Guild Wars 2

Hello, I recently decided to start capturing data from Solo Queue matches that I played in. I just completed my 100th data capture, and thought I’d share the results. At the least, I hope you find it interesting, here are some details on the data.

Method: I initially started doing screen-caps at the end of a match, but I soon realized that players often left matches before I hit print-screen, so I started doing a screen cap at the beginning and end of the match, so that I had more complete data, there are still some classes listed as “unknowns” where I forgot to screen-cap at the beginning of the match. There are also cases where I may not have noticed a player switch classes mid match, but I did capture some of those.

There are a total of 998 classes because of 2x 5v4s. There were more 5v4s than that, but I only counted it if the player was gone at the beginning of the game because that was an easier call to make. On the redistribution graph, rounding put the total classes up to 999. Go figure.

I mostly played between 6-9pm Server time, all games are on NA Servers and are Solo Queue. My rank ranged from 80th percentile at the beginning of the 100 matches and then traveled between there and the lower top 1k (at least every time I checked), the last 30 or so matches I was in the Top 1k, even sneaking up into the 400+ where I don’t belong.

I used the screen caps to record my personal score, the score of the game, the map played and the class distribution. The data was recorded by hand, so I expect I made a mistake or two, might have even put the losing team as the winners and the winning team as the losers. But I did go through all the scores and maps a second time, so I think it is pretty accurate.

I had 1 loss that I forgot to screen-cap, so that is excluded.

I don’t want to make too many judgments about the data, but I did provide some captions explaining the graphs with some of my thoughts. Keep in mind, this data may or may not be representative of the Solo Queue map population as a whole. I’ll probably continue to do this and add to my database just to see if/how it changes over time. Maybe I’ll update this post in the future with more samples added in.

Enjoy! Let me know if you have any questions or find any mistakes.

wow! nice research!
thanks for sharing!

hmmm looking forward to your future test results!

I’d like to see this with a player that isn’t below average; aka, 56%+ winrate.

Oxygen.5918:

I’d like to see this with a player that isn’t below average; aka, 56%+ winrate.

Yeah, when the gods of Guild Wars 2 established a 56% winrate as ‘above average’ and anyone with a 54.6% as below average.

Jesus Christ.

Also, to the original poster: Nice research! I spent a lot of time around where you are on the leaderboards and necromancers seemed really, really popular. Now I’m floating between 100-300 on team queue/solo queue on both my accounts and I’m not noticing nearly as many necros.

Interesting data, thanks for posting.

Something like this probably would have saved you a lot of time, hey.

Pixels: thanks for compiling this! Nice to see a bit of numerical data on class popularity and balance and I look forward to any updates – especially if we can get some data pre- and post-patches, when the largest shifts in the meta will occur.

As an EU player, you’ve inspired me, and I’ma start trying to get my own screencaps and seeing whether things are much different here – obviously a much larger sample than just 100 games will be needed for proper comparison, but a couple of months of steady play should provide that, especially if anyone else wants to help out? >:D

It’s nice research, but the problem is that every single one of these matches had the same player in them as well; it skews the numbers. Still, very interesting, appreciate that you’re sharing the data nonetheless.

Oxygen.5918:

I’d like to see this with a player that isn’t below average; aka, 56%+ winrate.

Not sure what your point is here.

Average win rate should be about 50% over time now that the MMRs have settled some.

Oxygen.5918:

I’d like to see this with a player that isn’t below average; aka, 56%+ winrate.

Someone with a 56% win rate since the leaderboard wipe would probably be ensconced in the top 50 if they played a lot. I can pretty confidently say you probably don’t have a 56% win rate since the matchmaking changes and wipe unless you have played a very small amount of games. The winning percentages shown still take into account the bloated w/l records from before the changes. Just because you have a greater then 56% win percentage doesn’t mean you have been winning 56% of the time since November 26th. Anything around 50% is where the vast majority of people should be even you wannabe pros. It’s how its designed to be.

good stats^^

but is this NA or EU? And how much was 4v5? cause when i read forum everyone but me has all day 4v5 games^^

Good job Pixels, like the data.

I especially appreciate the 2nd graph down on the left, where you take yourself out of the equation to see how popular professions are. Seems to line up with reality pretty well, Warriors and Necros are the most played soloQ professions.

+1.

Thanks for posting this, really cool stuff.

Just a suggestion though: plotting score vs. wins/losses would look great as two separate PDF or CDF plots.

Also have you considered posting your raw data? Might be fun for other people to poke around in.

This is amazing data. Thanks for doing this!

Thanks for the replies, glad some of you found it useful. I’ll continue to update the master file, maybe after 200 matches, I’ll post an updated version to see how things have changed.

Again, all this stuff was done during NA prime-time (Except for a few matches) on NA servers. I should add that these 100 matches took place between Dec 16, 2013 & Jan 2, 2014.

If Anet wants to drop me a csv file with the data, say 10k NA and 10k EU team and solo matches with MMR to gauge where the matches to place in the ranks, as well as class & rank data, individual scores and total scores, as well as match durations… I’d be happy to do more robust analysis… :o)

Brigg.6189:

Thanks for posting this, really cool stuff.

Just a suggestion though: plotting score vs. wins/losses would look great as two separate PDF or CDF plots.

Also have you considered posting your raw data? Might be fun for other people to poke around in.

Ask and you shall receive. It is attached. That is a tab-delimited file, this forum doesn’t allow csv format.

Silentshoes.1805:

Oxygen.5918:

I’d like to see this with a player that isn’t below average; aka, 56%+ winrate.

Not sure what your point is here.

Average win rate should be about 50% over time now that the MMRs have settled some.

If you are 100% certain that you are not the reason that you are losing (aka, always doing positively good), then your winrate should be at 56%. This rate takes into account the non-probability that you are a negative factor upon your team and that the other team are getting the deadweights instead.

9 (potentially bad players) /10 (potentially bad players + 1, you) = 1.11…

50*1.11… = ~56%

Anything above 50% is a legitimately good win rate to have in a 1v1 game, but this isn’t a 1v1 game.

This work similarly the other way around; bad players are most likely going to hover at or below 44%.

Very good players (someone who consistantly 2v1’s and wins) could see win rates much higher, and very bad players… you see where this is going.

The reason I was looking for someone that was statistically good (56!) was to see the distributions in the higher end of the spvp skill curve. Even if the leaderboards don’t mean anything, the MMR may be there.

Great data and it kinda confirms what I’ve been stating on these boards.

Warriors and Necros make you win.

Dee Jay.2460:

Great data and it kinda confirms what I’ve been stating on these boards.

Warriors and Necros make you win.

the data only shows that warriors and necros are more popular when compared against other professions in 100 games of solo arena at the moment, where does it says that it makes us win? O_O

if we look at the first graph on the top row, left side.

it says

warrior
90 wins
80 loses

necromancer
79 wins
66 loses

if warrior are truly overpowered, we will see the warrior wins number to be much more higher. not only a difference in 10.

Deimos Tel Arin.7391:

Dee Jay.2460:

Great data and it kinda confirms what I’ve been stating on these boards.

Warriors and Necros make you win.

the data only shows that warriors and necros are more popular when compared against other professions in 100 games of solo arena at the moment, where does it says that it makes us win? O_O

if we look at the first graph on the top row, left side.

it says

warrior
90 wins
80 loses

necromancer
79 wins
66 loses

if warrior are truly overpowered, we will see the warrior wins number to be much more higher. not only a difference in 10.

That’s because the sample is small, you indiscriminately unfit for analyzing statistics biased person.

Oxygen.5918:

Deimos Tel Arin.7391:

Dee Jay.2460:

Great data and it kinda confirms what I’ve been stating on these boards.

Warriors and Necros make you win.

the data only shows that warriors and necros are more popular when compared against other professions in 100 games of solo arena at the moment, where does it says that it makes us win? O_O

if we look at the first graph on the top row, left side.

it says

warrior
90 wins
80 loses

necromancer
79 wins
66 loses

if warrior are truly overpowered, we will see the warrior wins number to be much more higher. not only a difference in 10.

That’s because the sample is small, you indiscriminately unfit for analyzing statistics biased person.

yes, because it is only 100 solo arena games.

hmmm by the way what does " indiscriminately unfit for analyzing statistics biased person" mean? and are you referring to me? since i saw a “you” in front.

sorry english is not my first language.

Oxygen.5918:

Silentshoes.1805:

Oxygen.5918:

I’d like to see this with a player that isn’t below average; aka, 56%+ winrate.

Not sure what your point is here.

Average win rate should be about 50% over time now that the MMRs have settled some.

If you are 100% certain that you are not the reason that you are losing (aka, always doing positively good), then your winrate should be at 56%. This rate takes into account the non-probability that you are a negative factor upon your team and that the other team are getting the deadweights instead.

9 (potentially bad players) /10 (potentially bad players + 1, you) = 1.11…

50*1.11… = ~56%

Anything above 50% is a legitimately good win rate to have in a 1v1 game, but this isn’t a 1v1 game.

This work similarly the other way around; bad players are most likely going to hover at or below 44%.

Very good players (someone who consistantly 2v1’s and wins) could see win rates much higher, and very bad players… you see where this is going.

The reason I was looking for someone that was statistically good (56!) was to see the distributions in the higher end of the spvp skill curve. Even if the leaderboards don’t mean anything, the MMR may be there.

But the whole point of MMR is to assign you with players of equal skill: not dead weight.

As for the skill curve, I’d like to see all of it. But I don’t have access to that data, I only know what my rank was through out the matches, it ranged from 80th percentile up to 400+, as of right now, I’m in the 300s on the LB. Some of the data I provided includes top tier players, one of the top 3 was on my team the other night.

I would be interested in seeing how class distribution shifts along the skill curve, if it does, so again, if ANET wants to provide it… I’ll take a peak.

Thanks for the work! At least it paints some kind of picture for us players about the grand scheme of things.

Sadly ArenaNet will never release such information to us, and they have the entire statistical database of all matches since the game’s release

We clearly see necro imba win 20% more than looses. Warrior same

Oxygen.5918:

If you are 100% certain that you are not the reason that you are losing (aka, always doing positively good), then your winrate should be at 56%. This rate takes into account the non-probability that you are a negative factor upon your team and that the other team are getting the deadweights instead.

That assumption is extremely flawed and you know it. One can’t simply assume that they never contribute negatively to a team, nor can one say that they never contribute positively to a team. Viewing it from a statistics standpoint 50% means you contribute equally negatively and positively to the teams you are on, but even that is an assumption. There is far to many random variables to find a percentage that works well all the time. A simple estimation like you or I made holds very little actual baring on the skill of a player.

Oxygen.5918:

Silentshoes.1805:

Oxygen.5918:

I’d like to see this with a player that isn’t below average; aka, 56%+ winrate.

Not sure what your point is here.

Average win rate should be about 50% over time now that the MMRs have settled some.

If you are 100% certain that you are not the reason that you are losing (aka, always doing positively good), then your winrate should be at 56%. This rate takes into account the non-probability that you are a negative factor upon your team and that the other team are getting the deadweights instead.

9 (potentially bad players) /10 (potentially bad players + 1, you) = 1.11…

50*1.11… = ~56%

Anything above 50% is a legitimately good win rate to have in a 1v1 game, but this isn’t a 1v1 game.

This work similarly the other way around; bad players are most likely going to hover at or below 44%.

Very good players (someone who consistantly 2v1’s and wins) could see win rates much higher, and very bad players… you see where this is going.

The reason I was looking for someone that was statistically good (56!) was to see the distributions in the higher end of the spvp skill curve. Even if the leaderboards don’t mean anything, the MMR may be there.

It would be true if you are at a low enough matchmarking than the game can be split into deadweight and doing something.

However, take into account that the sample is small -as you said- that means that win rate may vary, because of its random component.

amazing data – anet needs to hire you and kick their balancing team. Not sure what these guys are doing maybe kittening and drooling all day.

Edit: nvm i missed your class choice

Time is just so important in this game.

Also look at the engineer stats xD 60 losses 40 wins and when theres 2 engies the win/loss ratio is the worst of them all only winning half as many games as losing xD

kitten this faceroll class that carries everyone while rolling your face over the keyboard.. so true xD

If u played sc2 u know that the win loss in a perfect world should be 50%. That’s because it also takes into account ur MMR and so in the enemy team there should be a player with somewhat equal skill as yours. In that game i had consistenly 48-52%, even when i improved a lot and reached higher leagues.
There can only be 2 reasons for a huge group of players not having a close to 50% w/l ratio, either the matchmaking is badly designed or the playerbase is low enough that matchmaking fails most of the time or both.

But, solo queue and even team queue when pugging, is more about having a build that can tank and do damage and some support, than actual skill, indeed one of the best recomendations to achieve higher mmr is to play a greatsword guardian or a warrior.

Baldric.6781:

If u played sc2 u know that the win loss in a perfect world should be 50%. That’s because it also takes into account ur MMR and so in the enemy team there should be a player with somewhat equal skill as yours. In that game i had consistenly 48-52%, even when i improved a lot and reached higher leagues.
There can only be 2 reasons for a huge group of players not having a close to 50% w/l ratio, either the matchmaking is badly designed or the playerbase is low enough that matchmaking fails most of the time or both.

But, solo queue and even team queue when pugging, is more about having a build that can tank and do damage and some support, than actual skill, indeed one of the best recomendations to achieve higher mmr is to play a greatsword guardian or a warrior.

Well, in a situation like that you’ll achieve 50% in an infinite amount of plays.

Considering no one can achieve that result, your win loss ratio in such situation will be represented by a Gauss distribution. That means there is a random factor in it. The more the games, the narrower the bell curve.

yolosmurf.8052:

amazing data – anet needs to hire you and kick their balancing team. Not sure what these guys are doing maybe kittening and drooling all day.

Edit: nvm i missed your class choice

Time is just so important in this game.

Also look at the engineer stats xD 60 losses 40 wins and when theres 2 engies the win/loss ratio is the worst of them all only winning half as many games as losing xD

kitten this faceroll class that carries everyone while rolling your face over the keyboard.. so true xD

I apologize, but I’m not sure if you’re serious, but I’ll answer you as if you are.

ANET could hire me… I wouldn’t mind, but the PvP and Balance team have a lot more data than is presented here, they don’t have samples, they actually have a census and, I assume, make balance decisions with holistic data while taking into account other game modes.

These 100 matches are a convenience sample in which I participated. They may indicate class distribution for a very specific time period on NA Servers for players with MMR close to mine. Not sure you can extrapolate too much beyond that.

I posted this for fun, and I’ll update it in the future. Also, I wouldn’t worry about the W/L records by class too much, over time (i.e., larger sample), those should approach 50% because classes end up playing each other often enough that even if they’re “OP,” they counter each other. I put it in to see how it works out, I already saw it a bit as Necro win % dropped down as time went on (if I remember correctly).

The key to seeing what the population thinks is OP is in class selection, right now, the player base I played with for those 100 matches felt that Necros and Warriors give them the best chance to win, and that kind of matches up with what you read here, esp with Warriors.

Again, I feel pretty confident that this sample is representative of NA, Solo Queue during 6-9pm Server time at my MMR. Beyond that, I have no idea, esp since I participated in the sample.

It would be nice; however, if ANET and someone with some coding knowledge worked out a system similar to the Black Lion Trading post websites that can grab snap-shots of data from the server and create much larger data points from all different times and servers. Not sure that’s possible.

TL/DR: ANET has better data.

Pixels.6532:

TL/DR: ANET has better data.

Thats what i’ve been telling myself since release… but after all this time and A LOT of completely random nerfs and buffs. Im not so sure anymore.

Edit: I work a lot with randomized trials as well so i know 100 is a very small sample size and all that yada yada yada. But either way you listed all of that which u did perfectly soo…

Still thanks for compiling this list!

yolosmurf.8052:

Pixels.6532:

TL/DR: ANET has better data.

Thats what i’ve been telling myself since release… but after all this time and A LOT of completely random nerfs and buffs. Im not so sure anymore.

Edit: I work a lot with randomized trials as well so i know 100 is a very small sample size and all that yada yada yada. But either way you listed all of that which u did perfectly soo…

Still thanks for compiling this list!

Hey, no problem, glad you found it interesting.

zapv.8051:

That assumption is extremely flawed and you know it. One can’t simply assume that they never contribute negatively to a team, nor can one say that they never contribute positively to a team. Viewing it from a statistics standpoint 50% means you contribute equally negatively and positively to the teams you are on, but even that is an assumption. There is far to many random variables to find a percentage that works well all the time. A simple estimation like you or I made holds very little actual baring on the skill of a player.

The assumption that some players are always better than other players is flawed? Would you like some quick duels to settle this one? Bring a friend, or 4.

Oxygen.5918:

zapv.8051:

That assumption is extremely flawed and you know it. One can’t simply assume that they never contribute negatively to a team, nor can one say that they never contribute positively to a team. Viewing it from a statistics standpoint 50% means you contribute equally negatively and positively to the teams you are on, but even that is an assumption. There is far to many random variables to find a percentage that works well all the time. A simple estimation like you or I made holds very little actual baring on the skill of a player.

The assumption that some players are always better than other players is flawed? Would you like some quick duels to settle this one? Bring a friend, or 4.

How did you end up so low in the leaderboard you are better than most other players and easily win 2v1s?

And there’s a different between fighting capability and the ability to make the right decision during a match.

redslion.9675:

Oxygen.5918:

zapv.8051:

That assumption is extremely flawed and you know it. One can’t simply assume that they never contribute negatively to a team, nor can one say that they never contribute positively to a team. Viewing it from a statistics standpoint 50% means you contribute equally negatively and positively to the teams you are on, but even that is an assumption. There is far to many random variables to find a percentage that works well all the time. A simple estimation like you or I made holds very little actual baring on the skill of a player.

The assumption that some players are always better than other players is flawed? Would you like some quick duels to settle this one? Bring a friend, or 4.

How did you end up so low in the leaderboard you are better than most other players and easily win 2v1s?

And there’s a different between fighting capability and the ability to make the right decision during a match.

I don’t play a lot, so my leaderboard spot isn’t representative of anything. Besides, I was just illustrating.

Oxygen.5918:

redslion.9675:

Oxygen.5918:

zapv.8051:

That assumption is extremely flawed and you know it. One can’t simply assume that they never contribute negatively to a team, nor can one say that they never contribute positively to a team. Viewing it from a statistics standpoint 50% means you contribute equally negatively and positively to the teams you are on, but even that is an assumption. There is far to many random variables to find a percentage that works well all the time. A simple estimation like you or I made holds very little actual baring on the skill of a player.

The assumption that some players are always better than other players is flawed? Would you like some quick duels to settle this one? Bring a friend, or 4.

How did you end up so low in the leaderboard you are better than most other players and easily win 2v1s?

And there’s a different between fighting capability and the ability to make the right decision during a match.

I don’t play a lot, so my leaderboard spot isn’t representative of anything. Besides, I was just illustrating.

I wasn’t trying to imply you are bad because you don’t have a high place in the leaderboard. I’d be no one to judge.

What I mean is that your win ratio depends on more issues, such as where you come from: sometimes it happens that a player starting soloQ enters the leaderboard in too high of a position (assuming the leaderboard actually works as intended) and then eventually falls down to his actual level. We call this player A.

Player B started from a low point in the leaderboard, and eventually went up to his level.

Now, let’s assume A and B are equally good. A will have a lower win ratio than B, because A has been stomped by veteran, and B has stomped rookies. Now they are at the same level, though, and both roughly maintain their position until they get better.

The higher the number of matches played, the more precise this mechanism becomes.

redslion.9675:

And there’s a different between fighting capability and the ability to make the right decision during a match.

Finally someone with brains in this forum. I always tell people that judge me for playing staff ele before the match starts that PvP isn’t all about the class you play – that’d be the case in a simple deathmatch – but we have conquest, and therefore it’s also about general understanding und the decisions you make.