Guild Wars 2

Nice analysis this is exactly what is needed and I agree with your conclusions. It is crystal clear that the random number range is too wide.

JonPeters.5630:

This is excellent and would be true if our rating currently had significant meaning. Once this new randomization is in place for a while we expect both ratings and deviation will stabilize a bit more. If at that point our simulation is still generating bad match-ups we can absolutely look at lowering the range.

Thanks for commenting! I only worked on what data was given. When you say they have no meaning, is this because you are resetting the ratings, or because the dev’s are expecting that the match ups will “scramble” the rankings a bit and let them be more stable? It is also unclear how the FINAL ratings, after these random match ups, will be calculated.

I also want to note that the simulations aren’t doing 5 million match ups, one after another assuming a winner or something. Just that given the current ratings and deviations, what the statistically possible match ups could be, and what their probabilities are. I could put in any initial ratings and deviations for the server and produce the statistical probabilities of potential match ups!

JonPeters.5630:

Yes, we expect the new match ups to help stabilize the ratings/deviation.

Can you give some clarification on how the final ratings will be calculated after each of these matches? Will they be the matchmaker ratings, or using the glicko ratings for those match ups?

JonPeters.5630:

Glicko will still be used to calculate ratings, we will simply properly be using deviation to match people up because glicko matchup was not accounting for the the fact that all matches were simultaneously created, and simply matching the same teams every single time.

Thank you for your speedy replies for clarification!

JonPeters.5630:

This is excellent and would be true if our rating currently had significant meaning. Once this new randomization is in place for a while we expect both ratings and deviation will stabilize a bit more. If at that point our simulation is still generating bad match-ups we can absolutely look at lowering the range.

I disagree that any random variation system would ever stabilize any system.
The new system will not bring more stability but variation, as a positive will allow imploding servers to fall faster through ranks and for the sake of randomness and variation a negative allowing uneven match-ups.
There is cause and effect, statistics only reflect the system and not fix it.

This image here:

JonPeters.5630:

Yes, we expect the new match ups to help stabilize the ratings/deviation.

To help people understand better, would it be possible for you guys to put somewhere what your simulations/programs are showing should be the expected ratings after they stabilize? Since you said yourself that if you base if off the current ratings, the predictions don’t mean much, but that is all you have showed us: the possible match ups with the current ratings. And, if they stabilize, about after how many matches are you guys expecting? Since they run for a week, even if they stabilize after 5 or 6 matches, thats over a month for us to play in.

JonPeters, the problem is not the matchmaking, but the score system itself. Fix the scoring system so that the server with most coverage doesn’t automatically win.

Az z.2746:

JonPeters, the problem is not the matchmaking, but the score system itself. Fix the scoring system so that the server with most coverage doesn’t automatically win.

I have a feeling what they’re hoping for is that if say, a rank 1 server faces a rank 4, and the rank 4 server does well, they’ll get a decent boost, much more than they would get facing the rank 5 and 6 servers in their own tier. However, I am worried this could also punish the rank 1 server more if they don’t win “by enough”, and not give the rank 4 server enough gains. If they still lose by a lot, they might lose rank, instead of gaining it. Meanwhile, if the rank 5 server faces a rank 7 or 8 and dominates them, they could move up.

Az z.2746:

JonPeters, the problem is not the matchmaking, but the score system itself. Fix the scoring system so that the server with most coverage doesn’t automatically win.

Devon has already hinted at some changes in this regard.

..though I didn’t get the impression it will be an entire overhaul, which is really what’s needed. Complete redesign.

Handin.4032:

kitten .2746:

JonPeters, the problem is not the matchmaking, but the score system itself. Fix the scoring system so that the server with most coverage doesn’t automatically win.

I have a feeling what they’re hoping for is that if say, a rank 1 server faces a rank 4, and the rank 4 server does well, they’ll get a decent boost, much more than they would get facing the rank 5 and 6 servers in their own tier. However, I am worried this could also punish the rank 1 server more if they don’t win “by enough”, and not give the rank 4 server enough gains. If they still lose by a lot, they might lose rank, instead of gaining it. Meanwhile, if the rank 5 server faces a rank 7 or 8 and dominates them, they could move up.

When a server is demoralized they don’t even show for the matchup anymore, this happened countless times already.

Az z.2746:

Handin.4032:

kitten .2746:

JonPeters, the problem is not the matchmaking, but the score system itself. Fix the scoring system so that the server with most coverage doesn’t automatically win.

I have a feeling what they’re hoping for is that if say, a rank 1 server faces a rank 4, and the rank 4 server does well, they’ll get a decent boost, much more than they would get facing the rank 5 and 6 servers in their own tier. However, I am worried this could also punish the rank 1 server more if they don’t win “by enough”, and not give the rank 4 server enough gains. If they still lose by a lot, they might lose rank, instead of gaining it. Meanwhile, if the rank 5 server faces a rank 7 or 8 and dominates them, they could move up.

When a server is demoralized they don’t even show for the matchup anymore, this happened countless times already.

Yup! I totally understand that that’s why I think their system parameters need to be tweaked a little to allow for less spread on a match by match basis, at least initially.

Az z.2746:

JonPeters, the problem is not the matchmaking, but the score system itself. Fix the scoring system so that the server with most coverage doesn’t automatically win.

I agree that score needs to be addressed too, I have mentioned a suggestion in previous thread that after you get more data you might consider applying a handicap. (all servers are no equal so its not a scratch game) By doing so camps, towers and keeps are worth more based on the numbers available by a particular server at a particular time.

As with most handicap leagues you need to establish a baseline by playing everyone first and then it gets adjusted throughout the season.

vanzan.1250:

I agree that score needs to be addressed too, I have mentioned a suggestion in previous thread that after you get more data you might consider applying a handicap. (all servers are no equal so its not a scratch game) By doing so camps, towers and keeps are worth more based on the numbers available by a particular server at a particular time.

As with most handicap leagues you need to establish a baseline by playing everyone first and then it gets adjusted throughout the season.

No one wants their play time to be worth less than someone else’s, end of story. SEAS/Asia, have no personal servers and (almost always) play on NA because of how data centers are localized for lowest ping. You’re going to punish these people for playing because the other servers didn’t get as many such players?

And what happens when your system is gamed hardcore by blackouts because a server maximizes a lead initially and then just stops showing up except for small defense forces, devaluing all possible point gain?

Zen.8497:

I disagree that any random variation system would ever stabilize any system.
The new system will not bring more stability but variation, as a positive will allow imploding servers to fall faster through ranks and for the sake of randomness and variation a negative allowing uneven match-ups.
There is cause and effect, statistics only reflect the system and not fix it.

The RNG isn’t what’s going to stabilize the system, specifically. The RNG adds more sample points on which the rating can be computed and averaged. The current system has, at any given time, literally 3 data points per tier on which to build its ratings which leads to servers being literally stuck in rating for months on end because there is no chaotic deviation of any sort or adequate data accumulation on which real data can be generated.

Look back to the SoS/HoD/etc collapses, the system had nothing but old data and few incoming data points with little deviation between them on which to “rate” the dead servers. So they remained in T1/T2 for months while barely even having a standing population on a single map. If you add a chaotic deviation such that they can be literally thrown in against an opponent two tiers below and still get demolished, their rating would then tank post haste because the rating system knows that they cant even compete against low ranked servers. Likewise on the climb, the same would be true.

As more data is gathered, the chaotic deviations will lessen because there is more data to outweigh the randomness. This is, all in all, pretty standard.

TOO MUCH NUMBERS ARRGH!

But good job haha I think the random number there has too big an impact to the overall base rating. But what do I know

Vena.8436:

The RNG isn’t what’s going to stabilize the system, specifically. The RNG adds more sample points on which the rating can be computed and averaged. The current system has, at any given time, literally 3 data points per tier on which to build its ratings which leads to servers being literally stuck in rating for months on end because there is no chaotic deviation of any sort or adequate data accumulation on which real data can be generated.

Look back to the SoS/HoD/etc collapses, the system had nothing but old data and few incoming data points with little deviation between them on which to “rate” the dead servers. So they remained in T1/T2 for months while barely even having a standing population on a single map. If you add a chaotic deviation such that they can be literally thrown in against an opponent two tiers below and still get demolished, their rating would then tank post haste because the rating system knows that they cant even compete against low ranked servers. Likewise on the climb, the same would be true.

As more data is gathered, the chaotic deviations will lessen because there is more data to outweigh the randomness. This is, all in all, pretty standard.

Sorry, but you and more importantly ANet do not account the human factor in their calculations. Let me explain shortly.

By now, the server population knows which server fields which guilds and how their coverage is. If you take a tier 4 server like mine and put it with this wide spread against Vizunah, people are just going to quit WvW after 2 days. Why? Because all the progress made during the day is going to be negated by the opposing server’s night crew.

Now, this would happen for a week and the new system is going to pick new opponents. BUT, the minute the new system picks up Gandara and Vizunah again, players are not going to play from the first day on – thus lowering the score of their server significantly. This will happen every single time a server faces another server knowingly that they will loose anyway. This is how people think and this is how people play.

And why is it so? Because there is no other reward right now other than fun. There are no shiny epix, no gold, no nothing – only fun. It is not fun starting a match and knowing you are going to loose anyway.

The problem isn’t with the rating system, it’s with the scoring system.

The winners are determined by coverage and numbers, combining the EU and NA servers (WvW wise) into half the servers there are now would diminish those dead-zone periods where 1 server gains superiority over the others (gaining points, making enemy places paper, and upgrading their own to T3 for the next time-zone).

Change the game from the winner winning by numbers to the winner winning by skill.

I’m sure this will reply will be ignored though, like every other reply.

Happy to go with this new matchup system when you change the way the scoring works. It was pointed out to you at the very beginning Anet. “Off hours” capping completely imbalances the game.

Please don’t comment on my off hours being your on hours… the debate has raged since launch. Servers with greater coverage gain more ppt.

Factor in the number of players from a server in the actual map and apply some rulesets around this to determine value of objectives.

Get rid of the most points for yaks for 10 ppl hitting it. A solo player should get the most points available for the slap, rather than the zerg.

Have you, Anet, actually played on a demoralized server ? Lets see what this new matchup system produces… im guessing alot of very very very disgruntled wvw players.

Doyen.7063:

Get rid of the most points for yaks for 10 ppl hitting it. A solo player should get the most points available for the slap, rather than the zerg.

This is off topic but I can’t help respond. I too feel it is somehow wrong that everyone in the Zerg gets the same point for a cap as a highly skilled small team. Not so much, in fact, for dolys, which anyone can kill given time, but for sure, for towers and keeps.But, what’s the alternative?

At the moment, although WvW is not perfect by any means, all people on a server do pull together and do support each other. You are always glad of an extra helping hand. With this kind of change people would be scared of assisting other players for fear of the verbal abuse that will come from some, even if a minority. Even if most people do not abuse others there will still be comments along the line of “why did you have to get involved, I had that covered”.

A change that awarded points by how many participated would lead to a destruction of team spirit, less focus on actual WvWvW and more on competing WITHIN a server for caps and points.

One idea, may not work but I’ll put it out there. How about a weekly reward (wXP or other) at the end of each matchup, based on your personal contribution to the week. This would be based in some way on your contribution to each battle/cap. That way, the reward is abstracted from individual caps and less likely (won’t say impossible) to cause intra-server strife!

I just want to point something out to the people that advocate servers with less population getting more points for holding objectives, or accounting for night capping through the point system:

Gathering more points will make you come up against opponents with even more coverage and will thus lead to worse fights. The current system is actually working more or less in getting the servers with more coverage pitting against each other. If you start giving more points to servers with less population they’ll end up against servers with even more people. I’d prefer fighting other servers with the same issues as mine to be honest.

I also don’t really understand the “I’m bored of fighting against these servers, let me fight against servers that will totally roll over mine with superior population” mentality, but Anet is adjusting the system for those. Of course the servers getting stuck in tiers they can’t get any decent fights in were right to complain, but the complaints from people in other tiers I really don’t understand.

Jon,

WvW rankings need a way for us to show what rank (title) we are to other players on our own server in WvW and PvE. Are we gonna get it? Even just a title and small graphic next to our name with it would be cool. i.e. WvW Major (insignia icon)
Something along the lines of what the military has http://www.aviationexplorer.com/Military_Insignia_Rank_Structure_Charts/US_Military_Enlisted_Ranks_All_Branches_Of_Service.jpg

Besides this, we need to take the matchups/competition between servers to another level. Have something like a Heavy/Middle/Light/Feather weigh divisions and servers are put in these divisions based on their performance(points). Have servers from each division battle it out for top prize.
The reward would be some banner that would be flown in Lion’s Arch (showing the matchup world date, the servers competing at that time, and the final rank of one’s own server). ALSO for the top 3 winners, some other prize as appropriate for 1-3. Stuff like Siege, Special Title, etc etc.

I mean, it’s all good that we have weekly matches, but currently the only thing we have to know the ‘winner’ of a round of matches between servers is only the ‘collective’ activity since launch. I feel there needs to be something more.

Az z.2746:

JonPeters, the problem is not the matchmaking, but the score system itself. Fix the scoring system so that the server with most coverage doesn’t automatically win.

Ya night capping or day etc.. is a pain but i am not sure if you could fix it. Maybe cutting off WvW at set times but then worlds would become exclusive for time zones not that best way to do it. The world doing the most work should tend to win or at least do well in wvw how do you size that to having more effective coverage?

Problem is you don’t have enough servers to have such randomization put into effect.

With 100 or 200 different servers, you would have movement between tiers without having one server meeting one or two other ones much to powerful for itself.

Right now, there is simply too much difference in power between tiers to allow such a randomization.

Blanche Neige.7241:

Problem is you don’t have enough servers to have such randomization put into effect.

With 100 or 200 different servers, you would have movement between tiers without having one server meeting one or two other ones much to powerful for itself.

Right now, there is simply too much difference in power between tiers to allow such a randomization.

I do not think its going to be true randomization. If it was it would be very well random lol. More of a system to make it so other worlds can play vs other worlds then the same 3 over and over mainly problems at the top and may be bottom.

JonPeters.5630:

This is excellent and would be true if our rating currently had significant meaning. Once this new randomization is in place for a while we expect both ratings and deviation will stabilize a bit more. If at that point our simulation is still generating bad match-ups we can absolutely look at lowering the range.

would it not be more prudent in this case to say…run a very large simulation to as many as possible and see how varied and well…out of balance…match up’s that occur compared to well balanced match ups? doesn’t that seem more viable for the future of a video game than just throwing it into the game, praying it works, then if it doesn’t say “oh (kitten)” and then trying to fix it. I think we’ve seen this problem with other updates as well where things weren’t clearly truly tested and just thrown in, then later told, that it would, at some point, when they got around to it, changed around. I just think there’s a better way of testing the system at first then just throwing in the system. Heck I would even say that it would be more prudent to start with a smaller random number, then raise it, instead of lowering it. If a rank 9 gets put against a rank 1 and other’s, it could lead to a blow out that’ll make all servers bored and not care anymore. Compared to if a rank 1 is put against a rank 4 and it’s a hard fight, and it’s later decided to increase it.

another option to prevent too much volatility due to randomness (but still allow it occasionally, just to keep things interesting) is to use a flattened bell curve distribution for random numbers, rather than a linear distribution.

with a linear distribution, you are just as likely to get a +0.9 as you are a 0 or a -0.9. with a flattened bell curve numbers near zero will be more likely than numbers far from zero, but the outliers will still happen from time to time.

I think a flattened bell curve is needed rather than a regular one because otherwise the outliers won’t happen nearly often enough to have a noticeable impact.

a computationally easy way to do something similar (but mathematically less accurate) is to simply take two linear random numbers between -0.5 and +0.5 and add them together. statistically, numbers near 0 will be more common than numbers farther away (similar to how 1-6 are all equally likely on a single die, but with 2 dice 7 is very common while 2 and 12 are rare).

-ken

Snowreap.5174:

another option to prevent too much volatility due to randomness (but still allow it occasionally, just to keep things interesting) is to use a flattened bell curve distribution for random numbers, rather than a linear distribution.

with a linear distribution, you are just as likely to get a +0.9 as you are a 0 or a -0.9. with a flattened bell curve numbers near zero will be more likely than numbers far from zero, but the outliers will still happen from time to time.

I think a flattened bell curve is needed rather than a regular one because otherwise the outliers won’t happen nearly often enough to have a noticeable impact.

a computationally easy way to do something similar (but mathematically less accurate) is to simply take two linear random numbers between -0.5 and +0.5 and add them together. statistically, numbers near 0 will be more common than numbers farther away (similar to how 1-6 are all equally likely on a single die, but with 2 dice 7 is very common while 2 and 12 are rare).

-ken

This would seem like a better approach indeed. I don’t mind the occasional snake eyes, but let’s keep those 7s coming a bit more frequently.

Snowreap.5174:

another option to prevent too much volatility due to randomness (but still allow it occasionally, just to keep things interesting) is to use a flattened bell curve distribution for random numbers, rather than a linear distribution.

with a linear distribution, you are just as likely to get a +0.9 as you are a 0 or a -0.9. with a flattened bell curve numbers near zero will be more likely than numbers far from zero, but the outliers will still happen from time to time.

I think a flattened bell curve is needed rather than a regular one because otherwise the outliers won’t happen nearly often enough to have a noticeable impact.

a computationally easy way to do something similar (but mathematically less accurate) is to simply take two linear random numbers between -0.5 and +0.5 and add them together. statistically, numbers near 0 will be more common than numbers farther away (similar to how 1-6 are all equally likely on a single die, but with 2 dice 7 is very common while 2 and 12 are rare).

-ken

The bell curve drawing is basically done in the mean method shown above. Since means, calculated from ANY probability distribution, will, with large enough numbers, go to a gaussian distribution. I used a small number of draws to let it start to go to a gaussian, but still maintain a decent amount of scatter, just very limited. It’d be much more like a poisson distribution than anything else. But, you’re right, using purely a constant distribution, with such a large potential for the deviation, will make the match ups far too random and too scattered. They might say “it will” even out, but as my simulations show, for the first match up, and probably the first several, the potential match ups are going to be have too high of a probability of several blow out matchs. Not because of a lack of skill, but a lack of numbers.

I think one way to implement this and the glicko ratings is for the first month or so make the match ups only like 3 days. This is really short, and not ideal, but then at least if a server is getting blown out, they won’t have to sit there for a week and deal with it. It will also give them plenty of data to stabilize the rankings, which apparently should just magically happen.

I really do wish that anet would give the players USEFUL data. If you expect them to stabilize, please show us. You gave us that wonderful table of data in your press release, then go on to tell people “It’s useless, don’t look at it, it means nothing”. Then why show it?! Show us what you, the programmers, are expecting. That will give us faith in the system. Not telling us “what we showed you won’t happen, it’s worthless, don’t look too much into it. Here is how it will work does jedi mind trick”

bump bump bumpity bump

Az z.2746:

JonPeters, the problem is not the matchmaking, but the score system itself. Fix the scoring system so that the server with most coverage doesn’t automatically win.

Exactly.

Already last year I gave a good suggestion for this:
The score obtained from each objective should be reversely proportional to the proportion of players per total number of players.

Sorry English is not my native language, so here is an example:
Let’s say server A has 100 players, server B 200 and server C 300. Total number of players in that tier is 600.

Let’s say objective X would give 100 base points, server A would get 100/ (100/600) = 600 points doing that, while server B would get 100 / (200/600) = 300 points doing that and server C would get 100 / (300/600 points doing that.

Objective Y would give 300 base points, server A would get 300/ (100/600) = 1800 pts doing that, while server C would get 300 / (100/600) = 600 points from it.

In other words, because server C has 3 times as many players, they need to earn 3 times more base points from the objectives to earn the same amount of actual points for their server.

Another way to balance the game is to make the outmanned “buff” to actually do something. Preferably increase the defenses of the weaker side e.g. give more health and toughness to the outmanned side (NPCs, players and structures). NPCs and objects should be buffed more than players, to avoid giving the outmanned side “godlike” powers.

33.3%/33.3%/33.3% <- perfectly equally distributed population on 3 servers

It could have multiple levels e.g.
Outmanned (<25% total population) +50% NPC & structure hp & thoughness, +10% player health & thoughness
Overpowered (<17% population): +100% NPC & structure hp & thoughness, +20% player health & thoughness
Last Stand (<10% total population): +200% NPC & structure hp & thoughness, +30% player health & toughness

So in other words, if there is a total 100 players on a map, outmanned would be triggered on that side who has less than 25, outpowered when that side has less than 17 and last stand when that side has less than 10 players. It is possible to have two sides having similar outmanned buffs e.g. one server totally dominating the numbers game with 90 players and the others have just 10.

My suggestion wouldn’t stop zerging or night capping, but it would make it a bit slower and give at least somewhat better chances for the weaker side. A more clever formula could take in account the total number of players in all maps to avoid abuses like having a big zerg teleport from one map to another and then leaving them empty (to trigger Last Stand on them).

The current system is bad, because it leads to even bigger imbalances in the long run. The losing side has even less motivation, a downward spiral. While some band wagoners will join the winning side.

I am actually curious how the current outmanned works. It seems to require quite bit population difference e.g. if there is 20 players on your side and while enemy has 29 (= significant advantage), it doesn’t seem to trigger. Can some developer please explain what are the current limits?

Denaria, then what is to stop servers from having blackouts to falsely lower their population? While I think it would be amazing if the score put into population, how do you make sure it doesn’t punish the players of big servers who have been there from the beginning? It also is much too drastic of a change, especially because if, in your example, server A takes a keep with 20 people, and server B takes it with 20 people, server B gets punished for it. Both servers did the same amount of work, but they got totally different levels of rewards, which really isn’t fair to the individual players. If two servers do the same amount of work, they shouldn’t be rewarded on totally different levels.

Handin.4032:

Denaria, then what is to stop servers from having blackouts to falsely lower their population? While I think it would be amazing if the score put into population, how do you make sure it doesn’t punish the players of big servers who have been there from the beginning? It also is much too drastic of a change, especially because if, in your example, server A takes a keep with 20 people, and server B takes it with 20 people, server B gets punished for it. Both servers did the same amount of work, but they got totally different levels of rewards, which really isn’t fair to the individual players. If two servers do the same amount of work, they shouldn’t be rewarded on totally different levels.

You do realize that in my example server B has twice the population compared to A, thus they should be able to take several keeps at same time.

Nothing would stop manipulating the numbers, but you do realize if you intentionally have just e.g. 9 men on the map to trigger the Last Stand buff and enemy has 91 players, you are not able to defend your keeps and towers very long. Bigger numbers would still guarantee you more control of the map.

Surely just 9 men could e.g. defend a keep quite long time, assuming it is full of siege, against a big force, but I don’t see them actually defending an entire map. So in other words if a dominating server would conquer all of map and then intentionally manipulate their player levels way too low, they would be unable to defend all their towers, camps and keeps, since the totally outnumbering enemy could attack several locations at same time.

My suggested system is based on the notion that you get points per work done per number of players.

The current system actually seems in worst cases reward your server more points if you have more players doing the same easy objective. If three players participate in killing a Dolyak, the server gets triple points compared to just one player doing it. Can some developer confirm this? And please explain why it is so? (this punishes solo roamers and servers with low coverage)

The current system also gives accumulating bonuses to health, healing, endurance regeneration, etc. Those bonuses are small, but the winning server slowly gets better and better stats than their enemies. In other words: winner takes it all, loser standing small.
http://wiki.guildwars2.com/wiki/Power_of_the_Mists

All I see that the current system totally favors the side with the better coverage and numbers. With the new randomized match up system, you will see that some EU tiers are gonna probably have even more unbalanced matches. And this lead to losing servers losing players and winning servers getting more.

For couple of months when Blacktide was bleeding numbers (they dropped from tier 1 to tier 9) I was fighting totally imbalanced matches, at worst having 2 vs 30-50. Now this Spring at Desolation we have been generally outmanned even during the evening on our own Borderland, not to mention enemy borderlands. 6+ months fighting outnumbered. Arenanet please give us quickly a different type of playing experience, instead of forcing to change servers! Make the outnumbered buff MATTER.

Deniara Devious.3948:

Another way to balance the game is to make the outmanned “buff” to actually do something. Preferably increase the defenses of the weaker side e.g. give more health and toughness to the outmanned side (NPCs, players and structures). NPCs and objects should be buffed more than players, to avoid giving the outmanned side “godlike” powers.

33.3%/33.3%/33.3% <- perfectly equally distributed population on 3 servers

It could have multiple levels e.g.
Outmanned (<25% total population) +50% NPC & structure hp & thoughness, +10% player health & thoughness
Overpowered (<17% population): +100% NPC & structure hp & thoughness, +20% player health & thoughness
Last Stand (<10% total population): +200% NPC & structure hp & thoughness, +30% player health & toughness

My idea (in the same vein):
Bring the orb of power back. However, this new orb doesn’t affect player stats. This orb applies what is currently the outmanned buff to the entire server (PvE sections too, if that’s possible) but to a lesser extent. (sans no repair cost, of course)

The reason that’s relevant is because the second part of this idea is to take what used to be the old orb buff, and make it the new and improved outmanned buff. This way outmanned is actually useful. This new outmanned buff would retain the property of no equipment damage upon death.

Not only does this make the outmanned buff useful, it motivates PvE players to come to WvW and claim the orb so they get additional magic find, etc.

JonPeters.5630:

This is excellent and would be true if our rating currently had significant meaning. Once this new randomization is in place for a while we expect both ratings and deviation will stabilize a bit more. If at that point our simulation is still generating bad match-ups we can absolutely look at lowering the range.

But would that not just recreate what we already have? i.e. static boring matchups where servers find themselves stuck in the same tier despite dominating for weeks. (…and everyone knows that the only reason GH won this week was because of the dc problems this week – RoS didn’t just throw a 10k lead).

I am willing to try the new system out but if we end up back to the start then what will be next?

I personally still think a one up, one down would mix things up while keeping servers at the right level… if a server goes up one week and then finds themselves out of their depth then they will simply go back down the following week.

Deniara Devious.3948:

kitten .2746:

JonPeters, the problem is not the matchmaking, but the score system itself. Fix the scoring system so that the server with most coverage doesn’t automatically win.

Exactly.

Already last year I gave a good suggestion for this:
The score obtained from each objective should be reversely proportional to the proportion of players per total number of players.

Sorry English is not my native language, so here is an example:
Let’s say server A has 100 players, server B 200 and server C 300. Total number of players in that tier is 600.

Let’s say objective X would give 100 base points, server A would get 100/ (100/600) = 600 points doing that, while server B would get 100 / (200/600) = 300 points doing that and server C would get 100 / (300/600 points doing that.

Objective Y would give 300 base points, server A would get 300/ (100/600) = 1800 pts doing that, while server C would get 300 / (100/600) = 600 points from it.

In other words, because server C has 3 times as many players, they need to earn 3 times more base points from the objectives to earn the same amount of actual points for their server.

Another way to balance the game is to make the outmanned “buff” to actually do something. Preferably increase the defenses of the weaker side e.g. give more health and toughness to the outmanned side (NPCs, players and structures). NPCs and objects should be buffed more than players, to avoid giving the outmanned side “godlike” powers.

33.3%/33.3%/33.3% <- perfectly equally distributed population on 3 servers

It could have multiple levels e.g.
Outmanned (<25% total population) +50% NPC & structure hp & thoughness, +10% player health & thoughness
Overpowered (<17% population): +100% NPC & structure hp & thoughness, +20% player health & thoughness
Last Stand (<10% total population): +200% NPC & structure hp & thoughness, +30% player health & toughness

So in other words, if there is a total 100 players on a map, outmanned would be triggered on that side who has less than 25, outpowered when that side has less than 17 and last stand when that side has less than 10 players. It is possible to have two sides having similar outmanned buffs e.g. one server totally dominating the numbers game with 90 players and the others have just 10.

My suggestion wouldn’t stop zerging or night capping, but it would make it a bit slower and give at least somewhat better chances for the weaker side. A more clever formula could take in account the total number of players in all maps to avoid abuses like having a big zerg teleport from one map to another and then leaving them empty (to trigger Last Stand on them).

The current system is bad, because it leads to even bigger imbalances in the long run. The losing side has even less motivation, a downward spiral. While some band wagoners will join the winning side.

I am actually curious how the current outmanned works. It seems to require quite bit population difference e.g. if there is 20 players on your side and while enemy has 29 (= significant advantage), it doesn’t seem to trigger. Can some developer please explain what are the current limits?

I posted this elsewhere:

Personally, other than a modest “outmanned” buff, that you could kinda justify in the real world down to adrenaline, I don’t like faking the battles to make it easier for a weaker team.

What I might prefer though is to move to more like golf handicap system – adjust the in-game scoring for weaker/ stronger teams. Weaker servers, get more points for a cap. Ratings could still be calculated on the basis of an un-handicapped score (as now). But the displayed score in-game could reflect the handicap. That way we would no longer have the situation where a team wins the matchup, but loses points. And a lower tiered team that fought hard would actually stand a chance of winning instead of losing badly but having the dubious “honour” of still gaining rating points!

Personally I also prefer this to adjusting scoring on current active population (as suggested in another thread). A server that puts out a full team on the night deserves reward, not punishment, for that.

It seems better to me.

- lower tiered servers still have a better chance to actually win, not just gain rating points few care about

- high tiered servers are handicapped for their higher average population (and, yes, ability/organisation)

- there is still an incentive for servers/guilds to “get their team out” rather than have it all lost by an “overpowered” nerf.

Another idea, that, I don’t know, might help a bit with server balancing.

How about handicap wXP for forts (not individual players) based on server rating. So players on a lowly server get more wXP for capping a high tiered server’s tower/keep/castle.

Seems to reflect the real-life reward due and might encourage some to move from high to lower tiers.

Jong.5937:

Another idea, that, I don’t know, might help a bit with server balancing.

How about handicap wXP for forts (not individual players) based on server rating. So players on a lowly server get more wXP for capping a high tiered server’s tower/keep/castle.

Seems to reflect the real-life reward due and might encourage some to move from high to lower tiers.

Well, for some people, such as myself, who have been in TC since the beginning, would see this as kind of a slap to the face. We’ve worked very hard, with many of us sticking around from the beginning, to get to where we are. So then to be told that we now have to do, for instance, double the work to get the same amount of reward that someone in a lower tier does. For instance, TC used to be in T4, and many of us worked long nights to get up to T3, then to T2. Yes, we got some transfers, but doing real time “handicaps” for rewards on high population server is only punishing people who have worked hard (not that those in Lower tier servers don’t, they most certainly do! We’ve seen it when they come up!). I think the better way to do it would be to NOT do real time handicaps, with things like lowering WXP or points. Instead, have a weight value in the glicko scores that puts into account the population difference, even something like was mentioned above with the inverse relationship. This would still give people the same rewards when playing, but allow the servers to move up more when they win, or do at least get rewarded for doing better!

Well I understand why you would feel like that, but (on average) I’d still say that it IS easier for a higher tier to cap than a lower tier, so the reward should reflect that.

Note: I am only talking a handicap relative to your opponents, not some kind of absolute scaling, so if you are playing teams close in rating the rewards will not be that different.

Jong.5937:

Well I understand why you would feel like that, but (on average) I’d still say that it IS easier for a higher tier to cap than a lower tier, so the reward should reflect that.

Note: I am only talking a handicap relative to your opponents, not some kind of absolute scaling, so if you are playing teams close in rating the rewards will not be that different.

I agree with that ,but remember, you’re still facing an opponent in your tier. If a T1 server was facing a T3 server, then it would be much easier. However, if you included a weighting factor in the glicko score, instead of the real-time rewards, then the server would raise/lower accordingly, without affecting the single player. But I do understand that idea I just think if you did real-time stuff, you’d have to only look at local populations (i.e. how many are defending the tower vs. how many are attacking it). But, that just my opinion, though I have fought in Tiers 1→4 (so I haven’t just been in high tiers hehe)

Remember from next week you could be matched with any server +/- 9 from you (roughly). Personally, as I said above, I don’t like the idea of handicapping based on current server population or tower population as that penalises people for getting their team out. Trying hard and having most people you can on the field should benefit you.

But I do think a team 8 places below their opponent in the table (for example) who caps towers should get greater rewards, both personally and as a team, and not just in their rating at the end of the week. But, also, just my opinion

Handin.4032:

I think the better way to do it would be to NOT do real time handicaps, with things like lowering WXP or points. Instead, have a weight value in the glicko scores that puts into account the population difference, even something like was mentioned above with the inverse relationship. This would still give people the same rewards when playing, but allow the servers to move up more when they win, or do at least get rewarded for doing better!

This is what I was trying to suggest. WXP rewards should stay the same, at least for all big objectives, like towers, keeps and castle, no matter how many people contribute to that. But a server with proportionally smaller active population should get more glicko points if they capture an objective.

Desolation finally dropped to tier 2. We wanted it badly, because we simply lacked the coverage and numbers for EU tier 1. The game is now so different and our side enjoys it more. Last night under my command we completely conquered our own borderlands using just map chat and PuGs. Was it all about skill? No, our strategy was just one big 20-30 man mono blob, plus good intel via map chat and couple of players force doing supplies and refreshing siege on keeps. One of our opponents (Kodash) was oumanned (TV showing champion’s league final might have something to do with it, but that was my intention to attack then ;-) ). And as soon as SFR started to get constantly beaten and wiped, their morale dropped and they also became outmanned. Once you played on the outmanned side for so many months, you desperately want victory. The reason why I told such a silly story that it just proves my point. In some big fights we outnumbered the enemy only like 30 vs 20, 1.5 to 1 ratio. This already a huge advantage, allowing an easy and complete wipe of enemy.

I’m confused because I thought that was exactly what happened now. Not directly using populations, I realise, but a lower tier server gets more points if it does well. It is precisely why it is possible to lose and still gain points or win and lose them.

Jong.5937:

I’m confused because I thought that was exactly what happened now. Not directly using populations, I realise, but a lower tier server gets more points if it does well. It is precisely why it is possible to lose and still gain points or win and lose them.

Thats for rating, not for the overall score… which is a bit backwards since your rating actually defines your server and not the score from any one week. All of these suggestions are about fixing the scoring system: i.e. trying to fight night/“someone else’s day” capping.

If the scoreboard for WvW showed rating along with score, things would probably change in regards to topics like this but, since that isn’t the case and most people don’t even know what rating is, we’ll continue to have people trying to “fix” when/how people play the game.

Jong.5937:

I’m confused because I thought that was exactly what happened now. Not directly using populations, I realise, but a lower tier server gets more points if it does well. It is precisely why it is possible to lose and still gain points or win and lose them.

That is true, but it normally isn’t proportional to the population difference. In some tiers (such as the lowest or the highest), there has to be several big blow outs in a row for anyone to be able to actually move up a tier. They might be able to bounce around theirs for a while, but for the most part, once you’re in a tier, the rating difference between them gets large that hopping between them is hard. So the rank 6 team might not ever face the rank 7 unless there are massive blow outs in rank 7’s tier.

so, after running a lot of Monte Carlo simulations, I have to say that the linear distribution of random numbers is probably fine.

I was worried that with a single ‘die roll’ the outliers were just as likely as the middle numbers, but in order to really get those outlier matches, all the servers around you have to get the right rolls too.

For a server in the middle of the US rankings, it’s not a single die roll that determines the outcome, but 24 die rolls. At the top of the US rankings, 14 die rolls contribute meaningfully to the result. The combination of all those rolls naturally leads to a bell-curve result.

For example (these results are based on a fresh run of 10 billion trials using the latest US ratings taken from the Leaderboards last night):

Sanctum of Rall
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 64.394177 JQ
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 58.941160 BG
xxxxxxxxxxxxxxxxxxxxxxxxxx 25.770207 Tarnished Coast
xxxxxxxxxxxxxxxxxxxxxxx 23.306251 Dragonbrand
xxxxxxxxxxxxxxxxxx 18.202404 Fort Aspenwood
xxxxx 5.071285 Maguuma
xx 1.952292 Kaineng
xx 1.595952 Yak’s Bend
x 0.749685 Sea of Sorrows
0.011832 Crystal Desert
0.003752 Stormbluff Isle
0.000988 Ehmry Bay
0.000015 Borlis Pass
0.000000 Anvil Rock
0.000000 Darkhaven
0.000000 Isle of Janthir
0.000000 Sorrow’s Furnace
0.000000 Gate of Madness
0.000000 Devona’s Rest
0.000000 Northern Shiverpeaks
0.000000 Henge of Denravi
0.000000 Ferguson’s Crossing
0.000000 Eredon Terrace

Ehmry Bay
0.000988 Sanctum of Rall
0.020389 Jade Quarry
0.049147 Blackgate
x 1.133204 Tarnished Coast
x 1.439178 Dragonbrand
xx 2.376285 Fort Aspenwood
xxxxxxxxxx 9.665702 Maguuma
xxxxxxxxxxxxxx 14.701245 Kaineng
xxxxxxxxxxxxxxxx 15.535723 Yak’s Bend
xxxxxxxxxxxxxxxxxxx 19.223970 Sea of Sorrows
xxxxxxxxxxxxxxxxxxxxxxxx 24.370725 Crystal Desert
xxxxxxxxxxxxxxxxxxxxxxxx 23.874882 Stormbluff Isle
xxxxxxxxxxxxxxxxxxxxxxxx 23.759505 Borlis Pass
xxxxxxxxxxxxxxxxxx 18.452743 Anvil Rock
xxxxxxxxxxxxxxxxx 16.617779 Darkhaven
xxxxxxxxxxx 10.919846 Isle of Janthir
xxxxxxxxxx 9.800816 Sorrow’s Furnace
xxxxx 5.050317 Gate of Madness
xx 1.685463 Devona’s Rest
x 1.084844 Northern Shiverpeaks
0.175680 Henge of Denravi
0.040929 Ferguson’s Crossing
0.020642 Eredon Terrace

Eredon Terrace
0.000000 Sanctum of Rall
0.000000 Jade Quarry
0.000000 Blackgate
0.000000 Tarnished Coast
0.000000 Dragonbrand
0.000000 Fort Aspenwood
0.000000 Maguuma
0.000000 Kaineng
0.000000 Yak’s Bend
0.000002 Sea of Sorrows
0.003185 Crystal Desert
0.007002 Stormbluff Isle
0.020642 Ehmry Bay
0.460162 Borlis Pass
xx 2.268529 Anvil Rock
xxx 3.241324 Darkhaven
xxxxxxx 7.442637 Isle of Janthir
xxxxxxxxx 8.670657 Sorrow’s Furnace
xxxxxxxxxxxxxxx 15.456251 Gate of Madness
xxxxxxxxxxxxxxxxxxxxxxxxxx 25.985287 Devona’s Rest
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 29.918847 Northern Shiverpeaks
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 47.912876 HoD
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 58.612601 F C

(edit: I am very amused that the forum software replaces the common two-letter abbreviation used for Ferguson’s Crossing with “kitten”)

-ken

If they implement a wider range of WvW match up they should remove gesting. Right now the PvE server are pretty much stuck together. They are generally of equal strength and usually offer good match up. As of right now everybody can enjoy the PvE side of any server and offering nothing to it WvW side.