Albuquerque said:
And back to the point, you're fixating on a logical fallacy... You are making the assumption that data is missing, yet you have no way to back that up. You are simply picking your numbers completely out of thin air, with absolutely ZERO to back you.
I never picked a number. In fact, I made it a point to say that I have no idea what the percentages and that I wouldn't even try to guess at one. The only statement I ever made on the odds was that 50% was probably an over-optimistic number. Reread my posts on this thread, there isn't a single place where I surmise the exact odds.
The NUMBERS from [H]ardforum, from Rage3D, from here, from ClubSi Out There, from numerous other places all suggest that YOU are in the wrong. I've seen a lot of people claim failure, I've seen the same number claim success. Every poll, every thread and every general discussion shows the same results: approximately 50/50.
Until YOU CAN FIND AND PROVE better numbers, the CURRENT DATA is 50/50. And contrary to your illogical assumption from above, nobody is being TOLD to EXPECT a 9700 out of any given 9500np. And anyone who DOES EXPECT a 9700 out of it is a t00L.
I never argued that the current numbers
aren't approximately 50/50, my contention is that the way we have arrived at these numbers is unscientific at best and wholly misleading at worst.
50/50 doesn't mean a guarantee, and anyone who's been through Basic Algebra knows this. In fact, even people who didn't pass Algebra likely already know this. In fact, I don't recall anyone even mentioning it was a sure thing, so you can be dropping that part of your argument as well.
Again, I don't know what you're arguing. Did I ever once say 50/50 meant a guarantee? The only thing I've been arguing (and I hate to repeat myself) is that the 50/50ish odds seen on these and other boards is misleading.
Here are a few reasons why:
1. Polls inherently are flawed. This is why all polls say they have a +-5% error (or whatever). Furthermore, most of the polls that state these types of errors generally have statiticians that do calculations on
why they think the polls are off by the amount they claim. These inaccuracies are due to a myriad of reasons including the population pool they're gathered from (for example, people who live in different regions often have vastly different opinions or whatnot when it comes to polls); general mathematical errors inherent in something as nebulous as collecting polls from humans; and also the number of people who submit votes on polls (for instance, a poll that that asks the opinion of one million Americans will be a lot more accurate than one that asks the same question to only one thousand).
AFAIK, the polls on message boards do not have such staticians conducting calculations on polls. So as flawed as polls are to begin with, they are even MORE erroneous given that we don't have the luxury of hiring someone to do such calculations.
2. Polls are inherently flawed Part II. On CNN, they often have polls asking "What do you think of the president's performance?" And then, you have something like 65% think he's doing a good job. But what does that mean? To New Yorkers, a good job might mean, "well, at least he hasn't blown up the country yet" while to other people, it might mean, "I don't like his dog, so he sucks." In short, we have to be careful as to the meaning these polls might generate. Now, when I said people might be lying when it comes to the effectiveness of their softmods, I do not necessarily think that the majority of people who answered "yes" are maliciously lying to skew the polls. A lot of us (myself included), sometimes stretch or bend the truth to better suit our perceptions. For instance, a lot of people consider "stable" being able to run Prime 95 for 6 hours straight, whereas other people won't be happy until 24 hours pass without a single hiccup. On the same token, some people, who experience very minor artifacting on their soft-modded cards may convincingly say to themselves, "well, it isn't THAT bad, so it pretty much is a 100% success!" On the other hand, some people may have applied the hacked driver incorrectly, and without trying to do it correctly, may have thrown up their hands in despair and said, "SCREW THIS! It doesn't work!!"
3. Numbers can and often do lie. What about the people, such as myself, who bought
multiple 9500NPs? AFAIK, you can only vote once on these polls, so someone like me wouldn't be able to vote on all three failed cards. Right now, there are a hundred votes on the poll on oc-forums. If I were able to vote more than once, my votes
alone would skew the results by nearly 3%. Now, what if there are 5 such people as myself who bought 3 9500nps? That means there ten cards unaccounted for. Let us further assume that these 5 people also all experienced failed softmods. That would constitute nearly a 15% change in the polls. On the flip side, maybe these people experienced multiple successful mods. In that case, our polls would severely
underestimate the success of the softmod. Whatever may be the case, it is obviously clear that the numbers cannot be completely trusted. And what about the people that bought a 9500NP, failed to softmod, RMAed the card until they got one that worked, and
then voted? There is no provision on these polls to deal with such problems.
Perhaps you're thinking that these problems would all cancel each other out? Maybe. I don't know enough about stats to wager a guess on that. However, this observation as well as the fact that there are only 100 votes on the poll make me very suspect. I don't think 100 votes is nearly enough to definitively show what the true percentages are. This would be similar to flipping a coin only ten times, getting 7 heads and 3 tails, then claiming that the odds of getting heads is 70%.
