Guild Wars 2

I’ve haven’t played any matches today so I’m pretty sure the leaderboard is up to date for me at the time of this post.

I’ve gone 22-16 for a total of 26 leaderboard points or nearly 1.2 points per win.

I haven’t spent a lot of time pondering this but I see one of two basic explanations for this.

A) Nearly all my losses occurred in even odds (50%/50% chance of winning) matches and my team scored at least 400.

B) I’ve been in a lot of matches were I was favored to lose and I either won or it was a relatively close game.

I recall maybe 5-6 losses to my part where my team scored 400. Which points more heavily to explanation B.

Does the matchmaking system make any effort to balance your number of games where you are favored to win with the number of games where you are favored to lose?

How good is the system overall? Whats the ratios of 50/50 games to 60/40 games and 80/20 games?

Does the matchmaking system make a binary decision (win/lose) or does it have a predicted point spread as well?

It has a percentage win “chance” based on the average MMRs of the teams, how many are solo on a team or the other.

When you have points per win, greater than one, that simply means you’re scoring more points than the odds predicted, on average over all your games.

It’s not really possible to tell if you’re winning when you’re not predicted to, or simply not losing as badly as you’re predicted to. Can’t sort them out because we can’t see the odds. The system may have a binary win/loss prediction, we wouldn’t know, we just know we get rewarded for the more points we score against bad odds of success.

The matchmaking system does its best efforts to put you in a balanced match on a per match basis. When it does matchmaking, it won’t look at someone and be like: “oh they had many matches recently where they were not favoured to win, let’s give em an easy one”

How good is the system overall…….. it succeeds relatively well in the big picture for average MMR people: You will 500-0 spawn camp people just as often as you get 500-0 spawn camped, usually.

However, that’s just the macro look at the matchmaking. What players really want is for every match to be as close as possible.

Also if you’re a high MMR player, you get a 500-0 spawn camped more often because you’re always stuck as the messiah to somehow bring 4 new players to success against a 5 person premade.

High MMR players can only have a good experience in premade vs. premade ranked matches,

New to medium MMR players can only have a good experience if they are alright with rather high occurrences of 500-0ing people and getting 500-0’d.

I’d say the overall new system is worse than the previous, that’s just my opinion.

Chaith.8256:

Also if you’re a high MMR player, you get a 500-0 spawn camped more often because you’re always stuck as the messiah to somehow bring 4 new players to success against a 5 person premade.

pretty much the most true thing ever said

Justin ODell.9517:

Matchmaking does not know about or operate according to win/loss streaks or account averages.

There may be some merit to adding logic to bust up streaks, but it isn’t a planned feature afaik.

there may be merit in adding memory to matchmaking to bust up losing streaks just to keep people playing… but i dont think there is merit in purposefully trying to bust up a winning streak. plus, both streak types are already supposed to be taken care of by changing a players mmr. you lose, next game should be slightly easier. you win, next game should be slightly harder. but i guess its all assuming the player pool is large enough to accommodate nonrepetitive matchmaking. which it isnt, because we hear stories of premades playing the same premades 10x in a row during off hours.

Chaith.8256:

When you have points per win, greater than one, that simply means you’re scoring more points than the odds predicted, on average over all your games.

It’s not really possible to tell if you’re winning when you’re not predicted to, or simply not losing as badly as you’re predicted to. Can’t sort them out because we can’t see the odds. The system may have a binary win/loss prediction, we wouldn’t know, we just know we get rewarded for the more points we score against bad odds of success.

The matchmaking system does its best efforts to put you in a balanced match on a per match basis. When it does matchmaking, it won’t look at someone and be like: “oh they had many matches recently where they were not favoured to win, let’s give em an easy one”

How good is the system overall…….. it succeeds relatively well in the big picture for average MMR people: You will 500-0 spawn camp people just as often as you get 500-0 spawn camped, usually.

However, that’s just the macro look at the matchmaking. What players really want is for every match to be as close as possible.

Also if you’re a high MMR player, you get a 500-0 spawn camped more often because you’re always stuck as the messiah to somehow bring 4 new players to success against a 5 person premade.

High MMR players can only have a good experience in premade vs. premade ranked matches,

New to medium MMR players can only have a good experience if they are alright with rather high occurrences of 500-0ing people and getting 500-0’d.

I’d say the overall new system is worse than the previous, that’s just my opinion.

I’d like to start out that I agree with your points they more of less match my take away of the system.

With that being said are they based on speculation or have you gotten a behind the scenes looks are more of the numbers than the average player :-)?

For you first point, it doesn’t really matter if you’re winning when the system said you should lose or not losing as bad as the system predicted; you’re still getting the short end of the matchmaking stick. Based on the test season I agree that over time the system seems to be balancing out since virtually no one with a ton of games had more than 1 point per win.

I also agree that a couple of weeks ago a good MMR player would seem to be the outlier on a bad team to bring the average in line. Personally I haven’t seen this happen nearly as much lately so maybe they tweaked the allowed variance or more people are playing.

I think the system we have is good wau of doing things on a per match basis given the fact that the system can’t find a perfect 50/50 matchup all the time. I would recommend another test season with some type of cap on the number of games that counted (already suggested by other people).

Dixo.3876:

I’d like to start out that I agree with your points they more of less match my take away of the system.

With that being said are they based on speculation or have you gotten a behind the scenes looks are more of the numbers than the average player :-)?

For you first point, it doesn’t really matter if you’re winning when the system said you should lose or not losing as bad as the system predicted; you’re still getting the short end of the matchmaking stick. Based on the test season I agree that over time the system seems to be balancing out since virtually no one with a ton of games had more than 1 point per win.

I also agree that a couple of weeks ago a good MMR player would seem to be the outlier on a bad team to bring the average in line. Personally I haven’t seen this happen nearly as much lately so maybe they tweaked the allowed variance or more people are playing.

I think the system we have is good wau of doing things on a per match basis given the fact that the system can’t find a perfect 50/50 matchup all the time. I would recommend another test season with some type of cap on the number of games that counted (already suggested by other people).

I don’t have any insider knowledge, just observations and the process of elimination mostly.

Justin ODell.9517:

Matchmaking does not know about or operate according to win/loss streaks or account averages.

There may be some merit to adding logic to bust up streaks, but it isn’t a planned feature afaik.

Taking streaks into consideration would be a great idea!
By experience, there are few things more frustrating than a long losing streak. I guess there are many people that gave up on PvP after feeling that it is unwinnable.
Usually it hasn’t been happening that much with me, but there were days that I lost with my guardian, like, 10 times in a row.
As much as I am not a good guardian, I’m pretty sure I’m not able to bring down my entire team with my badness… I hope.

The system should probably try to give you as many 50/50 type games as possible. Long losing streaks are very possible with this system that have almost nothing to do with your skill level if you solo. During the test season I got mired about a 35% win percentage for quite a long time. I persevered and got relatively close to 50% overall by the end but a lot of players won’t keep playing like I did if after 50-75 games they are winning not much more then 1/3 of them.

Speaking of streaks:Dec 28. The day I almost threw my pc out the window. I went 17 wins and 39 losses and still gained 15 lb points. My obviously got put into match after match that I was destined to lose.

http://www.gw2score.com/PvP/player/Dixo3876

I will say that I’ll take the changed weighting were I’m less likely to face premises as a solo queuer in exchange for an increased chance of duplicate classes any day of the week.