Deimos.4263 Posted December 16, 2017 Share Posted December 16, 2017 > @"Rodzynald.5897" said: > I wish we got back to pre-HoT pvp where there were no "legendary" players running builds that *allowed* to be so incredibly efficient, and baddies had to git gud. A "classic" PvP mode with no elite specs at all would be fantastic. Link to comment Share on other sites More sharing options...
Abelisk.4527 Posted December 17, 2017 Share Posted December 17, 2017 > @"Deimos.4263" said: > > @"Rodzynald.5897" said: > > I wish we got back to pre-HoT pvp where there were no "legendary" players running builds that *allowed* to be so incredibly efficient, and baddies had to git gud. > A "classic" PvP mode with no elite specs at all would be fantastic. > Incoming 5 Core Thief comps Link to comment Share on other sites More sharing options...
zealex.9410 Posted December 17, 2017 Share Posted December 17, 2017 > @"DalinarKholin.6283" said: > I don't want to be rude but the top 5% of players desires for quick queue times should not be driving the bus for the majority. I know of no smart business that would lower the overall quality of there product because of the desires of the smallest group of customers. That being said surely the range could expand after a reasonable wait if queues got too intolerable for some people. But it should be an ironclad range for at least 3-5 minutes. The 5% is usually the party that also advertises an streams the game. Link to comment Share on other sites More sharing options...
Ledavi.1803 Posted December 17, 2017 Share Posted December 17, 2017 > @"Deimos.4263" said: > > @"Rodzynald.5897" said: > > I wish we got back to pre-HoT pvp where there were no "legendary" players running builds that *allowed* to be so incredibly efficient, and baddies had to git gud. > A "classic" PvP mode with no elite specs at all would be fantastic. > yeah, lets just take it down to 2 classes and no weapons, just fists and snowballs. Everyone loves lack of options, freedom from choice!! Link to comment Share on other sites More sharing options...
Bossun.2046 Posted December 19, 2017 Share Posted December 19, 2017 Hey Ben, did you notice any positive changes during this experiment? Link to comment Share on other sites More sharing options...
ArenaNet Staff Ben Phongluangtham.1065 Posted December 19, 2017 Author ArenaNet Staff Share Posted December 19, 2017 Hey folks, We promised some data from the experiment, so here it is! The long and short of it is that, as expected, queue times went up for everyone. People with skill rating 0-1000 and people with skill rating 1600+ were much more heavily affected. Numbers are % of people queuing. Sorry this is a bit hard to read. Tables... **Overall**| | :-----:|:-----:|:-----: Queue Time|0-5 Min|5+ Min Reduced Range|0.8748|0.1322 Old|0.965|0.0365 | | **0-1000**| | Queue Time|0-5 Min|5+ Min Reduced Range|0.6922|0.3078 Old|0.8427|0.1573 | | **1001-1400**| | Queue Time|0-5 Min|5+ Min Reduced Range|0.9103|0.0897 Old|0.9817|0.0183 | | **1401-1600**| | Queue Time|0-5 Min|5+ Min Reduced Range|0.8718|0.1282 Old|0.9745|0.0255 | | **1601+**| | Queue Time|0-5 Min|5+ Min Reduced Range|0.7538|0.2462 Old|0.8848|0.1152 Link to comment Share on other sites More sharing options...
Ghos.1326 Posted December 19, 2017 Share Posted December 19, 2017 > @"Ben Phongluangtham.1065" said: > Hey folks, > > We promised some data from the experiment, so here it is! > > The long and short of it is that, as expected, queue times went up for everyone. People with skill rating 0-1000 and people with skill rating 1600+ were much more heavily affected. > > Numbers are % of people queuing. Sorry this is a bit hard to read. Tables... > > **Overall**| | > :-----:|:-----:|:-----: > Queue Time|0-5 Min|5+ Min > Reduced |0.8748|0.1322 > Old|0.965|0.0365 > | | > **0-1000**| | > Queue Time|0-5 Min|5+ Min > Reduced|0.6922|0.3078 > Old|0.8427|0.1573 > | | > **1001-1400**| | > Queue Time|0-5 Min|5+ Min > Reduced|0.9103|0.0897 > Old|0.9817|0.0183 > | | > **1401-1600**| | > Queue Time|0-5 Min|5+ Min > Reduced|0.8718|0.1282 > Old|0.9745|0.0255 > | | > **1601+**| | > Queue Time|0-5 Min|5+ Min > Reduced|0.7538|0.2462 > Old|0.8848|0.1152 So, Ben, will this also mean that players won't be put into a queue anymore with people detrimentally lower than their own skill rating? I noticed improved queue times and improved quality of matches overall as well when I was playing on this day in unranked. Link to comment Share on other sites More sharing options...
ArenaNet Staff Ben Phongluangtham.1065 Posted December 19, 2017 Author ArenaNet Staff Share Posted December 19, 2017 > @"Ghos.1326" said: > > @"Ben Phongluangtham.1065" said: > > Hey folks, > > > > We promised some data from the experiment, so here it is! > > > > The long and short of it is that, as expected, queue times went up for everyone. People with skill rating 0-1000 and people with skill rating 1600+ were much more heavily affected. > > > > Numbers are % of people queuing. Sorry this is a bit hard to read. Tables... > > > > **Overall**| | > > :-----:|:-----:|:-----: > > Queue Time|0-5 Min|5+ Min > > Reduced |0.8748|0.1322 > > Old|0.965|0.0365 > > | | > > **0-1000**| | > > Queue Time|0-5 Min|5+ Min > > Reduced|0.6922|0.3078 > > Old|0.8427|0.1573 > > | | > > **1001-1400**| | > > Queue Time|0-5 Min|5+ Min > > Reduced|0.9103|0.0897 > > Old|0.9817|0.0183 > > | | > > **1401-1600**| | > > Queue Time|0-5 Min|5+ Min > > Reduced|0.8718|0.1282 > > Old|0.9745|0.0255 > > | | > > **1601+**| | > > Queue Time|0-5 Min|5+ Min > > Reduced|0.7538|0.2462 > > Old|0.8848|0.1152 > > So, Ben, will this also mean that players won't be put into a queue anymore with people detrimentally lower than their own skill rating? I noticed improved queue times and improved quality of matches overall as well when I was playing on this day in unranked. It's not turned on yet. It was just on unranked for that specific day. Overall queue times increased, but yours going down could have just been an outlier. We haven't made specific decisions yet on this part of the match making changes, but we'll let you all know when we do. Link to comment Share on other sites More sharing options...
saerni.2584 Posted December 19, 2017 Share Posted December 19, 2017 Did data collection include average difference between teams scores? Will we be able to get this data as well? Take 500-losing score (prevent outliers from Foefire) and compare before and after. Even if we go from 10% of games are competitive to 12% that is a 20% increase and worth taking for a 9% decrease in the under 5 minute queues. But, obviously we need to know what exactly the quality increase was objectively rather than just relying on a few random anecdotes. Link to comment Share on other sites More sharing options...
ArenaNet Staff Ben Phongluangtham.1065 Posted December 19, 2017 Author ArenaNet Staff Share Posted December 19, 2017 > @"saerni.2584" said: > Did data collection include average difference between teams scores? Will we be able to get this data as well? > > Take 500-losing score (prevent outliers from Foefire) and compare before and after. > > Even if we go from 10% of games are competitive to 12% that is a 20% increase and worth taking for a 9% decrease in the under 5 minute queues. > > But, obviously we need to know what exactly the quality increase was objectively rather than just relying on a few random anecdotes. We did pull that information. It was actually about the same. Less than 10 points difference. This isn't surprising, as we didn't make any additional MM changes besides tightening the range. Also, using match scores as a way to measure the effectiveness of a change is problematic. A lot of factors go into score differential besides relative skill of the teams. Things like "snowball" potential of the map, the tendency of people to give up after they are down a certain number, etc. Also, since this is unranked, people do a lot of things like play with classes they are learning or with experimental builds that cause them to play below their normal skill level. Link to comment Share on other sites More sharing options...
Bossun.2046 Posted December 19, 2017 Share Posted December 19, 2017 The 1601+ range doesn't seem as bad as I thought. Would you be able to apply a test like this in Ranked Off-season? As you said, people play in unranked with builds/classes that they aren't as familiar with, and thus the MMR differs there. Also, can you tell me my unranked MMR compared to my ranked one? Thanks! :) Link to comment Share on other sites More sharing options...
saerni.2584 Posted December 19, 2017 Share Posted December 19, 2017 > @"Ben Phongluangtham.1065" said: > > @"saerni.2584" said: > > Did data collection include average difference between teams scores? Will we be able to get this data as well? > > > > Take 500-losing score (prevent outliers from Foefire) and compare before and after. > > > > Even if we go from 10% of games are competitive to 12% that is a 20% increase and worth taking for a 9% decrease in the under 5 minute queues. > > > > But, obviously we need to know what exactly the quality increase was objectively rather than just relying on a few random anecdotes. > > We did pull that information. It was actually about the same. Less than 10 points difference. This isn't surprising, as we didn't make any additional MM changes besides tightening the range. Also, using match scores as a way to measure the effectiveness of a change is problematic. A lot of factors go into score differential besides relative skill of the teams. Things like "snowball" potential of the map, the tendency of people to give up after they are down a certain number, etc. > > Also, since this is unranked, people do a lot of things like play with classes they are learning or with experimental builds that cause them to play below their normal skill level. True, but I meant more on a map by map comparison. And sure, there are people messing around but all things being equal the odds are an increase in match quality will be reflected in final score spread especially on non-snowball maps. Link to comment Share on other sites More sharing options...
Kapax.3801 Posted December 19, 2017 Share Posted December 19, 2017 and what then defines if a person is bad or good in PvP? I guess I must be a loser because they always group me with losers ... if the opposing team was as mediocre as my team then it would be a long and maybe fun battle, but it seems that the system likes to make people feel more miserable by matching it with a team of "professionals". Link to comment Share on other sites More sharing options...
Ithilwen.1529 Posted December 20, 2017 Share Posted December 20, 2017 Initial Impression: Because of the way I've seen statistics and other data used on this forum, I distrust this. I think it's very clear that matchmaking needs to improve. It's also clear that low ranked players vs high ranked players will not improve things. Link to comment Share on other sites More sharing options...
Reikou.7068 Posted December 20, 2017 Share Posted December 20, 2017 > @"Ben Phongluangtham.1065" said: > Hey folks, > > We promised some data from the experiment, so here it is! > > The long and short of it is that, as expected, queue times went up for everyone. People with skill rating 0-1000 and people with skill rating 1600+ were much more heavily affected. > > Numbers are % of people queuing. Sorry this is a bit hard to read. Tables... > > **Overall**| | > :-----:|:-----:|:-----: > Queue Time|0-5 Min|5+ Min > Reduced Range|0.8748|0.1322 > Old|0.965|0.0365 > | | > **0-1000**| | > Queue Time|0-5 Min|5+ Min > Reduced Range|0.6922|0.3078 > Old|0.8427|0.1573 > | | > **1001-1400**| | > Queue Time|0-5 Min|5+ Min > Reduced Range|0.9103|0.0897 > Old|0.9817|0.0183 > | | > **1401-1600**| | > Queue Time|0-5 Min|5+ Min > Reduced Range|0.8718|0.1282 > Old|0.9745|0.0255 > | | > **1601+**| | > Queue Time|0-5 Min|5+ Min > Reduced Range|0.7538|0.2462 > Old|0.8848|0.1152 Can you share with us what the specific search-range restrictions were that were tested? Were they hard limits of +/-150 MMR or something? Or is it something thats just "tighter at the start, and still expands out until infinity eventually," because that honestly does nothing for the current issue. Link to comment Share on other sites More sharing options...
Ithilwen.1529 Posted December 20, 2017 Share Posted December 20, 2017 That looks like a success to me. I am well and truly sick of matches that I know I will lose going in. I'd vastly prefer to wait 5 minutes or more for a good match. Link to comment Share on other sites More sharing options...
ArenaNet Staff Ben Phongluangtham.1065 Posted December 20, 2017 Author ArenaNet Staff Share Posted December 20, 2017 > @"Reikou.7068" said: > > @"Ben Phongluangtham.1065" said: > > Hey folks, > > > > We promised some data from the experiment, so here it is! > > > > The long and short of it is that, as expected, queue times went up for everyone. People with skill rating 0-1000 and people with skill rating 1600+ were much more heavily affected. > > > > Numbers are % of people queuing. Sorry this is a bit hard to read. Tables... > > > > **Overall**| | > > :-----:|:-----:|:-----: > > Queue Time|0-5 Min|5+ Min > > Reduced Range|0.8748|0.1322 > > Old|0.965|0.0365 > > | | > > **0-1000**| | > > Queue Time|0-5 Min|5+ Min > > Reduced Range|0.6922|0.3078 > > Old|0.8427|0.1573 > > | | > > **1001-1400**| | > > Queue Time|0-5 Min|5+ Min > > Reduced Range|0.9103|0.0897 > > Old|0.9817|0.0183 > > | | > > **1401-1600**| | > > Queue Time|0-5 Min|5+ Min > > Reduced Range|0.8718|0.1282 > > Old|0.9745|0.0255 > > | | > > **1601+**| | > > Queue Time|0-5 Min|5+ Min > > Reduced Range|0.7538|0.2462 > > Old|0.8848|0.1152 > > Can you share with us what the specific search-range restrictions were that were tested? > > Were they hard limits of +/-150 MMR or something? Or is it something thats just "tighter at the start, and still expands out until infinity eventually," because that honestly does nothing for the current issue. It still had the expansion over time, but we capped the +/- range at a much tighter amount. We plan to do a another experiment in the future, probably after Christmas where we leave the range as it currently is on live and adjust the expansion time. Link to comment Share on other sites More sharing options...
Reikou.7068 Posted December 21, 2017 Share Posted December 21, 2017 > @"Ben Phongluangtham.1065" said: > > @"Reikou.7068" said: > > > @"Ben Phongluangtham.1065" said: > > > Hey folks, > > > > > > We promised some data from the experiment, so here it is! > > > > > > The long and short of it is that, as expected, queue times went up for everyone. People with skill rating 0-1000 and people with skill rating 1600+ were much more heavily affected. > > > > > > Numbers are % of people queuing. Sorry this is a bit hard to read. Tables... > > > > > > **Overall**| | > > > :-----:|:-----:|:-----: > > > Queue Time|0-5 Min|5+ Min > > > Reduced Range|0.8748|0.1322 > > > Old|0.965|0.0365 > > > | | > > > **0-1000**| | > > > Queue Time|0-5 Min|5+ Min > > > Reduced Range|0.6922|0.3078 > > > Old|0.8427|0.1573 > > > | | > > > **1001-1400**| | > > > Queue Time|0-5 Min|5+ Min > > > Reduced Range|0.9103|0.0897 > > > Old|0.9817|0.0183 > > > | | > > > **1401-1600**| | > > > Queue Time|0-5 Min|5+ Min > > > Reduced Range|0.8718|0.1282 > > > Old|0.9745|0.0255 > > > | | > > > **1601+**| | > > > Queue Time|0-5 Min|5+ Min > > > Reduced Range|0.7538|0.2462 > > > Old|0.8848|0.1152 > > > > Can you share with us what the specific search-range restrictions were that were tested? > > > > Were they hard limits of +/-150 MMR or something? Or is it something thats just "tighter at the start, and still expands out until infinity eventually," because that honestly does nothing for the current issue. > > It still had the expansion over time, but we capped the +/- range at a much tighter amount. We plan to do a another experiment in the future, probably after Christmas where we leave the range as it currently is on live and adjust the expansion time. Thanks for the reply. Can you share the maximum +/- range that was tested this time? I believe that capping the +/- range to something like Divisions would be the best way forward. That way it actually gives meaning to divisions. Where bronze players only play with/against other bronzes, Silvers with Silvers, Gold with gold, Plat and legend with plat and legends. And while hard limits on match making searches may unfortunately lend to some hours being completely unplayable at the extreme-end of MMRs due to simple lack of population, it would go a LONG way in improving matchmaking quality. Link to comment Share on other sites More sharing options...
Elegie.3620 Posted December 21, 2017 Share Posted December 21, 2017 Hello, > @"Ben Phongluangtham.1065" said: > Also, since this is unranked, people do a lot of things like play with classes they are learning or with experimental builds that cause them to play below their normal skill level. Careful here, the very concept of "normal" skill level is flawed. Skill level does include unvarying elements such as strategic or tactical insight (rotations, map awareness, fight management), moderately varying elements such as mood or shape, and strongly varying elements such as class / build knowledge (mechanical skill of the class / build being played). I understand you cannot include class MMR in the matchmaking because players can change classes before or during games, but don't forget that it still remains one of the primary reason for unbalanced matches. Also, I wanted to thank you for your presence and your communicating with the community. I don't play PvP anymore because reasons (verbal violence and toxicity, too many bunker builds, too much conquest, not enough playing), but I really appreciate your regular input, and wish you the best in your experiences. Link to comment Share on other sites More sharing options...
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now