Curse

Darth The Xx · Mar 20, 2009, 08:39 AM // 08:39

So I'm sure were all familiar with this post up the top of the forum. But for those who arn't

Quote:

If you think your side of preference is better than the other, by all means, keep believing it. However, unless you're actively keeping statistics on which side is winning more, please refrain from posting claims that one side is better than the other, or from stereotyping members of either faction.

At best, it's e-peen flexing. At worst, it leads to arguments between two people based on nothing more than unverifiable anecdotal personal experience. None of this leads to any form of intelligent discussion.

As such, posts or threads based on such claims will be deleted.

So I saw this and decided hey, why the hell not keep some stats.

So what I decided to do over the past 3 or so weeks I have been recording the scores of all AB games I've been in, to be fair the stats are from 100 games. 20 from each map.

I realise there is no way to validate this data but trust me, there's no way i'd go though this much effort to make fake results. Do take into account that I usually run with a friend or guild group but I PuG maybe 40% of time or so. While some people may not believe the results, these are the kind of outcomes that I get when I AB.

If you don't wanna read through all the results scroll down to the bottom where all the stats are listed. You will see quite clearly which side is better, for those who can't be bothered looking at the table here are the most important stats. There is also a graph showing what % each map was played, and another one showing Luxon vs Kurzick maps.

Average Faction Luxon/Kurzick

Ancestral Lands 1528.35 / 997.4
Grenz Frontier 1421.1 / 797.4
Saltspray Beach 1016.85 / 726.9
Etnaran Keys 1214.8 / 852.4
Kaanai Canyon 1449.65 / 751.4
Overall 1326.15 / 825.1

Average Score Luxon/Kurzick

Ancestral Lands 359.45 / 447.4
Grenz Frontier 410.55 / 397.4
Saltspray Beach 416.85 / 376.9
Etnaran Keys 466.8 / 313.7
Kaanai Canyon 499.65 / 233.8
Overall 430.66 / 353.84

Winrate (Luxon)

Ancestral Lands 0.45
Grenz Frontier 0.6
Saltspray Beach 0.65
Etnaran Keys 0.75
Kaanai Canyon 0.95
Overall 0.68

PS : I am luxon

EDIT: Outliers were any games where either side scored less than 100 points and the scores were not recorded and so do not count towards the final results.

elypta · Mar 20, 2009, 09:00 AM // 09:00

well, first of all gj, I can see your point but got a few things to say about it :P
1. I think most Kurzick players that want the faction are doing JQ so maybe less experienced people in AB.
2. You said you would be running some guild teams so that may differ the scores too. only 12 people and with ~4 organised it goes better then without them

but still a remarkeble result, didnt expected it :P *I am a Kurz*

turbo234 · Mar 20, 2009, 09:11 AM // 09:11

doesn't surprise me too much to be honest. i always seem to win more when playing for the luxon side. i play for both sides btw.

Darth The Xx · Mar 20, 2009, 09:14 AM // 09:14

Quote:

Originally Posted by elypta

well, first of all gj, I can see your point but got a few things to say about it :P
1. I think most Kurzick players that want the faction are doing JQ so maybe less experienced people in AB.
2. You said you would be running some guild teams so that may differ the scores too. only 12 people and with ~4 organised it goes better then without them

but still a remarkeble result, didnt expected it :P *I am a Kurz*

Its true that running with organised teams make a difference, however I'm not always in a team of four, usually just 2-3 and we take extras, even so the difference is huge. I was also very surprised, I always thought it was closer to 50/50.

tmakinen · Mar 20, 2009, 09:34 AM // 09:34

This is interesting and clearly a commendable amount of work has gone into collecting the statistics. However, I am concerned about two different factors of selective bias, those of non-random sampling of play time and teams. The latter is especially hard to compensate for since in order to record the result of a match one must participate in it and therefore the sampling of teams is by definition not random. Non-random sampling of time is likely to happen as well unless you're willing to spread your playtime equally across 24 hours of a day.

There is a way around the observer bias by just keeping track of maps which reflect the balance of power between the sides. Thus, I'd like to ask how you recorded the percentage of each map played? Was it a 'random' sample (whenever you happened to check it) or a systematic one (all maps for a certain period/periods)? Did you take into account that sometimes the map doesn't change at the 3-hour poll time? What was the temporal coverage (there might be a daily cycle and if you repeatedly sample at one particular time of the day the results may be very misleading)?

Darth The Xx · Mar 20, 2009, 09:45 AM // 09:45

Quote:

Originally Posted by tmakinen

This is interesting and clearly a commendable amount of work has gone into collecting the statistics. However, I am concerned about two different factors of selective bias, those of non-random sampling of play time and teams. The latter is especially hard to compensate for since in order to record the result of a match one must participate in it and therefore the sampling of teams is by definition not random. Non-random sampling of time is likely to happen as well unless you're willing to spread your playtime equally across 24 hours of a day.

There is a way around the observer bias by just keeping track of maps which reflect the balance of power between the sides. Thus, I'd like to ask how you recorded the percentage of each map played? Was it a 'random' sample (whenever you happened to check it) or a systematic one (all maps for a certain period/periods)? Did you take into account that sometimes the map doesn't change at the 3-hour poll time? What was the temporal coverage (there might be a daily cycle and if you repeatedly sample at one particular time of the day the results may be very misleading)?

The map percentage was just tagged on at the end. I realise that it is not really that reliable of an indication. It was recorded based on how many games I played on a certain map, the thing was once i passed 20 for grenz but the map kept coming up I thought it a good idea to record how many times past 20 it took before all results for the full 100 were collected.

So basically it is the percentage of games that I played on those particular maps during the data collection.

I didn't even consider time of day to be an issue but now that you mention it I see how that might affect the outcome. However I can only play at one time during the day so there isn't much I can do there

edit: I suppose it is fair to say that these results are only valid for a certain time frame, however a large portion of them were collected on weekends, which is when my play times vary a lot.

Edit2: I ran into a guy who was also collecting stats but in a different way, he was looking at the maps. Ours were collected over the same time period but he was looking specifically at the maps. He posted his stuff here.

http://wiki.guildwars.com/wiki/User:Alleji

lewis91 · Mar 20, 2009, 11:17 AM // 11:17

Wow good job. Im kurz and i admit to loosing more often, i try to make "Organised Teams" but most of the time, the only organised teams you see are all in the same guild.

Majority of teams i go in always has at least 2 assassins, and either a ranger or a warrior. I love to play my warrior in AB, but theres too many of them, but i also love to play ranger, but theres far too many of them aswell, so i usually end up being a monk for a team that runs in different directions, away from the group, and then when theyre out of range and die, they blame me ¬_¬

So i dont play monk often any more. Infact ive stopped playing AB alot all togeather now, JQ i prefere.

zling · Mar 20, 2009, 11:33 AM // 11:33

honestly these results dont mean much at all. first of all AB is very random, a single guild group on either side can swing the battle in their favour. as others have said time of day also affects these results. lastly you, assuming you're at least a decent player, have probably influenced the results yourself by being at least decent, as most people in AB are either entirely clueless about PvP or play random builds for kicks, fun or whatever else they use as an exuse to be bad.

so, while it's interesting to see actual statistics about the issue, there's a reason why that post you quoted was written, and it will need a lot more extensive statistics from different people on different servers(Euro, American, International, Asian) on different times to really get results that are more or less accurate.

glacialphoenix · Mar 20, 2009, 11:38 AM // 11:38

I have to admit that's a commendable effort, but as many others have said, it's biased for a lot of reasons you can't control. Guild teams, number of people playing at that hour... I'm not really surprised by the stats, though. I've played both Kurz and Luxon, and the only surprising one is Grenz Frontier. I usually see wins on my side (Kurzick) when it's in our territory, shallow or deep notwithstanding.

Arduin · Mar 20, 2009, 11:45 AM // 11:45

I wonder if Anet keeps a log of these battles, and is able to show it to the public, much like Valve does (eg. TF2).

Junato · Mar 20, 2009, 12:20 PM // 12:20

Let me just say that you have done a great job implementing statistics. Are you a statician? If not, you should be!

Well, I have taken statistics and the main point in statistics in order to mean anything to a population is how you compare the whole population to a sample. So, just giving statistics to related AB doesn't really mean much if you can not determine how many people play pve, gvg, ab and etc.

own age myname · Mar 20, 2009, 12:45 PM // 12:45

Quote:

Originally Posted by elypta

well, first of all gj, I can see your point but got a few things to say about it :P
1. I think most Kurzick players that want the faction are doing JQ so maybe less experienced people in AB.
2. You said you would be running some guild teams so that may differ the scores too. only 12 people and with ~4 organised it goes better then without them

but still a remarkeble result, didnt expected it :P *I am a Kurz*

You sure? Yesterday I played JQ and won 10 out of 12. 5 or 6 of those games we had a leecher(s), and we still won.

Songbringer · Mar 20, 2009, 12:50 PM // 12:50

I read the OP's whole post and a couple after that but as I am in a rush to go to work I am going to post what I think and if it was already said I am sorry.

When I was luxon I always felt like I won more then when I play as a kurzick

100 games aren't enough to make a statistic

You should have had like 40 people all taking the stats

If the results aren't able to be remade then they can't be taken as truth. From a scientific perspective. So while I won't do it because I don't ab anymore. Simply because I already got Savior of the Luxons and vanquishing>AB. Less idiots faster faction <3. I do suggest and even would push someone to doing it.

However to turn this into a real expirement I suggest this. A group that is doing all PuGs. No running with friends regoing with the same group more then once or or running with guildies. Then a group that is all friendslist. No guildies or PuGs. Then a group made of just Guild and Alliance members. I suppose you could combine the friend and guild group. However if your guildies don't know your friend and don't know how he plays then he might as well be playing with a PuG.

Shayne Hawke · Mar 20, 2009, 12:54 PM // 12:54

Now do it for JQ and FA and we might have something that measures overall PvP ability of both sides. The fact that Luxons beat out Kurzicks in general in AB doesn't surprise me in the least.

How would you measure PvE ability? Who can farm faction the fastest from vanquishing?

Sandros · Mar 20, 2009, 02:21 PM // 14:21

Maybe I'm wrong here but from what I remember from my college statistics classes, for any statistic to be viable in science, you need to gather at least 1000 stats to be 95% +-5% right on any subject. Not also that but the 1000 stats you gather must be gathered in a specific way to have any weight at all (aka make sure its random by representing the most variables possible).

If you didnt observe 1000 games with around the same number of "premade" team on all the maps, at different times of the day, then I doubt your statistics are even close to being right. You would also need to know how much people there is on each side, how much "premade" teams each side run, and who tend to join whatever side, how many people left before the end etc.

Then again I got the feeling luxons win more, but thats because I never play AB alone maybe.

glacialphoenix · Mar 20, 2009, 02:43 PM // 14:43

Quote:

Originally Posted by Sandros

Then again I got the feeling luxons win more, but thats because I never play AB alone maybe.

I've played both sides before. (Don't kill me, people.) Honestly, I think Kurz AB suffered from the nerfing of HFFF - the first few days after that nerf, the influx of people into AB was terrible, and many of them were coming in on the Kurz side - which has more people anyway.

And the more players there are, the less chance you have of getting a good team on your side. I'm assuming all of us here think we have decent teams - but one decent team isn't enough to make up for two terribad teams. I'm pretty sure that if you were to take the statistics from, say, AB weekends, you'd get results pretty different from the OPs.

Edited to add:

I might also add that because of the sheer number of people on Kurzick (significantly more than Lux, I believe) - good teams on Luxon will presumably be playing more rounds than good teams on Kurzick. So, well, quite obviously they're going to factor in turning the tide more frequently. I

Tharg · Mar 20, 2009, 02:53 PM // 14:53

I like the approach - gj.

You don't need a certain specific number of results, although with an increased sample sze the results become statistically more significant. Even with the limited sample size, the statistics on the winrate for the Luxon side (68%) seem statistically significant to me:

Ancestral Lands 0.45
Grenz Frontier 0.6
Saltspray Beach 0.65
Etnaran Keys 0.75
Kaanai Canyon 0.95
Overall 0.68

You would want to run the same test again - same sample size - and compare the second test with the first. This will give you an idea about the reliabiity of the test.

To the guys making the point that one group can change the tide of an AB - that is just the point of the research. The statistics actually shows that the groups having a deciding role play more for the Luxon side than for the Kurzick side.

I don't think that the time of the day / week would affect the sample, unless one would want to make a point that there are better/more Luxons/Kurzicks playing during specific periods. I would say that it's evened out pretty much.

The only thing that is not random is the observer's participation in the AB. If he is not an average player, he is influencing the outcome of the game and therefore tainting the sample.

What the research does NOT prove is that Luxon players are better players than Kurzick players in general. It only shows that Luxon players in AB are better than Kurzick players in AB.

Eragon Zarroc · Mar 20, 2009, 03:03 PM // 15:03

statistics are not as consistant as they would be if u ran with the same team everytime. if u ran with pugs all the time, it would be more likely that the stats would be lower in favor of the luxons. if u ran with guild groups all the time, it would raise the stats in favor of the luxons. it is also impossible to know the experience of the random people u aquire for teams. it would require a much larger sampling for a more accurate result. hundreds of games per map would be preffered. good job though, at least someone decided to take the time to acquire some statistics. it still gives u a good idea that luxons dominate ab battles more than the kurzicks i suppose

TheodenKing · Mar 20, 2009, 03:03 PM // 15:03

Quote:

Originally Posted by Darth The Xx

I didn't even consider time of day to be an issue but now that you mention it I see how that might affect the outcome. However I can only play at one time during the day so there isn't much I can do there

edit: I suppose it is fair to say that these results are only valid for a certain time frame, however a large portion of them were collected on weekends, which is when my play times vary a lot.

My observation of trends in AB (I'm r11 luxon/r6 kurz) is that time of day has a tremendous impact on AB outcome. I can go for 3 or 4 hours without losing a match as a luxon (except on Ancestral Lands they win their share), then there will be a 2 or 3 hour stretch when the Kurziks absolutely dominate us.

I'm impressed with your effort, but I think for your data to be more relevant, you need to do xx amount of matches during specific hours. I also think it should be done with pugs... no help from a friend, except that your friend could help you collect data by also pugging. If your friend IS going to play in your group, he needs to be with you for all the matches, not just some.

I'm working on other things lately, but when I get back to more ABing, I might just do some data collection playing on both sides of each map.

Krill · Mar 20, 2009, 03:15 PM // 15:15

Doesn't surprise me either, although if you wanted to be more scientific you'd need to play a similar amount from the Kurzick side and note the results. When I use to play AB with a Kurz guild I bet our overall win % was similar to yours, perhaps a little higher on Grenz (compared to your Keys, very rarely lost) and a little lower on Canyon (compared to your Ancestral). But like TheodenKing said, time of day matters quite a bit because decent players tend to leave if their side keeps losing or stay for multiple rounds, often hours if their side is doing well.