Lets just get this right out.  Today, 10/30/08, I predict based on xfire data and a single data point on subs of a peak at 750k.  I say that today War has between 350-450K subs.    Further I'll say that as the number in xfire drops from a peak for the week of 40k currently to a lower number the percentage of subs will correspondingly drop.

So lets watch for some real numbers or look at mmodata and see what numbers are forthcoming you can then compare them to the numbers i've given here and if the numbers are off then you can put to rest the assumption that xfire is a good predictor of subs.

Stop using xfire to try and guess subscription amounts, it's not used by all gamers, and some gamer types use it more than others.

Look at xfire use for EVE Online, which is what, 200-250k subs?  by that comparison, Warhammer would have well over 500k active subs still, hell it'd have had more active subs a few weeks ago than boxes it shipped.

Originally posted by Malthros

Stop using xfire to try and guess subscription amounts, it's not used by all gamers, and some gamer types use it more than others.

 

Look at xfire use for EVE Online, which is what, 200-250k subs?  by that comparison, Warhammer would have well over 500k active subs still, hell it'd have had more active subs a few weeks ago than boxes it shipped.

You can't use xfire to do sub counts between games without a lot of potential error.  It gives you a rough sense but like you point out with more error margin.  However for the numbers of a specific game it is quite accurate.  ie If you see Eve online xfire numbers double in 6 months you can say the sub count doubled and be very accurate.  NOT exact because there is always a margin of error but you will be accurate to within maybe + or - 10%.

I haven't seen a game drop like this in a long time it's now at number 12 and by the looks of it it's still droping, I thought this game would at least beat Guild wars but by it's not even doing that now.

Originally posted by LondonMagus

Quite correct, but the most important point is that all the mathematical calculations about survey reliability based on 'Actual Sample Size' Vs 'Total population Size' are compeletly dependant on the sample being truly representative. If the sample is biased then everything else is meaningless. 

Before it is possible to use any sample to predict future activity, you first have to demonstrate that it is an unbiased & representative sample of the population you are trying to measure.

Unfortunately the Xfire advocates just dont seem to be able to grasp that.

Just my opinions of course.

Thank you, thank you and THANK YOU! I have been trying to get people to understand this for months only to be told "sample size does not matter" or "you don't know statistics", etc....

Since there is no way to prove if the sample of people that uses X-fire it truly representative of the population you want survey (i.e. - all MMO gamers), then any data culled from that sample is worthless.

Not to mention there are way too many variables that you must account for when surveying, one would be lost trying to forecast any type of conclusion based on the data.

I just wanted to say thanks LondonMagus - I am glad someone out there "gets it".

Originally posted by phluux

Originally posted by megafluxmega

im thinking the vast majority of xfire users are people who just click their way through installs in general and probably have comrade, weatherbug, and several other not so useful things installed as a result of that fact. THOSE are the people we are supposing represent gamers in general? hmm....i gotta say to me it sounds like the only thing xfire could measure is "people in too much of a rush to uncheck a box who play games".

use visa, use internet providers, use anything that provides solid data, dont use a cross section that represents the fanboy hordes. shit, fanboys dont even count as gamers, they jump from 1 fad to another; it is like counting fleas on a dog.

 

Right on. Don't expect any of the knuckleheads here to understand that though.

Explain again.

A player that never played and installed WAR and uses XFires.
How do that person affect whatever trend WAR is showing on Xfire?

If 100 out of 1000  people playing a certain game uses Xfire. And 20 of those quit. Please convince why I should believe that the, relative, number of quitting Xfire uses are more or even the opposit trend then the rest of WAR players.  Megafluxmages quoted post did not convince me.

Originally posted by templarga

Originally posted by LondonMagus

Quite correct, but the most important point is that all the mathematical calculations about survey reliability based on 'Actual Sample Size' Vs 'Total population Size' are compeletly dependant on the sample being truly representative. If the sample is biased then everything else is meaningless. 

Before it is possible to use any sample to predict future activity, you first have to demonstrate that it is an unbiased & representative sample of the population you are trying to measure.

Unfortunately the Xfire advocates just dont seem to be able to grasp that.

Just my opinions of course.

 

Thank you, thank you and THANK YOU! I have been trying to get people to understand this for months only to be told "sample size does not matter" or "you don't know statistics", etc....

Since there is no way to prove if the sample of people that uses X-fire it truly representative of the population you want survey (i.e. - all MMO gamers), then any data culled from that sample is worthless.

Not to mention there are way too many variables that you must account for when surveying, one would be lost trying to forecast any type of conclusion based on the data.

I just wanted to say thanks LondonMagus - I am glad someone out there "gets it".

What do you mean with biased? Do gamers play whatever game they play to intentionally raise the Xfire numbers and then they choose to not play a game they don't like with the intent to not raise the Xfire number.

Hmm... wait

Bias has nothing to do with intent. Bias can be anything. What if X-fire is used by people who play single player games more than MMO's? What if X-fire is used more by people in the US than in other areas of the world? What if the average X-fire players plays 6 hours a day but the average MMO'er plays 3 hours a day? What if they average age of a gamers according to X-fire is 18 but the real average age is 28?

Any of this can skew your data and make it worthless if you cannot: a - prove the bias; b - take it into account with your poll and prediction.

I could keep going but I think you get the idea. Bias can be anything at all and many times there is no intent behind, it is simply bias. Until all possible bias is accounted for and taken into account by the pollster, any data gathered is unreliable.

Originally posted by templarga

Bias has nothing to do with intent. Bias can be anything. What if X-fire is used by people who play single player games more than MMO's? What if X-fire is used more by people in the US than in other areas of the world? What if the average X-fire players plays 6 hours a day but the average MMO'er plays 3 hours a day? What if they average age of a gamers according to X-fire is 18 but the real average age is 28?

Any of this can skew your data and make it worthless if you cannot: a - prove the bias; b - take it into account with your poll and prediction.

I could keep going but I think you get the idea. Bias can be anything at all and many times there is no intent behind, it is simply bias. Until all possible bias is accounted for and taken into account by the pollster, any data gathered is unreliable.

Yeah. I just want to move the focus in comparing to other games. A game that is lower on the chart on Xfire could very well have more players then one that is higher and so on.

And I have to agree with some things. Just because the Xfire number goes down doesn't have to mean it is loosing subscribers. It could very well be the same amount of subscribers, I mean some quit and others join. And those that stay have settled and the new intruiging game thing have wore off. They have done their share of leveling and are following a "schedule" of some kind, I don't know.

I don't know what to believe really when it comes to the Xfire thing.

where can i get something to measure how many people install, UNINSTALL, and patch xfire?

idiocy lads, there is nothing to explain because you just dont get it.

why not just take a COMPLETELY useless cross section...like fkn mac users?

Originally posted by strategy

Originally posted by LondonMagus

Originally posted by strategy

Xfire is used by PC game players and these players use the internet.

As  long as there are a few thousand WAR players on it, it is a good indicator.

So the drop is too significant.

It is also in accordance with what we see on the server populations. Half of the servers are on low/low in prime time...

Even GW surpassed WAR yesterday. So it is not a problem of "lesser" MMO play, it is a problem of lesser "WAR" play.

http://www.xfire.com/games/who/Warhammer_Online_Age_of_Reckoning/

Ignoring this can only be viewd as a fanboy defense.

So if for arguments sake the few thousand players were composed entirely of retired clowns would that still be OK then?

It is never statistically valid to argue that the size of a sample is more important than how it was gathered.

Ignoring this can only be viewed as a politician's defence. 

 

Sorry I only now saw your remark of "clowns".

Well of course., why should clowns not count as gaming is concerned? .. the only thing we are interested in is the number of players that play WAR.

No matter what their profession. You see, Xfire measure people that play PC games with Xfire installed. That's all, race, cult, profession doesn't matter.

These guys live in our western world, have installed War on their machines and are quite up front in their games. Look at the position of brand new games (only published this week) and bang they show up on Xfire.

A pre patch for a Wow expansion and hop you see the trend on Xfire.

5 weeks ago 15000 of these guys were playing War, now that number is down to 7500 or even less.  We aren't analysing a social study, we look at how much Xfire players play PC games.

Samples of this order (into the thousands) show more than anyone could hope for.

Lucky, because otherwise FunCom would still brag about millions of subscribers. Now we have at least a sample of a few hundred thousand players if all games are counted.

If that sample would mean that only clowns were playing, no problem. PC game players are PC game players - on line - in our world and with the game installed they are playing.

No matter if they are bold, young or old. BTW What IS your typical WAR player 

The 'Retired Clowns' remark was of course in jest, but carefully chosen because of all minority groups at least they should be able to take a joke.

When you commented "That's all, race, cult, profession doesn't matter", I wasn't sure if you were joking because the point I was trying to make is so fundamentally obvious to me that I honestly can't understand why you still can't see it.

For example, imagine you were a confectioner trying to decide on a new range of up-market desserts & had just created potentially the most delicious pastry ever but which involved a lot of custard. If your test sample audience bizarrely consisted mainly of 'Retired Clowns' who were so heavily traumatised that the mere mention of custard made them burst into tears, you might decide it was a failure & miss the chance at inventing the most famous dessert ever made.

Obviously that is a silly example, but it does illustrate how demographic bias in samples can lead to grossly distorted results. To use a more serious example, the American elections are fast approaching & judging by the UK coverage, the news over there must be full of opinion polls & predictions. Newspapers the world over are known to have heavy political bias, so if two newspapers print completely contradictory predictions based on very large samples taken from their own readership, which is correct?

As for what is a typical WAR player, I don't know because no-one has conducted a decent survey yet to find out, but that still doesn't mean that is valid to assume any old biased sample will do.

Originally posted by Orphes

Originally posted by templarga

Originally posted by LondonMagus

Quite correct, but the most important point is that all the mathematical calculations about survey reliability based on 'Actual Sample Size' Vs 'Total population Size' are compeletly dependant on the sample being truly representative. If the sample is biased then everything else is meaningless. 

Before it is possible to use any sample to predict future activity, you first have to demonstrate that it is an unbiased & representative sample of the population you are trying to measure.

Unfortunately the Xfire advocates just dont seem to be able to grasp that.

Just my opinions of course.

 

Thank you, thank you and THANK YOU! I have been trying to get people to understand this for months only to be told "sample size does not matter" or "you don't know statistics", etc....

Since there is no way to prove if the sample of people that uses X-fire it truly representative of the population you want survey (i.e. - all MMO gamers), then any data culled from that sample is worthless.

Not to mention there are way too many variables that you must account for when surveying, one would be lost trying to forecast any type of conclusion based on the data.

I just wanted to say thanks LondonMagus - I am glad someone out there "gets it".

What do you mean with biased? Do gamers play whatever game they play to intentionally raise the Xfire numbers and then they choose to not play a game they don't like with the intent to not raise the Xfire number.

Hmm... wait

The bias doesn't refer to deliberate attempts to distort the Xfire stats for a given game, rather it concerns inherently atypical behaviour in non-representative samples.

Please see the Wikipedia entry below as it explains it in far greater detail than I have the patience for.

http://en.wikipedia.org/wiki/Sampling_bias 

Originally posted by templarga

Bias has nothing to do with intent. Bias can be anything. What if X-fire is used by people who play single player games more than MMO's? What if X-fire is used more by people in the US than in other areas of the world? What if the average X-fire players plays 6 hours a day but the average MMO'er plays 3 hours a day? What if they average age of a gamers according to X-fire is 18 but the real average age is 28?

Any of this can skew your data and make it worthless if you cannot: a - prove the bias; b - take it into account with your poll and prediction.

I could keep going but I think you get the idea. Bias can be anything at all and many times there is no intent behind, it is simply bias. Until all possible bias is accounted for and taken into account by the pollster, any data gathered is unreliable.

LOL, you sound like you hate bad statistics as much as I do. 

         I found the Ops post to be fairly educational myself but there are a couple of things I have to question......First off is the hours played thing......Ive seen players that were listed at 168 hours for the previous week in a game (Literally 24/7) and Ive seen others with zero hours played of anything (I think almost all of us play something for a few minutes  at least once a week)......Second I didnt know that it was mainly intended for FPS games and that may explain why I had never seen it before in any of the games I have bought (Not an FPS fan myself)......Third he claims that xfire has 11 million users......That one I find hard to believe and the numbers I got had xfire at only 150-180k users......

      We've went through the xfire thing many times.....Yeah I know some of you use it but I dont think the majority of us do......If you want to use it for curiousity then great but dont generalize that the sky is falling because of some stats shown on it.........As for WAR it doesnt matter if they have 1 million players or 100k players....If youre having fun and enjoying the game then that is what matters.

Originally posted by LondonMagus

Originally posted by templarga

Bias has nothing to do with intent. Bias can be anything. What if X-fire is used by people who play single player games more than MMO's? What if X-fire is used more by people in the US than in other areas of the world? What if the average X-fire players plays 6 hours a day but the average MMO'er plays 3 hours a day? What if they average age of a gamers according to X-fire is 18 but the real average age is 28?

Any of this can skew your data and make it worthless if you cannot: a - prove the bias; b - take it into account with your poll and prediction.

I could keep going but I think you get the idea. Bias can be anything at all and many times there is no intent behind, it is simply bias. Until all possible bias is accounted for and taken into account by the pollster, any data gathered is unreliable.

LOL, you sound like you hate bad statistics as much as I do. 

I do! I hate bad statistics but I hate drawing bad conclusions from bad statistics even more. I hold a Master's in Public Administration and statistics were integrated into many of our classes not to mention the actual statistics courses. Heck, the first half of one course was all about how to sample and gather data and biases....and this was before we did our first calculation.

Next Fall, I start on the PhD and will have multiple courses in statistics including courses specifically on surveying and gathering data and some on polling alone. I also teach in my American Government courses about statistics as they relate to opinion polls do I have a decent background in stats....so anytime I read bad stats or bad stat conclusions, I cringe. Glad there is someone out there that feels the same way I do.

Originally posted by templarga

Originally posted by LondonMagus

Originally posted by templarga

Bias has nothing to do with intent. Bias can be anything. What if X-fire is used by people who play single player games more than MMO's? What if X-fire is used more by people in the US than in other areas of the world? What if the average X-fire players plays 6 hours a day but the average MMO'er plays 3 hours a day? What if they average age of a gamers according to X-fire is 18 but the real average age is 28?

Any of this can skew your data and make it worthless if you cannot: a - prove the bias; b - take it into account with your poll and prediction.

I could keep going but I think you get the idea. Bias can be anything at all and many times there is no intent behind, it is simply bias. Until all possible bias is accounted for and taken into account by the pollster, any data gathered is unreliable.

LOL, you sound like you hate bad statistics as much as I do. 

I do! I hate bad statistics but I hate drawing bad conclusions from bad statistics even more. I hold a Master's in Public Administration and statistics were integrated into many of our classes not to mention the actual statistics courses. Heck, the first half of one course was all about how to sample and gather data and biases....and this was before we did our first calculation.

Next Fall, I start on the PhD and will have multiple courses in statistics including courses specifically on surveying and gathering data and some on polling alone. I also teach in my American Government courses about statistics as they relate to opinion polls do I have a decent background in stats....so anytime I read bad stats or bad stat conclusions, I cringe. Glad there is someone out there that feels the same way I do.

I smile when people who should be intelligent mix-up everything just because they want to defend a game.

:))

Purpose: of XFIRE: showing what people play on their PC when XFIRE is installed on their PC's.

Results for WAR: (like AoC): After 6 weeks a sharp decline of player activity for WAR on the users of X-FIRE. Up to 55% on average.

http://www.xfire.com/games/who/Warhammer_Online_Age_of_Reckoning/

Meaning: War is being played LESS with X-Fire users.

Sample: 1 to 2% of the total players that play War have XFire installed.

Poll value?: Since the profile of a gamer of WAR is undefined, unless you can say the person must have a PC and access to the internet ----> the sample is EXTREMELY in relation to what the goal is: looking at how many players play WAR in general.

You guys mix up social and marketing studies with the very limited goal of Xfire: and that is: WHAT PC game is being played once  you install XFIRE on a client.

And just like AoC we see a VERY downward trend in number of players AND playing time for WAR after the initial free period.

We see it on the servers, we see it on X-Fire.

Nothing more nothing less. And I keep saying: history has proven that these Xfire give a day to day analysis of a game being played less and less (AoC).

1 to 2% sample AND have the right (very limited) sample (PC gameplayers with an internet connection) does the rest.

Sleep well.

Originally posted by strategy 

I smile when people who should be intelligent mix-up everything just because they want to defend a game.

:))

Purpose: of XFIRE: showing what people play on their PC when XFIRE is installed on their PC's.

Results for WAR: (like AoC): After 6 weeks a sharp decline of player activity for WAR on the users of X-FIRE. Up to 55% on average.

http://www.xfire.com/games/who/Warhammer_Online_Age_of_Reckoning/

Meaning: War is being played LESS with X-Fire users.

Sample: 1 to 2% of the total players that play War have XFire installed.

Poll value?: Since the profile of a gamer of WAR is undefined, unless you can say the person must have a PC and access to the internet ----> the sample is EXTREMELY in relation to what the goal is: looking at how many players play WAR in general.

You guys mix up social and marketing studies with the very limited goal of Xfire: and that is: WHAT PC game is being played once  you install XFIRE on a client.

And just like AoC we see a VERY downward trend in number of players AND playing time for WAR after the initial free period.

We see it on the servers, we see it on X-Fire.

Nothing more nothing less. And I keep saying: history has proven that these Xfire give a day to day analysis of a game being played less and less (AoC).

1 to 2% sample AND have the right (very limited) sample (PC gameplayers with an internet connection) does the rest.

Sleep well. 

Nice try at avoiding answering all the points by adopting a smug attitude, that doesn't just make me smile it makes me laugh out loud!

A biased sample may have coincidental behaviour with the main population but it cannot be considered reliably predictive of anything other than itself & no-one really cares about Xfire users anyway. Either you understand that perfectly well & are being intentionally awkward or you have no understanding of statistics at all & should read up on the theory before embarrasing yourself in public.

Its always funny when you encounter people that are so thick skinned they are unvulnerable to logic!

Maybe a Good Nights sleep will help you understand things better.

Originally posted by LondonMagus

Originally posted by strategy 

I smile when people who should be intelligent mix-up everything just because they want to defend a game.

:))

Purpose: of XFIRE: showing what people play on their PC when XFIRE is installed on their PC's.

Results for WAR: (like AoC): After 6 weeks a sharp decline of player activity for WAR on the users of X-FIRE. Up to 55% on average.

http://www.xfire.com/games/who/Warhammer_Online_Age_of_Reckoning/

Meaning: War is being played LESS with X-Fire users.

Sample: 1 to 2% of the total players that play War have XFire installed.

Poll value?: Since the profile of a gamer of WAR is undefined, unless you can say the person must have a PC and access to the internet ----> the sample is EXTREMELY in relation to what the goal is: looking at how many players play WAR in general.

You guys mix up social and marketing studies with the very limited goal of Xfire: and that is: WHAT PC game is being played once  you install XFIRE on a client.

And just like AoC we see a VERY downward trend in number of players AND playing time for WAR after the initial free period.

We see it on the servers, we see it on X-Fire.

Nothing more nothing less. And I keep saying: history has proven that these Xfire give a day to day analysis of a game being played less and less (AoC).

1 to 2% sample AND have the right (very limited) sample (PC gameplayers with an internet connection) does the rest.

Sleep well. 

Nice try at avoiding answering all the points by adopting a smug attitude, that doesn't just make me smile it makes me laugh out loud!

A biased sample may have coincidental behaviour with the main population but it cannot be considered reliably predictive of anything other than itself & no-one really cares about Xfire users anyway. Either you understand that perfectly well & are being intentionally awkward or you have no understanding of statistics at all & should read up on the theory before embarrasing yourself in public.

Its always funny when you encounter people that are so thick skinned they are unvulnerable to logic!

Maybe a Good Nights sleep will help you understand things better.

See the evidence of the correct applied sample now ? :))
 

It's predictive of anything other than itself: meaning Xfire War players. Pity you don't care about Xfire players but that's just your cognitive dissonance playing, not the reality of data assembled of Xfire users playing PC games.

Originally posted by strategy

Originally posted by LondonMagus

Originally posted by strategy 

I smile when people who should be intelligent mix-up everything just because they want to defend a game.

:))

Purpose: of XFIRE: showing what people play on their PC when XFIRE is installed on their PC's.

Results for WAR: (like AoC): After 6 weeks a sharp decline of player activity for WAR on the users of X-FIRE. Up to 55% on average.

http://www.xfire.com/games/who/Warhammer_Online_Age_of_Reckoning/

Meaning: War is being played LESS with X-Fire users.

Sample: 1 to 2% of the total players that play War have XFire installed.

Poll value?: Since the profile of a gamer of WAR is undefined, unless you can say the person must have a PC and access to the internet ----> the sample is EXTREMELY in relation to what the goal is: looking at how many players play WAR in general.

You guys mix up social and marketing studies with the very limited goal of Xfire: and that is: WHAT PC game is being played once  you install XFIRE on a client.

And just like AoC we see a VERY downward trend in number of players AND playing time for WAR after the initial free period.

We see it on the servers, we see it on X-Fire.

Nothing more nothing less. And I keep saying: history has proven that these Xfire give a day to day analysis of a game being played less and less (AoC).

1 to 2% sample AND have the right (very limited) sample (PC gameplayers with an internet connection) does the rest.

Sleep well. 

Nice try at avoiding answering all the points by adopting a smug attitude, that doesn't just make me smile it makes me laugh out loud!

A biased sample may have coincidental behaviour with the main population but it cannot be considered reliably predictive of anything other than itself & no-one really cares about Xfire users anyway. Either you understand that perfectly well & are being intentionally awkward or you have no understanding of statistics at all & should read up on the theory before embarrasing yourself in public.

Its always funny when you encounter people that are so thick skinned they are unvulnerable to logic!

Maybe a Good Nights sleep will help you understand things better.

See the evidence of the correct applied sample now ? :))
 

It's predictive of anything other than itself: meaning Xfire War players. Pity you don't care about Xfire players but that's just your cognitive dissonance playing, not the reality of data assembled of Xfire users playing PC games.

I don't care about Xfire users because they are an irrelevant subset of the game playing community, much as I don't read too much into the opinions of people that still think flared trousers are still fashionable.

Sorry to have broken your sleep, do you have an alarm clock keyed into forum posts or something. 

Originally posted by templarga

Originally posted by LondonMagus

Originally posted by templarga

Bias has nothing to do with intent. Bias can be anything. What if X-fire is used by people who play single player games more than MMO's? What if X-fire is used more by people in the US than in other areas of the world? What if the average X-fire players plays 6 hours a day but the average MMO'er plays 3 hours a day? What if they average age of a gamers according to X-fire is 18 but the real average age is 28?

Any of this can skew your data and make it worthless if you cannot: a - prove the bias; b - take it into account with your poll and prediction.

I could keep going but I think you get the idea. Bias can be anything at all and many times there is no intent behind, it is simply bias. Until all possible bias is accounted for and taken into account by the pollster, any data gathered is unreliable.

LOL, you sound like you hate bad statistics as much as I do. 

I do! I hate bad statistics but I hate drawing bad conclusions from bad statistics even more. I hold a Master's in Public Administration and statistics were integrated into many of our classes not to mention the actual statistics courses. Heck, the first half of one course was all about how to sample and gather data and biases....and this was before we did our first calculation.

Next Fall, I start on the PhD and will have multiple courses in statistics including courses specifically on surveying and gathering data and some on polling alone. I also teach in my American Government courses about statistics as they relate to opinion polls do I have a decent background in stats....so anytime I read bad stats or bad stat conclusions, I cringe. Glad there is someone out there that feels the same way I do.

Ok I've seen this posted by you a zillion times.  And what you say taking from a mathmatician point of view is absolutely correct.  However taken from a practical point of view like an engineer of social scientist is not.  What statistical polling type data is always going to be bad to some degree in almost all cases.  And if you don't know the badness of the statistic it is hard to determine the inherent error rate in the results.  Or in other words the error becomes unbounded.  But again we are speaking in absolutes and in the extreme.  The reality is not absolute but rather analog.  So we know that xfire sampling is not entirely random or unbiased.  This means that the data has a range of potential error that we can't quantify from looking at the source of the statistic.  This does NOT mean that the data is meaningless.  It merely means that you can't quantify the error.

So instead you need to quantify the error using other means through experimentation.  Kinda like if put 1000 people in a room and I go into the room and only talk to people in one corner of the room I'm getting a biased data sample of people that only liked to hang out in the sw corner maybe because the lighting was better there or some other self selecting rational.  So maybe 30% of my sample prefer snickers bars.  Since it was a biased sample I can't say that 30% of the entire room like snickers and while I could say that 30% of the people in the corner liked them and the distribution of the people in the corner relative to the rest was a certain percent and from that calculate and error level.  I could also measure the entire room and calabrate that against my results and do that several differnt times and that also would allow me to quantify the significance of my bias and the errors.

Meaning that if the data from xfire is in reality valid in test cases then it is probable that it is valid in other cases.  And while again this doesn't hold up in terms of absolutes very little ever does. What you can determine is that it is valid in test cases and therefore is not exhibiting errors.  This doesn't mean you can toss out your logic there are obvious things that can poison your data.  Like maybe xfire sends out a promo to all gamers telling them that for the next month playing on wow will be free.  This is gonna create a surge of xfire users playing wow that won't be reflective of the community.

But in a general sense xfire does clearly map to real world trends.  When a game launchs xfire shows a large serge up.  No surprise.  When a game dies xfire shows the game dying.  If a game is really popular xfire shows it doing better then other games.  Can I compare AoC to LInage II using xfire data and say that one game is doing better then the other.  No there is a margin of error and while I don't know what it is I do expect it to be something.  If things are plus or minus 10 or 20% I consider that too close to tell.  However I can say that there are more players playing WoW then Warhammer.  I can say that there are more people Playing War then LoTr.  Or there are more people playing LoTr then AoC.  I feel rather confident making these statements and I'd put money on the table any day saying I'm right.

So in summary for those that haven't fallen asleep.  Is xfire data absolute NO WAY.  Is xfire meaningless also NO WAY.  The truth is in the middle and xfire is the best data that we have access and it is meaningful in doing some preditions as long as we understand that there is a significant margin of error.