Public Statistics

Salt · January 2013

<div class="IPBDescription">i can not stress this enough</div>It feels like a lot of people still don't quite understand the natural selection 2's game statistics.
The only thing we publically have access to that shows statistics is ns2stats.org.
The statistics the actual development team possess and balance by is not what we have access to.

So i can not stress enough that ns2stats.org ONLY takes in statistics from servers that has this third party 'mod' installed.
It are not statistics of every single game ever played in ns2. (And even then the server can disable sending statistics to the ns2stats.org service.)
So for the love of whatever you hold dear stop using this as your 'source' for saying the game is unbalanced and broken and dead or whatever weird threads have been popping up lately.
So get lost with your 60% alien win rate please. because this really is a moderately baseless assumption.

people have pointed this out in answers to topics, but i feel as though this needs to be highlighted AS a topic.

Edit:
I forgot to mention as someone else did,
Servers can have mods too that can unbalance the whole of statistics pulled even more.
On top of servers only playing one map, or clan servers that mostly have one particular team stacked because it contains a majority of good/famous players.

Comming to think of this, there are actually a lot of things that can skewer the statistics around a lot. i'm genuinly curious as to how UWE themselves measure these statistics and if they filter for combat mod servers and/or servers only running one map etc...

Shino · January 2013

Unless you have data explaining a correlation between having ns2stats and player skill/win rates (For example- a server with ns2stats will experience a higher win rate for marines in comparison to a server without ns2stats), using ns2stats is completely legitimate.

There is most likely no correlation between player skill and whether a server has ns2stats or not. Because of this, it is a good sample size which is probably indicative of the total population.

YMICrazy · January 2013

Yup the devs have stated aliens have an edge. I think the last time they officially gave an article on the win rate was in November

<a href="http://www.unknownworlds.com/ns2/news/2012/11/build_230_is_now_live_on_steam" target="_blank">http://www.unknownworlds.com/ns2/news/2012...w_live_on_steam</a>

They say it was 60% aliens victories to 40% marine victories.

If you go to ns2 stats and type in the same date it is pretty close. Of course now it has lowered to less than 60% win rate meaning their changes did even things a bit. It is probably the best thing we will get when considering win rates. However it is probably not the best way since I think a few combat servers may alter those results but you can filter those out even though it might take time.

Salt · January 2013

Basing an average on a percentage of games played is faulty, It does not represent the whole, Because game win/losses are not contingent, This has to do with higher / lower population servers or servers only playing one single map, where a majority / minority of have NS2Stats.

An average needs to be taken upon the absolute total amount of server played games, the statistics are flawed if it is only partial.

for example, Lets say ns2stats is installed on 50% of the total amount of servers.
That means the game split is 50/50, The average for the partition of servers with ns2stats can be completely different than the partition of servers that don't have ns2stats.

It's the same as conventional statistic pulling, Lets say you ask what people find about the crime, And you say you ask 3000 people, but only 1000 people from 3 towns.
You will get wildly different results if you ask 3000 people from a single large city, but 'statistically' your values are 'correct'.

It just isn't accurate, At least, not accurate enough to be true.

I have played 150 some such hours of NS2 since launch, and even more in beta, but according to ns2stats, i've only played 19 games?
Go look up your own name and see how many matches you've allegedly played.

i'll retort by saying it's not completely baseless, but simply invalid. it does not represent the whole.

Shino · January 2013

It only doesn't represent the whole if you can prove that servers with ns2stats display a different correlation between servers without ns2stats.

Are rookies more attracted to servers with ns2stats installed? Are they repulsed? Chances are, there is a completely random sample of rookies in both servers, so the win rate isn't any different.

You state the above as if it were fact very casually as if it were a fact.

With your three towns example, let's say you want to ask 3000 people. If 1000 people in each city are representative of the same population then the results will all be the same for each city.

In our scenario, we are polling everybody that plays NS2. Players of NS2 randomally (I am asserting that it is completely random) choose a random server to play on. Since the server selection is not changed in any way shape or form due to ns2stats being installed, I assert that the win rate on ns2stats server is the same as vanilla server, since there is no discrimination between picking servers between the population of NS2 (You do not have people saying "This server has ns2 stats so I will play on it / I will not play on it). Unless you can prove that people selectively choose NS2stats server, as well as this having a confounding effect on the win/loss%, it is a good sample.

Salt · January 2013

the example i used of the crime rates polling was only used to say how closely this kind of statistic polling correlate to conventional statistic polling.

Why?
Because no one knows just how many good / bad players play on servers with ns2stats.

I'm not saying ns2stats's statistics are false, i'm saying they're inaccurate to the point where you can't base facts upon them.

Let me try explain my point more clearly;
<span style="color:#FF8C00">Statistically correct player-wise:</span>
Lets say 5000 players are polled, of which a quarter are pro players (1250) and half are average (2500) and another quarter are rookies/generally new players (1250)

Now, obviously these player numbers are by not 'ordered' like this, and are genuinely random.

So on 'average' that would mean the following is true:

16 players: 4 are pro, 4 are rookies 8 are average.

This equates to the total of 5000 and is statistically correct.

Now lets say that half of these players are polled by ns2stats.
Does that mean that per 2500 players, 625 player are pro, 625 players are rookies and 1250 players are average?
False, because it is 'genuinely' random chances are, the first batch (taken in by ns2stats) has 1000 pro players, and the second batch, only has 250. The same 'could' go for the rookies and average players.

Thats what i mean with how these statisticas can be skewered.

<span style="color:#FF8C00">The anomalies server-wise:</span>
Modded servers.
Clan servers.
Servers that favor certain maps.
Servers with low / high population counts (does matter on balance issues.)

Now combine the anomalies with the statistically 'correct' player counting, you'll get one chaotic hotchpotch of statistics that can be completely different when polled separately.

On top of that comes that i keep swinging 50/50 as a number around for simplicity , when really only like 20/25% of all servers run ns2stats (servers lighting up yellow in your game list 'can' contain ns2stats others do not)

SquishpokePOOPFACE · January 2013

I'm having flashbacks to statistics class...

Shino · January 2013

At no point in your OP do you reference the stats being wrong for 8v8/6v6, rather you flat out say the stats are wrong in general. With the population differences (12v12 being used in 8v8 and 6v6 stats) I agree, however in your example

If at first you get 1000 pro / 250 average / 250 rookies, the next time you poll you could have 250 pro / 1000 average / 1000 rookies, and so on and so on. This is why sample size is important. Sample size takes care of and normalizes this out, causing it to be representative of its population over a long enough period of time. Now you're not saying that ns2stats is inaccurate, rather you're saying the sample size is too small which is not the case.

There is no data to show that the level of skill in modded servers vs regular servers is any different. Combat and Regular NS2 could confound the data, but iirc someone addressed this saying that combat is statistically insignificant in comparison to normal ns2 playing in terms of % played (It is not big enough of a confound to alter the data in any significant way)

Clan servers are most often flagged as 'Competitive' which would not bunch itself up in 'public' stats (Check the filter, it's actually separated into two different categories)

Servers that favor different maps doesn't matter, as we're looking at total win/loss rate, not win/loss rate on certain maps. If we say "Marines have a 40% win rate and aliens have a 60% win rate", there are probably maps which are 35/65, and there are probably maps that are 45/55. It doesn't matter. Overall, marines 40, aliens 60. We're not going into that level of detail.

And as stated before, you didn't discriminate between 8v8/6v6 in the title. We're assuming out of all games, which includes 12v12.

I'm sorry your bull###### isn't working :(

Salt · January 2013

"I'm sorry your bull###### isn't working :(" - That really is unnecessary, if you cannot discuss an issue in a civilized way please refrain from posting at all.

And you're still missing my point of saying that polling 50% of a whole does not mean that each and every game played that is polled has been played as an average.
Take my explanation of the pro / rookie / average players as a concept and apply to to every single factor that can determine the outcome of a game, both 6v6 / 8v8 / 12v12, maps in general, game modes, player counts.

Once again i'll say the win/loss rates polled by ns2stats do not represent the whole, at least, not accurately.
Because it is only 20 or 30% of the entire amount of games played.

In order to normalize statistics you have to take it from the WHOLE not just 20 or 30 percent, because in that 20 / 30 percent, per chance those games polled could be the 'good' or 'worse' games played, where as the 'whole' equates that the 'good' and 'worse' games played can amount up to completely different results, BECAUSE of chance.

Not taking in the whole of games played can not represent the whole. it really is as simple as that.

PaLaGi · January 2013

It would be neat if you could prove in your argument that servers running and reporting NS2stats are statistically different than servers that don't before you start using your big city/small town analogy and other false assumptions.

<b>More than likely, ns2stats gives a random sampling of the greater whole</b> and the anomalies (servers running balance mods, rfk) aren't contributing significantly to skew either stat.

Also your idea that clan servers populated with good/famous people constantly stacking one team and skewing a large pool of data (are you serious with that one?).

The argument you should be making is that stats are far more irrelevant in pub games than people realize right now due to the nature of how new that game is and that the majority of players are still at the bottom of a learning curve. You could make the hypothesis that the alien team requires less teamwork/coordination to win perhaps due to the atonomy of the alien commander and the ability to complete tasks more easily without the teams' participation and understanding of the strategic side of the game. You could also make the hypothesis that despite more new players being familar with playing a human with a gun, proficiency at shooting plays less of a role in these victories (likely due to the prominent RTS element).

hakenspit · January 2013

Win loss rates are not the be all and end all, UWE have said so themselves.

They are not balancing a game on win loss ratios. They want the game to be fun, enjoyable, have a decent skill curve etc etc.

People take NS2 stats and think that they mean a game is not balanced, thankfully UWE dont share this limited perspective on balance.

They have bigger ideals for what a balanced game means (it involves words like fun for both team to play) and are balancing with these design goals in mind.

Well UWE have said that the win loss from their data is similar and has been for a while..BUT the changes they are about to make focus on other areas of their design brief.

I played aliens throughout most of the beta, got stomped a lot ...and I mean a lot often being down atleast 1 player but often 2.

Heck for a while there the highest alien lifeform was the lerk, so you can imagine how much chop aliens where after about 5-8 min into a game.

So people saying its not fun to lose and people wont play is garbage, I did as did a lot of other people.

|strofix| · January 2013

Say you have 100 people. Now you randomly choose 20 of them and analyse some part of their every day life in order to garner information about the population as a whole. This is completely legitimate and would yield data that is as accurate as you could expect a 20% sample size to be. The data gained should accurately represent the entire population.

Now say you did the same test, except instead of picking 20 people at random, you chose to analyse the 20 wealthiest people. Because your sample set is influenced by the selection process, you cannot assume that the results accurately represent the entire population. Since random selection is not influenced by anything, you have a good chance of getting relatively representative data.

What this means is, if you can't show some sort of correlation between a server having the ns2stats mod and the type of players that play on said server, then you have no reason to think that the stats acquired are skewed in any way. It is essentially a random subset of the entire population, which should be relatively representative.

Your idea of having 100% coverage of all games is not feasible. Lets say UWE had a way to track every single game that ever occured on any online server anywhere in the world. What about me and a friend having a 1v1 on an offline server, using a cracked version of NS2? Now they have missed a game, does that throw out all the stats they have acquired just because its 99.99999999999% coverage?

Of course not.

Salt · January 2013

'False assumptions'? i said not taking in the whole of results can result to the same results as conventional statistic pulling.
Merely a comparison to faulty statistic pulling.

Don't pry compartmental examples i give out of context, they contribute to the whole of my point.
The point being that ns2stats statistics cannot be taken as completely 'true' in people's posts of saying how messed up everything is.
UWE doesn't, and i know this, but i'm just generally getting tired of topics starting with saying that the win/loss rates are ridiculously different and generally blaming it on out of context mechanics.

i don't need 'evidence' because there is none, the thing i'm trying to do is take away the evidence ns2stats apparently seems to hold. When it does not represent 100% of all games played.

I'll try give another example, despite it might be taken out of context.
This example goes without saying, they are not real results or real percentages, but just to try illustrate and emphasize on the main point i'm trying to make given that all games are equal and no anomalies like you guys say don't matter.

Orange: games in favor of alien wins
Blue: games in favor of marine wins

<span style="color:#0000FF">Server 1: 25 games played, Marine/Alien winrate: 15/10.</span>
<span style="color:#0000FF">Server 2: 30 games played, Marine/Alien winrate: 18/12.</span>
<span style="color:#FF8C00">Server 3: 33 games played, Marine/Alien winrate: 13/20.</span>
<span style="color:#FF8C00">Server 4: 15 games played, Marine/Alien winrate: 5/10.</span>
<span style="color:#FF8C00">Server 5: 45 games played, Marine/Alien winrate: 15/30.</span>
<span style="color:#0000FF">Server 6: 23 games played, Marine/Alien winrate: 13/10.</span>
<span style="color:#FF8C00">Server 7: 19 games played, Marine/Alien winrate: 9/10.</span>
<span style="color:#0000FF">Server 8: 22 games played, Marine/Alien winrate: 20/2.</span>
<span style="color:#0000FF">Server 9: 20 games played, Marine/Alien winrate: 12/8.</span>
<span style="color:#FF8C00">Server 10: 14 games played, Marine/Alien winrate: 6/8.</span>
total marine/alien winrate 126/120 = 51/49%
NS2stats polls servers 3, 5, 9, 10 with a rate of 46/66 = 41/59% - like it is now on 40% of all (in this example 10 servers)
the remaining un-polled servers 1, 2, 4, 6, 7, 8 with a rate of 80/54 = 60/40%.

While the total really can be 50/50% like in this example of all (10) servers.

Shino · January 2013

The error you show is an error of sample size. You are claiming that given certain configurations, you can show a 50-50 split of win:loss, when it is not true.

Ns2stats as you stated above:

<div class='quotetop'>QUOTE (Salt @ Jan 9 2013, 03:19 AM) <a href="index.php?act=findpost&pid=2058296"><{POST_SNAPBACK}></a></div><div class='quotemain'>like 20/25% of all servers run ns2stats</div>

This is good enough to remove any random variance. Hell, this is more than good enough.

I had a ns2stats link me to round/33000s. We could have 1/30th of this sample size and it would still be more than accurate.

Yuuki · January 2013

Shino is right, the additional sampling that ns2stats introduce by not recording all the servers only increase the variance, but does not introduce a systematic bias, unless your assume that the player that play on ns2stats servers are somehow different from the ones that play elsewhere, which is dubious. The variance goes down like one over the sample size, so that solve this. We can also estimate the error on the winrate estimate, but it's pretty small for large numbers of game.

mushookees · January 2013

<div class='quotetop'>QUOTE (Salt @ Jan 9 2013, 01:48 PM) <a href="index.php?act=findpost&pid=2058330"><{POST_SNAPBACK}></a></div><div class='quotemain'>stuff</div>
and if the stats that UWE have collected confirm what NS2stats is showing us ... then you have no argument right ?

Isnt this thread kind of pointless ? you have no evidence what so ever and you cant really prove anything really. Rather than start this thread, you could of tried to email charlie/other devs/strayan or ask some playtesters to pester the devs and try get the statistical data that UWE have for this latest build ... and if that data was different from NS2stats then you could prove your point ...

but currently your just wasting everyones time :/

statikg · January 2013

Salt I'm afraid your argument is based on a basic lack of understanding of statistics and how random sampling works.

bongofish · January 2013

<div class='quotetop'>QUOTE (statikg @ Jan 9 2013, 08:40 AM) <a href="index.php?act=findpost&pid=2058348"><{POST_SNAPBACK}></a></div><div class='quotemain'>Salt I'm afraid your argument is based on a basic lack of understanding of statistics and how random sampling works.</div>

Exactly. It's obvious he's not dumb but it also obvious that he knows nothing about statistics and polling.

Ender_74 · January 2013

Whether we know about statistics or not, I tend to agree that the ns2stats.org global wins per race cannot be taken blindly as a real indicator of the current balance.

As far as I know it takes into account all game mods as explained above. It seems to me that the current balance of the combat mod is in the marines favor (I base my point of view on the wins per map). Although there are a lot fewer games of co than of ns2, it still must influence the global win ratio.

To get a more precise ratio, I would only count the games played on official maps, which would give a 58.2% alien win rate. (3644 wins out of 6258 when I'm typing this). All the mods that are played on the official maps may still influence this number though.

gnoarch · January 2013

still the number of games played with mods really changing the game is quite low. also it'S not like ALL mods favor marines. This leads to mods balancing out each other.
If really 20-25% of all games are monitored in NSstats it's a samplesize big enough.

I mean of course you cant say that it is <b>exactly</b> 58.2% and not 58.4%. Still it is save to say it is not 45%.

BentRing · January 2013

My biggest issue with stats is that it promotes personal player play rather than team play.

And if you think it doesn't, start a thread or do a search about removing k/d or score from the scoreboard.

Troubleshooter · January 2013

@Salt : Under what criteria would you consider a sampling of games to accurately reflect the balance of the game? Lay out your criteria completely please, don't bother with hypothetical examples as they will only confuse the issue.

The only way to resolve the point of contention here is to directly state your position on what constitutes a valid sample and not all the various issues you have with a particular data set are because until you establish a point from where you are satisfied the sample is valid everything else is nit-picking and spin no matter how sincere you are about your objections.

Kalabalana · January 2013

The reporting seems to be broken, with at least 50 games in NS2stats servers, only data from a few games is recorded.
Also the stats seem to ignore the majority of tangible game data, and there doesn't seem to be any adequate interpretation of the data actually recorded.

Savant · January 2013

I don't know why anyone would make a thread like this. Is it to try and suggest that win rates are even? We know they're not and the developers have REPEATEDLY said the aliens have the edge right now. (and they have access to stats from 100% of games) Trying to argue otherwise seems pretty pointless, and trying to slag NSstats for providing a service to the community - because you disagree with the results - is just plain childish.

Aliens win more often. I know it and the developers know it; you seem to be the only one in denial here.

ogz · January 2013

So what is the 95% comfidence level of ns2stats data

why not just report that then everyone can be happy

eg winrate is 60% aliens +/- 5%
(made up figures)

GORGEous · January 2013

<div class='quotetop'>QUOTE (Kalabalana @ Jan 9 2013, 11:41 AM) <a href="index.php?act=findpost&pid=2058434"><{POST_SNAPBACK}></a></div><div class='quotemain'>The reporting seems to be broken, with at least 50 games in NS2stats servers, only data from a few games is recorded.
Also the stats seem to ignore the majority of tangible game data, and there doesn't seem to be any adequate interpretation of the data actually recorded.</div>

Have you checked your filters? I see you as having played 33 games in the last 2 weeks via ns2stats.

(there is a bug where ns2 stats fails to properly parse logs, but I find that 80-95% show up correctly)

Zaggy · January 2013

I've deleted a number of offtopic posts, please note: if you don't have anything to add, do not post, thanks.

hakenspit · January 2013

<div class='quotetop'>QUOTE (Savant @ Jan 10 2013, 03:00 AM) <a href="index.php?act=findpost&pid=2058437"><{POST_SNAPBACK}></a></div><div class='quotemain'>I don't know why anyone would make a thread like this. Is it to try and suggest that win rates are even? We know they're not and the developers have REPEATEDLY said the aliens have the edge right now. (and they have access to stats from 100% of games) Trying to argue otherwise seems pretty pointless, and trying to slag NSstats for providing a service to the community - because you disagree with the results - is just plain childish.

Aliens win more often. I know it and the developers know it; you seem to be the only one in denial here.</div>
I am assuming to hopefully try to shed some light for the masses who seem blindly obsessed with win loss rates.

UWE, unlike most people here, are not saying the win:loss rates indicate any issues with the fundamental balance of the intended design goals.

In fact they are setting about tweaking aliens (some say buffing) to make them more fun to play.

Unlike most people here UWE see there is more to balancing a game than win loss ratios, longevity, skill curves and FUN are equally valid measures.

Bullsh*t, lies and statistics.

Think about why these 3 are grouped together in this saying...hint this is not a sesame street "one of these things is not like the other" games.

Heck for years the increasing size of drug busts has been seen as evidence of good policing....despite its total inability to actually stifle or even reduce the supply, cost or quality.

Salt · January 2013

Sorry for the somewhat late reply, i went to work and then sleep after that.

Let me first be clear in saying i'm not trying to force a point here. i posted this in General discussion to prompt it's discussion, and it's been lively in a positive sense.
As far as 'proving a point' goes all i'm saying is that the NS2stats statistics can't be hurled around like facts in all the threads about balance i've read, sure UWE themselves have said the winrate was in favor of aliens, but this was reported in november last year, and have taken appropriate action.
I was just curious as to what motivation people have to keep up these statistics when it could've balanced out already.

Sorry if people thought i was attacking the system or any some such nonsense.
Just got a little tired of the blind ignorance i've seen, if i may be so blunt.

Second thing, to answer Troubleshooter's inquiry;
To be honest the best possible statistic polling ofcourse would be 100% of all games played. Because then it fully and accurately shows the winrate.
As far as balance goes i think there's a thing or two to say, because ultimate balance feedback should only really apply in one particular case where a lot of things are set in stone, like no modded servers, a set number of players on every team, no players leaving mid-game. equally skilled players on each team.
But unfortunately we can never balance like this, since the odds of this happening are slim to none.

In order to get an average, and what i find to be accurate with the smallest error margin, is at least over 50% of the total.
Because after this half mark the variation on win/loss rates gets gradually less and less variation and more and more consistent.
People seem to be content with 25% of all servers (which run ns2stats), the controversy for me lies with sample size representing a whole.
Say this 25% is such a large number that polling results forming an average 'could' potentially represent the whole, there still is 75% of data (the larger part) that could completely turn these statistics around if fully taken into account.

Not saying it DOES /because/ there is no evidence., but just the sheer chance that it can form a large turn around makes the end result janky to use as fact, as so many people seem to believe it does.

I would love people to prove me wrong, that the random factor of this larger portion of results not being taken into account doesn't matter for the average.
Because when i lay this out on paper and go over the numbers myself it just doesn't seem accurate enough.

gnoarch · January 2013

<div class='quotetop'>QUOTE (Salt @ Jan 10 2013, 11:45 AM) <a href="index.php?act=findpost&pid=2058790"><{POST_SNAPBACK}></a></div><div class='quotemain'>In order to get an average, and what i find to be accurate with the smallest error margin, is at least over 50% of the total.
Because after this half mark the variation on win/loss rates gets gradually less and less variation and more and more consistent.
People seem to be content with 25% of all servers (which run ns2stats), the controversy for me lies with sample size representing a whole.
Say this 25% is such a large number that polling results forming an average 'could' potentially represent the whole, there still is 75% of data (the larger part) that could completely turn these statistics around if fully taken into account.

Not saying it DOES /because/ there is no evidence., but just the sheer chance that it can form a large turn around makes the end result janky to use as fact, as so many people seem to believe it does.

I would love people to prove me wrong, that the random factor of this larger portion of results not being taken into account doesn't matter for the average.
Because when i lay this out on paper and go over the numbers myself it just doesn't seem accurate enough.</div>

So take a country like Germany. We have more than 80 mio people living here.
Whenever there are elections there are opinion polls made to predict the outcome of these elections. These polls in most cases are correct within a window of (in extreme cases) 5% variance.
This is despite some weeks before the actual election no more of these opinion polls are allowed so they don't influence the outcome by themselfes. So major events and influences on the elections can not even be considered by these opinion polls. If there are no such "last minute" events the opinion polls are mostly spot on +-2%.

So you could say these opinion polls are quite accurate despite them being fairly outdated by the time of the acutual election.

So now guess how big the sample size of these polls is. 50% ? 40%? 10%?

The sample size are a ridulous between 1000 and 1004 people. So the sample size is about 0,0016% (about 62 mio people have the right to vote)

From this example it should be very clear that a sample size of 25% is ridiculously high.

Of course this example is not 100% correct as these 1000 people are not purely choosen by random but also by statistical criterias but with a sample size roughly 15000 times larger these additional criterias have not more measureable impact on the accuracy of the resulting statistics.

What you are saying is like: "sure, I threw my dice 600 times and every number came up about 100 times. BUT maybe if I throw it another 3000 the ratio of orrurence of these numbers suddently changes in a way so 6 occures considerably more often than any other number."
In theory this could occur if you have a sudden 6-streak and proofe me wrong.
But in theory it could also happen that my girlfriend this evening brings home my hot co-worker as well as BeyoncÃ© and my favourite porn star to join in on a fivesome in our new villa she bought after winning the lottery.

Theoretically both is possible. But think it's no mistake to not order 5 boxes of 500$ Champagne for tonight ...

In short: Maybe you should read some stuff on statistics because I dont think you really get something wrong about the whole concept of sample size etc.

Brony · January 2013

<div class='quotetop'>QUOTE (Salt @ Jan 10 2013, 11:45 AM) <a href="index.php?act=findpost&pid=2058790"><{POST_SNAPBACK}></a></div><div class='quotemain'>Say this 25% is such a large number that polling results forming an average 'could' potentially represent the whole, there still is 75% of data (the larger part) that could completely turn these statistics around if fully taken into account.</div>

Salt, try this thought experiment:

Flip a coin 1000 times. Note how many "heads" you get the first 250 flips. Then note how many "heads" you get the last 750 flips. Do you think its reasonable to assume that your percentage of "heads" will be any different in the first sample than the last? And what if you repeat this experiment 100 times, wouldn't you agree that the percentage of "heads" will be very close to 50% in all the samples?

The reason most people argue so strongly against you is not because they are defending NS2Stats, its because what you are doing is questioning a very fundamental concept in statistics and the scientific method. Sampling has been used for centuries and a lot of research is based on it, which has been performed by folks a lot smarter than you and me.

Public Statistics

Comments