Matchmaking By Way of the MATHS

Matchmaking By Way of the MATHS

in PvP

Posted by: Spooko.5436

Spooko.5436

like many players, i am certain there is a skew to matchmaking, and that the matchmaking system is severely broken and cannot evenly distribute the classes….

so i took a sample of 32 matches, played as thief, here are my results:

of 32 matches, in 15 of them i had more thieves on my team than the enemy team, giving me a 47% chance of having more thieves on my team than the enemy team.

of 32 matches, in 0 of them did the enemy team have more thieves than my team. giving me a 0% chance of facing a team with more thieves than mine.

of 32 matches, in 6 of them i was the only thief player on either side, giving me a 19% chance of the being the only thief player.

of 32 matches, in 11 of them there was an even number of thieves on both sides, giving me a 34% chance of having thieves equally distributed.

of 32 matches, in 8 of them my team had more than 1 thief, and the enemy team had 0, giving me a 25% chance of having a thief on my team and not facing one on the other side (this number is also represented in the 15 matches where i had more thieves on my team over-all)

during these 32 matches, i had on my team in total: 60 thieves (38% chance of having another thief on my team) 10 warriors (6%) 18 guardians (11%) 7 engineers (4%) 8 elementalists (5%) 27 necromancers (17%) 12 mesmers (8%) 14 rangers (9%) 4 revenants (3%)

during these 32 matches i faced in total: 26 thieves (16%) 12 warriors (8%) 19 guardians (12%) 12 engineers (8%) 10 elemantalists (6%) 34 necromancers (21%) 12 mesmers (8%) 22 rangers (14%) 13 revenants (8%)

regardless of the outcomes of these matches (yeah, i carry pretty hard) there is a significant disparity between class distribution in matchmaking—at least when i queue as thief.

Matchmaking By Way of the MATHS

in PvP

Posted by: Lexander.4579

Lexander.4579

nice math homework son but 32 matches are nothing

Alex Shadowdagger – Thief – Blacktide

Matchmaking By Way of the MATHS

in PvP

Posted by: Spooko.5436

Spooko.5436

32 is a statistically significant sample.

Matchmaking By Way of the MATHS

in PvP

Posted by: Lexander.4579

Lexander.4579

maybe if you are playing only 50 matches per year

Alex Shadowdagger – Thief – Blacktide

Matchmaking By Way of the MATHS

in PvP

Posted by: kuritsutian.2987

kuritsutian.2987

taking more samples would change at most +/-3% on class distribution, so no biggie the statistic can somewhat show us current class popularity more than class distribution, you can say popularity affects distribution. To bad it was centered only on the thief and only measured class distribution, I am more concerned about the skill of players… meaning putting skilled players along with less skilled in order to compensate and balance teams… this would happen if they are matched by the average alone without having the standard deviation into account.

Suddenly in the Forums Everyone is now a Game designer!

Matchmaking By Way of the MATHS

in PvP

Posted by: Lexander.4579

Lexander.4579

Suddenly in the Forums Everyone is now a Game designer!

Alex Shadowdagger – Thief – Blacktide

Matchmaking By Way of the MATHS

in PvP

Posted by: kuritsutian.2987

kuritsutian.2987

Suddenly in the Forums Everyone is now a Game designer!

xD so true … I build my own indie games tho :P

Suddenly in the Forums Everyone is now a Game designer!

Matchmaking By Way of the MATHS

in PvP

Posted by: brannigan.9831

brannigan.9831

Suddenly in the Forums Everyone is now a Game designer!

Not sure why you put them on so high a pedastool? There are many people that post on these boards regularly that are at least as smart if not smarter then the devs at A-net and many of them have backgrounds in computer science and mathmatics that could match any of theres. That’s not a knock on them its just a fact. Some of these companies have actually hired players to siginificant spots over the years. Mike Lescault at Mythic back in the day on Dark Age of Camelot one day he was just the Guild Master of one the largest guilds in the game the next he was the main balance honcho at Mythic. Most of these dudes are like 25-30ish years old with like 5-10 years experience tops you act like they are all the Ben Carson’s of computer gaming. They are far from infallable. Again not meant as an attack just stating facts.

(edited by brannigan.9831)

Matchmaking By Way of the MATHS

in PvP

Posted by: Lexander.4579

Lexander.4579

Suddenly in the Forums Everyone is now a Game designer!

Not sure why you put them on so high a pedastool? There are many people that post on these boards regularly that are at least as smart if not smarter then the devs at A-net and many of them have backgrounds in computer science and mathmatics that could match any of theres. That’s not a knock on them its just a fact. Some of these companies have actually hired players to siginificant spots over the years. Mike Lescault at Mythic back in the day on Dark Age of Camelot one day he was just the Guild Master of one the largest guilds in the game the next he was the main balance honcho at Mythic. Most of these dudes are like 25-30ish years old with like 5-10 years experience tops you act like they are all the Ben Carson’s of computer gaming. They are far from infallable. Again not meant as an attack just stating facts.

idk m8 i was just quoting the signature of the guy above

nice of you to take your time to write all of that though

Alex Shadowdagger – Thief – Blacktide

Matchmaking By Way of the MATHS

in PvP

Posted by: Revolutionen.5693

Revolutionen.5693

I play thief aswell, and I’ve been seeing the same pattern. The reason is Anet’s communist vision of everyone having a 50% win rate. This philosophy seems to outweigh every other factor, and a team containing duplicate classes isnt as important. If you win more games than you lose, why not throw in 2 more thieves to make it an almost 100% chance of losing that game??

(edited by Revolutionen.5693)

Matchmaking By Way of the MATHS

in PvP

Posted by: daggerhands.9634

daggerhands.9634

32 is too small of a sample and if you had played any other class then thief you wouldnt have gotten the same numbers
+ if you played on a thief daily day it isnt that odd too see that many thiefs

Matchmaking By Way of the MATHS

in PvP

Posted by: Spooko.5436

Spooko.5436

i doubt any of your “armchair” critics are going to watch this..

but 32 is statistically significant…….

if i played other classes than thief my results might have been different, we will never know. i play thief, i expect a decent match when i play a thief. i have no interest in how my matchmaking would play out differently if i played on other classes, because i have no interest in other classes.

so, instead of dealing with hypotheticals, i got down to business and actually put in the work. the work shows that matchmaking is broken AF. this data is not hypothetical data from a bubble, it is the observed results of 32 actual matches. regardless of what hypotheticals you may believe influence the data, this is how the data plays out. this is how the matchmaking actually works. and it is totally broken.

Matchmaking By Way of the MATHS

in PvP

Posted by: Crinn.7864

Crinn.7864

So basically you’re convinced that the matchmaker has some conspiracy against you personally to put thieves on your team?

Also the reason you are getting more thieves is because you yourself are a thief which guarantees that in every match that your team will always have at least 1 if not more thieves, while the same is not true for the other team.

Also 30 is only significant for the purpose of sampling, that doesn’t make said sampling without error.

Sanity is for the weak minded.
YouTube

(edited by Crinn.7864)

Matchmaking By Way of the MATHS

in PvP

Posted by: Spooko.5436

Spooko.5436

i am convinced that the matchmaking is broken, which is what i stated.

any other inferences would be an error.

Matchmaking By Way of the MATHS

in PvP

Posted by: Stand The Wall.6987

Stand The Wall.6987

you are proving a known fact. matchmaking is far from perfect, and imo its because it is choosing from a limited pool of players.

Team Deathmatch for PvP – Raise the AoE cap for WvW – More unique events for PvE

Matchmaking By Way of the MATHS

in PvP

Posted by: Teutos.8620

Teutos.8620

So, in this math you have only included your own perspective.

Can you maybe include the one of every thief in those matches?

of 32 matches, in 15 of them i had more thieves on my team than the enemy team, giving me a 47% chance of having more thieves on my team than the enemy team.

That also means, that in 15 matches, your opponent thieves faced more thieves on the opponent side.

EU – Multiple times #1 SoloQ pre Dec 2014 (pure MMR based ladder)
Primoridal (S1) & Exalted (S2) & Illustrious (S3) Legend

Matchmaking By Way of the MATHS

in PvP

Posted by: Valkyriez.6578

Valkyriez.6578

What’s the point of collecting statistics for matches regarding classes when you can swap at a anytime before the gate opens?

Matchmaking By Way of the MATHS

in PvP

Posted by: Barnesy.5839

Barnesy.5839

during these 32 matches, i had on my team in total: 60 thieves (38% chance of having another thief on my team)

This original post is really bad, and possibly intentionally misleading. Of those 60 thieves that spooko counted, he was also counting himself. You didnt have a 38% chance of having another thief on your team. 38% of your team (including you) were thieves! You are 20% of your team. So NOT counting you, 18% of your team were thieves.

There’s nine classes in GW2, so lets assume one ninth of the gw2 pvp population plays thief. (The exact proportion of people who play thief really doesn’t matter for this illustration, so don’t get hung up on this assumption).

In every game there is spooko + 9 other people. So on average per game, spooko should expect to see one other thief. 4 times out of 9, that thief will be on your team. That means on average your team will have more thieves roughly 44% of the time (4/9).

Spooko stated that 47% of the time his own team has more thieves. That is to be expected!

Everything that spooko has stated is more or less just expected values plus some deviation. There is nothing profound in his results.

(edited by Barnesy.5839)

Matchmaking By Way of the MATHS

in PvP

Posted by: Jack Daniels.5270

Jack Daniels.5270

Last match 3 thief in my team vs 3 necro in other team. Camped at spawn all game.

Matchmaking By Way of the MATHS

in PvP

Posted by: Gunlaugr.2716

Gunlaugr.2716

‘’32 is statistically significant’’ – What nonsense is this?

Your hypothesis is that the system unevenly distributes classes. There is not a specific minimum sample size required to test this hypothesis. The greater the sample size, the greater the precision of the test.

Depending on the distribution of the data points you collect, a sample size of 32 would or would not be enough to disprove the null hypothesis (no difference between equal class distribution and the group of data points you’ve collected).

If the class distribution is massively skewed, you would be able to disprove the null hypothesis with a sample size less than 32.

Matchmaking By Way of the MATHS

in PvP

Posted by: insanemaniac.2456

insanemaniac.2456

i doubt any of your “armchair” critics are going to watch this..

but 32 is statistically significant…….

if i played other classes than thief my results might have been different, we will never know. i play thief, i expect a decent match when i play a thief. i have no interest in how my matchmaking would play out differently if i played on other classes, because i have no interest in other classes.

so, instead of dealing with hypotheticals, i got down to business and actually put in the work. the work shows that matchmaking is broken AF. this data is not hypothetical data from a bubble, it is the observed results of 32 actual matches. regardless of what hypotheticals you may believe influence the data, this is how the data plays out. this is how the matchmaking actually works. and it is totally broken.

you are misunderstanding this concept.

if you were to play the same match with the same people 32 times, it would be a statistically significant sample size to say something about the conditions.

since you were playing against different people, you are changing test conditions in such a way that there can be no test.

youll need to play about 30 games against each specific composition, or at least against extremely similar compositions where maybe only 1-2 people change or 1-2 classes change to similar (but different) classes/roles.

additionally, when your error bars are like +-3% when 2 classes are showing up ~8% of the time, the true numbers for frequency could easily be 5% and 10%, or in other words, one class could actually show up twice as often as another while you theorize they show up at the same rate and its all within the error of your sample. that means your sample is not statistically significant.

JQ: Rikkity
head here to discuss wvw without fear of infractions

(edited by insanemaniac.2456)