow for fast queues, ergo if not enough people of the appropriate rating are on then you get whatever is available. At times i'm ok with this, since well its like opening a dungeon chest, sometimes I get what I want, sometimes I don't.
I'm a mathematician, and the problem tickled my mind, I wondering if you are allowed to tell me (I know information about proprietary code is very hard to clear with the bosses for even generalized release of information about it) if the tracking system for the matchmaking code has a lowish hard cap on the number data points it can hold per character before passing the information to the algorithm (software capped), or is the variable cap one of the practicality of processing power (hardware capped)?
If they would be placed evenly then it would be much better.
Your answer is NO it does NOT work. and anyone who disagrees is oblivious to any facts.
This is False, The ELO system in NW, has no real value, Believe me.... Queing with 5/5 18K GS players (top 10 pages) and we go against 5/5 8-10K players.,.... No im sorry but there is no value or math or method to how the Leaderboard/ELO works in Neverwinter, Along with many other problems in PvP. HR's are OP obviously.... Best class in the game Currently, Heavily Broken, and WAY more powerful then GWF's.
But ya ELO system has no math or method behind it, If u get lucky and u get matched with equal players, sure....you might have a chance to move up a few spots.....
The Problem is...You never get matched into a Feasible match...... lmao what a joke.... Mod 3 was the biggest Fail in Neverwinter's history imo.
In addition, when you win against someone significantly lower than you, you'll actually drop in rank (sometimes significantly).
The entire Leaderboard concept is a terrible idea. I say get rid of it. PVP was so much more enjoyable when everyone wasn't so concerned about their ranking.
Join Essence of Aggression: PVP-ing Hard Since Beta!
That is actually pretty sensible and I agree. was in a match this morning with my CW , where I was 12 and 0 and I was fighting myself 2-3 at a time.. again on my CW.. flip a switch next match, my 15k with a 16k gf and then nothing, 7 and 8 kers for the other two.. where the other team completely trounced them over and over. So instead of putting one 7k gs on each team, I do not know why they seem to stack one team with the majority of them vs the other.
No. We are not.
Really guys?
Instead of fixing it, now you hide the only way that we could prove your system is broken?!?
WHY?
It is insanely counterproductive what you guys are doing. Please stop hiding the real mechanics.
Yup, I would like to see it gone as well.
If someone below you got a lot of kills / wins in the same time frame you did, against tougher opponents because of DC or whatever, they are likely to rise above you.
I have a couple of suggestions.
1) Create another PVP queue that allows people to do choice match-ups... No AD rewards gained from this queue.
2) For this new queue, make it so you can choose ranked or non-ranked matches. Ranked matches have to be against opposing guilds/pugs - non-ranked against anyone (so you can practice and stomp guildies for fun).
Don't have to worry about adding tons of code for ELO, or gear score. Let the true PVP players decide who they will face, GS be damned. This will bring more people to the game for PVP. At least I strongly believe that.
Based on previous posts it looks like ELO ratings range from about 22 - roughly 50. So I will use these as a metric for example. This is NOT exact science as I am simplifying WHAT happens and not HOW it happens.
Team 1 consists of (ELO ratings)
Player 1: 46.243
Player 2: 45.67
Player 3: 31.89
Player 4: 30.34
Player 5: 32.56
Team 2:
Player 1: 36.243
Player 2: 35.67
Player 3: 37.89
Player 4: 39.34
Player 5: 38.56
Now on paper this may look "fair". Ok team 1 has 2 obviously very high players but has 3 "lower" players. Where team 2 looks to be more "average"
The issue is it doesnt factor into account ALOT of things. How any one of those players could be SUPER crappy GS. How one of those players could be horrendous and lose the match for them. Ontop of that, how apparently the "confidence" interval can INCREASE due to the imbalance of the higher ELO players causing them to LOSE ELO rating.
That 39.34 player on Team 2 could be an 18k GS GWF who may not be very good - but can still hit the 30.34 ELO CW in full PVE gear with no tenacity and 25k HP....
So this GWF almsot 1 shots this guy who cries foul on the forums. Or the two high ELO guys on Team 1 carry the team (after a painful match of their PUGS going 0-34 K/D and not standing on nodes) to vitory and LOSING ELO because of their confidence interval.
DEVS:
What you NEED to do is "band" ELO like you do with levels. So if your over say ELO of 40, you ONLY have a "pool" of players to que with that are over 40. I dont know the total spread, but maybe its as simple as:
20-30
30-40
40+
Essentially you create "bands" that separate the "elite (40+)" from the average (30-40) from the casual "20-30".
"What if an ELO rating 25 pairs with a 40+ though?"
Simple!
It will be based on the HIGHEST person in the group. So it "shops" in the 40+ range for players.
This MAY create long wait times for players on the "elite" end. For the majority of players this wont make a huge difference, but what this WILL do is create more "fair" matches.
You have a point. Every single person says they don't "care" about the leaderboard because it is broken, yet they cry everytime something happens to their ranking.
I think cryptic wanted to implement a leaderboard to make PVPing seem to have an end goal.
However, having an uncertainty variable that increases on the lower end, means the entire system is a less accurate system than if winning did not increase uncertainty on the lower end. It doesn't make sense to increase the uncertainty on the lower end for winning or on the upper end for losing.
It also means that for players to move down in rank on the leaderboards, their rank is represented on the leaderboards based in some fashion on their uncertainty range. My guess would be either the median or lower end of the range is a players elo and how the are represented in the ranks, but it's only my guess. But it would explain how a player's displayed rank can drop after a win if their uncertainty range increases on the lower end.
Basically the system is no good. It creates poor estimates of players' rankings and at any given time it is almost certain that a good number of players' ranking will not accurately reflect their true ranking, regardless of how many matches happen, even if no new players ever enter the system.
Now, it's nice to finally see a dev post about their elo system that actually matches what players are experiencing in game. However, because of the small player pool and despite what the dev said, it actually does either have a significant effect on the leaderboards because it means that player can drop rank much faster than they should for losing. For example, if a player wins match A and his uncertainty increases on the lower end and then he loses match B, that player will lose more ranking than if they only played match B. Again, it means the system is no good. The system poorly estimates players true skill level.
Add to this that it's quite discouraging for players to lose rank for winning and again, it means the system is no good.
They should have never introduced a leaderboard system without good reason and to compound the issue their current elo system doesn't work that well as it doesn't accurately reflect many ranks.
With a combined lack of competitive community or competitive ranked PVP their is no reason to have leaderboards because their negatives outweigh the positives. The system we have now is only with a poorly designed ranked system with individual elo ranking that doesn't accurately reflect players' true skill level and no casual PVP and still no competitive ranked system or community.
It makes be wonder if somebody thought it would be a good idea to invest money, time, and hours into developing a ranked leaderboard system based on the elo system. If makes be wonder if the people designing this game are competent
If what you said is right then this ELO system is certainly not working. I did an experiment with my brother and another friend. We queued individually at the SAME time. Each of us got paired up with 3 17k+ gs premade teams who apparently hit with perfect accuracy. It is quite obvious that if premades are matched with premades first then at least 2 of the 3 premade teams should have faced another premade in the queue instead of picking up 5 individual randomly solo queued players.
On our second trials, we queued solo the same way. This time my brother and his friend were put in the same group with 3 other random noobs. I was put in another teams of randoms. Each of our teams got 2 of the 3 premades teams in our first trial. Again, the 2 premades did not see each other. Instead, they picked up individual noobs in the queue as their opponents.
Aelar Hawkwind - Archer
Karrin Feywinter - Mistress of Flame
Errin Duskwalker - Executioner
Darquess - Soulbinder
Aelar Hawkwind - Archer
Karrin Feywinter - Mistress of Flame
Errin Duskwalker - Executioner
Darquess - Soulbinder