This site makes extensive use of JavaScript.
Please enable JavaScript in your browser.
Live
PTR
10.2.5
PTR
10.2.6
Why Are Horde Better In BattleGrounds?
Post Reply
Return to board index
Post by
Monday
The question is, were the Horde winning just recently? Were they winning a few months ago? Were they winning today, between the hours of twelve and two? Were they winning two months and six days ago between the hours of 5 P.M. and 9:45 P.M.?
That's what HsR is talking about. The Horde have a total of more wins, but it's impossible to tell if it was because they won more every day, whether they had a huge winning streak that died off for awhile, etc.
Edit: I'm assuming that's what he's referring to, anyways. Sometimes it can be hard to tell <3
Post by
Hyperspacerebel
The question is, were the Horde winning just recently? Were they winning a few months ago? Were they winning today, between the hours of twelve and two? Were they winning two months and six days ago between the hours of 5 P.M. and 9:45 P.M.?
That's what HsR is talking about. The Horde have a total of more wins, but it's impossible to tell if it was because they won more every day, whether they had a huge winning streak that died off for awhile, etc.
Edit: I'm assuming that's what he's referring to, anyways. Sometimes it can be hard to tell <3
lol <3
@Pallyalt Essentially what I'm saying is that the raw data provided is not enough by itself to support the implicit statistical claim being made is that being horde and winning are significantly correlated. I go back to the coin flipping scenario as a simplified example. Just simply comparing the number of times I flip heads versus the number of times I flip tails tells me very little about the probabilistic nature of coin flipping. Maybe the heads side is 10% heavier than the tales side, but by chance I am getting near equal representation. Just comparing the two numbers tells me what has happened, but not what will happen. And it's the same situation here. You're listing raw data, but you have not demonstrated that the data actually provides a significant correlation.
And I'm not telling you that you have to go and run those numbers through R or SPSS or anything, but I am saying that without that you can't claim there is any correlations present.
Post by
Pallyalt
The question is, were the Horde winning just recently? Were they winning a few months ago? Were they winning today, between the hours of twelve and two? Were they winning two months and six days ago between the hours of 5 P.M. and 9:45 P.M.?
That's what HsR is talking about. The Horde have a total of more wins, but it's impossible to tell if it was because they won more every day, whether they had a huge winning streak that died off for awhile, etc.
Edit: I'm assuming that's what he's referring to, anyways. Sometimes it can be hard to tell <3
lol <3
@Pallyalt Essentially what I'm saying is that the raw data provided is not enough by itself to support the implicit statistical claim being made is that being horde and winning are significantly correlated. I go back to the coin flipping scenario as a simplified example. Just simply comparing the number of times I flip heads versus the number of times I flip tails tells me very little about the probabilistic nature of coin flipping. Maybe the heads side is 10% heavier than the tales side, but by chance I am getting near equal representation. Just comparing the two numbers tells me what has happened, but not what will happen. And it's the same situation here. You're listing raw data, but you have not demonstrated that the data actually provides a significant correlation.
And I'm not telling you that you have to go and run those numbers through R or SPSS or anything, but I am saying that without that you can't claim there is any correlations present.
If you read back to my original reply, you'll see that I'm only talking about the 8 months I have had the ability to see bg stats. During that time, the horde have held a consistent lead over alliance for wins on a daily basis without exception (on
my
server). I'm in no way implying horde are better than alliance, or that it is written in stone that horde will always win. This may not have always been the case and tomorrow the tides could change. I'm simply speaking about what
has
happened and why I believe the numbers fell the way they did.
As you stated, I'm dealing with the "raw data". The end of day numbers that show which faction won the most bgs, has had horde on top on a daily basis for the past 8 months. It is what it is, just like the coin toss you mentioned. If I flip that coin 500 times and 450 of those times I flipped heads, I don't need to analyze it to know that I flipped heads more times than tails. It's simply a fact. Does it mean I'll always flip heads, or that tails is heavier than heads? I have no way of knowing that without doing that "analysis" you mentioned. But I
can
state with total honesty and know that I'm 100% right when I say I flipped heads more than tails.
I believe I was pretty straightforward when I discussed my opinions and pretty clear that I was stating an opinion. What I referred to as fact was nothing more than "raw data" and other than maybe a poor choice of words, I think I did relay this.
Post by
Hyperspacerebel
You're overlooking a word I've used several times: "significant". You're parading these results around, but you have not disproved the null hypothesis (proved the results significant). They are irrelevant numbers to the discussion if they are the result of pure randomness, and all your efforts to explain them are meaningless. You're doing the same thing that a lot of pop psychology does: taking insignificant results (or at least in this case results that haven't been shown to be significant yet) and trying to explain them.
Post by
Adamsm
I've seen the same thing on both sides myself: I'll enter into a BG on either side, and it's the losing one.
Post by
Pallyalt
You're overlooking a word I've used several times: "significant". You're parading these results around, but you have not disproved the null hypothesis (proved the results significant). They are irrelevant numbers to the discussion is they are the result of pure randomness, and all your efforts to explain them are meaningless. You're doing the same thing that a lot of pop psychology does: taking insignificant results (or at least in this case results that haven't been shown to be significant yet) and trying to explain them.
Dude, come back down to earth. "disproved the null hypothesis"... "proved the results significant"... LMAO All that technical jargon doesn't make you sound intelligent, it only makes you look arrogant. As for significance, in all your attempts to impress, you've missed the original question of this post and in the context of this post, my statements are quite relevant.
As for the statement that my numbers "are the result of pure randomness", is that an opinion, or did you quantify it?
Post by
Hyperspacerebel
If you don't want statistical jargon, don't make claims about statistical fact. The words I'm using have very specific meanings, that deal with very specific concepts within the field of statistics. Whether they make me look arrogant or not isn't relevant if they make me right.
As for significance...
You misunderstand. Significance has a
very specific
statistical meaning. I'm not talking about the relevance of your statements, I'm talking about whether your numbers have a high probability of not having been caused by chance or not. A purely statistical consideration.
And as for the statement that my numbers "are the result of Pure randomness", is that an opinion, or did you quantify it?
I didn't say your numbers were the result of pure randomness. I said that you haven't demonstrated that they aren't. But the real divisive issue is not even that you haven't done that... it's the fact that you are so vehemently against any notion that that might be a consideration.
But the crux of the matter appears to be that you do not have a working knowledge of statistics, neither the terms is uses nor the concepts it deals with, which is why this is going nowhere. So, I'm going to just take all your statements on the subject as anecdotal, based purely on personal experience instead of making any statistical claim. That way there is no longer any contention on the issue, since I don't have access to anything other than anecdotal evidence either.
Post by
Pallyalt
If you don't want statistical jargon, don't make claims about statistical fact. The words I'm using have very specific meanings, that deal with very specific concepts within the field of statistics. Whether they make me look arrogant or not isn't relevant if they make me right.
As for significance...
You misunderstand. Significance has a
very specific
statistical meaning. I'm not talking about the relevance of your statements, I'm talking about whether your numbers have a high probability of not having been caused by chance or not. A purely statistical consideration.
And as for the statement that my numbers "are the result of Pure randomness", is that an opinion, or did you quantify it?
I didn't say your numbers were the result of pure randomness. I said that you haven't demonstrated that they aren't. But the real divisive issue is not even that you haven't done that... it's the fact that you are so vehemently against any notion that that might be a consideration.
Again, you overlook the fact that I wasn't really discussing cause and effect, nor was I discussing probability. I already admitted I made a poor use of words. The use of "statistical" was out of context, but the point I was making wasn't that hard to get, even with the accidental misuse of one word.
As for demonstrating my numbers weren't random, anyone with a BG stat addon can see for themselves that they aren't. We already went over the whole screenshot thing and while I can't provide physical "proof" that the numbers were accurate, anyone with a computer and WOW account can get the same stats as me.
Post by
Hyperspacerebel
As for demonstrating my numbers weren't random, anyone with a BG stat addon can see for themselves that they aren't.
No, they can't. Unless they are themselves running the proper statistical tests. Randomness and correlation don't work the way you seem to think they do.
Post by
Squishalot
I haven't seen the data, but in fairness, the data sounds like it would be statistically significant especially when running paired tests (which you can, since each day's alliance / horde wins can be matched up).
Post by
Hyperspacerebel
I haven't seen the data, but in fairness, the data sounds like it would be statistically significant especially when running paired tests (which you can, since each day's alliance / horde wins can be matched up).
Them matching up only shows correlation between one's wins and the other's losses (a perfect +1 correlation), but not the significance of being a horde to winning.
Again, back to the coin. Flipping heads correlates to not flipping tails perfectly. But that doesn't change the fact that any set of coin-flip data you obtain is the result of randomness (or pseudorandomness depending on how much determinism you attribute to the act of coin-flipping.... for all intents and purposes though, the coin-flip example assumes randomness), whether it's 50/50 or 90/10.
Specific to the numbers at the top of the page, imagine several thousand alliance and horde flipping coins instead of playing matches. Heads horde wins, tails alliance wins. Now it's perfectly possible that using that method, those exact same numbers at the top of the page for those four days could be reached. So how do we determine the significance of those numbers and whether we can try to draw meaningful conclusions from them? We figure out what the chances are that those numbers came about by chance. If that chance is low enough, we can conclude that the results are significant. There's a bit more math involved than that, but that's the gist of it.
Post by
Squishalot
You're speaking to me like you've forgotten I've got degrees in actuarial science :P
Your null hypothesis is that your level of horde wins is equal to your level of alliance wins on any given day. So we can:
a) Test # horde wins vs # alliance wins (lowest power, since we're ignoring that they're perfectly correlated)
b) Test % horde wins (i.e. horde wins / (horde + alliance wins)) vs null 50%
c) Test margin of horde daily wins (i.e. horde wins - alliance wins) vs null 0
Again, inferring what we can about the data, it sounds like Pallyalt has around 200+ data points that each show:
a) Horde wins > alliance wins
b) Horde win% > 50%
c) Horde wins - alliance wins > 0
... with very few / no data points that demonstrate the reverse. So on the balance of probabilities, I would suggest that we are likely to draw a significant result in favour of horde winning more frequently, caveated by the fact that I agree we won't know for certain until we stat it all out.
Post by
Hyperspacerebel
You're speaking to me like you've forgotten I've got degrees in actuarial science :P
I honestly had no idea until now what you went to school for until now :P
Your null hypothesis is that your level of horde wins is equal to your level of alliance wins on any given day.
That's a fairly weak null hypothesis, no? The initial hypothesis isn't that "the level of horde wins is not equal to the level of alliance wins on any given day" (in which case that would be a perfect null hypothesis). Rather the initial hypothesis is that horde have a higher chance of winning any given battleground. For that to be true does not require that the level of horde wins be equal to the level of alliance wins in a given day. That's one possibility, but does not cover the full reach of what the null hypothesis should account for.
Again, inferring what we can about the data, it sounds like Pallyalt has around 200+ data points that each show:
a) Horde wins > alliance wins
b) Horde win% > 50%
c) Horde wins - alliance wins > 0
There are many factors at play, not the least of which is an extremely narrow sample population many of whom are being sampled again and again over the sample period. Does this data have horde winning more bgs? Yes. Is that significant to the hypothesis that being horde means you will win more bgs? No idea. Is it significant to the hypothesis that if you played on Pallyalt's battlegroup for the last 200 days between certain hours that you were more likely to win as horde? Yes.
So on the Balance of probabilities, I would suggest that we are likely to draw a significant result in favour of horde winning more frequently, caveated by the fact that I agree we won't know for certain until we stat it all out.
When it comes down to it, we don't have a proper sample of data points, which I would have very much liked to see from that post ElhonnnaDS linked, so I don't know how I feel about your claim. We have one person's biased sample and an overall average of another person's (presumably unbiased) sample. It's perfectly fine to form a hypothesis about them..... that's what you do before running the data in science afterall. But I don't think there's
significant
evidence of a significant correlation just from an immediate glance.
Post by
Squishalot
That's a fairly weak null hypothesis, no?
Ok, we can increase the threshold of the test by changing the null to '<= alliance wins', '<=50%', and '<=0' respectively. I don't think that changes things tremendously, but certainly makes the statistical significance a bit harder to achieve.
There are many factors at play, not the least of which is an extremely narrow sample population many of whom are being sampled again and again over the sample period. Does this data have horde winning more bgs? Yes. Is that significant to the hypothesis that being horde means you will win more bgs? No idea. Is it significant to the hypothesis that if you played on Pallyalt's battlegroup for the last 200 days between certain hours that you were more likely to win as horde? Yes.
Is the sample data biased? Certainly, but not necessarily due to the person's viewpoint, but rather, due to the method of collection. Is it reflective of the broader population? We can't be certain, but the whole point of sampling is to try to infer what the population is in lieu of having perfect information.
When it comes down to it, we don't have a proper sample of data points, which I would have very much liked to see from that post ElhonnnaDS linked, so I don't know how I feel about your claim. We have one person's biased sample and an overall average of another person's (presumably unbiased) sample. It's perfectly fine to form a hypothesis about them..... that's what you do before running the data in science afterall. But I don't think there's significant evidence of a significant correlation just from an immediate glance.
I'm not saying that there
is
significant evidence, I'm just suggesting that based on what the one person has said about his data, it would be likely to draw a significant result in line with his assertion. In that respect, criticising him for not conducting the analysis and lecturing about statistics and probability is a bit of a cheap shot when the analysis would support his theory - rather, you should be criticising the data for being incomplete and biased, as long as you could demonstrate a valid reason why his sample wouldn't be reflective of the broader population.
Post by
Hyperspacerebel
I'm not saying that there is significant evidence, I'm just suggesting that based on what the one person has said about his data, it would be likely to draw a significant result in line with his assertion. In that respect, criticising him for not conducting the analysis and lecturing about statistics and probability is a bit of a cheap shot when the analysis would support his theory - rather, you should be criticising the data for being incomplete and biased, as long as you could demonstrate a valid reason why his sample wouldn't be reflective of the broader population.
I brought those data criticisms up to you, because you have demonstrated an understanding of the need for a proper statistical analysis to achieve sure knowledge on the matter. I think Pallyalt commited an error that is more fundamental than just sample bias, namely in his claim that it's a statistical fact that Horde dominate bgs (which he has since retracted) and his claim that anyone looking at their own data can know for sure that variations in their results aren't the result of random variation.
Post by
Squishalot
I brought those data criticisms up to you, because you have demonstrated an understanding of the need for a proper statistical analysis to achieve sure knowledge on the matter. I think Pallyalt commited an error that is more fundamental than just sample bias, namely in his claim that it's a statistical fact that Horde dominate bgs (which he has since retracted) and his claim that anyone looking at their own data can know for sure that variations in their results aren't the result of random variation.
Fair enough.
I honestly had no idea until now what you went to school for until now :P
Really? I'm pretty sure we've discussed it before... Oh well, never mind then!
Post by
Rankkor
You know, I just wondered......... why is this topic here? Its related to wow, therefore it should be in the General section, or at least in the pvp section, not Off-topic.
Post by
ElhonnaDS
You know, I just wondered......... why is this topic here? Its related to wow, therefore it should be in the General section, or at least in the pvp section, not Off-topic.
It's been here since 2008. I think that gives it squatter's rights.
Post by
kwizzlix
Probably because we have to wait 8+ minutes for our BG que to pop, therefore we dont want to lose. Alliance on the other hand get instant ques, so it doesn't matter if they lose or not.
Post by
Lordplatypus
The most annoyingly true answer?
Hordealts.
I've seen hordies get on alliance alts and do things like take the flag after the FC drops it to give it to someone else and
run it to the horde base
. I've seen them literally run naked into mobs over and over in AV and IoC to lower our reinforcement counts. I've seen them call false incs. I've seen them lifegrip the FC off a cliff in Eots.
Basically, hordies are willing to get on alliance alts and repeatedly pull low tricks like that in order to win.
EDIT: This isn't just stupidity. i've had people who while the horde was just about to win even say FOR THE HORDE ALLIANCE NOOBS or #$%^ like that.
This is basically like in dota where one person goes a hero like wisp and teleports into the opponent fountain over and over with a victim.
Post Reply
You are not logged in. Please
log in
to post a reply or
register
if you don't already have an account.