Curse

Ashius · May 31, 2010, 11:02 AM // 11:02

Quote:

Originally Posted by Rahja the Thief

So... care to point out to me where she mentioned these "false positives"? I certainly am not seeing them... Computer error rate is about 25,000 to 75,000 failures per billion hours of operation. Unless ANET/NSSoft have the WORST LUCK IN HISTORY, I'd say this is a cut and dry issue.

My PhD (and nearly second degree in Electronic Computer Engineering) argues with you...

UMAD?

But you seem to be missing the point that it is highly unlikely that A-Net's methods utilised a completely computer-based method for detecting these bots, and the complexity of making judgements on individual cases cannot be left for a computer to decide. So the two options are to use a system that is completely computer-based and cannot differentiate case-by-case examples, or use a combination of both computer and human based operations. As it is likely the latter is true, then human error is a highly probable possibility.

And while we're quoting titles here,
<PhD in Organisational Psychology.

EDIT:

Quote:

Originally Posted by Rahja the Thief

Bingo, and the likelihood of said "bookkeeping errors" would likely be no more than 1 in every 250 (and that's being extremely generous to the guilty here...). So by that math, that means out of the 3,700 accounts there were banned, ~15 were banned by mistake. That number is likely even fewer. How many threads have we had thus far in Riverside claiming innocence? Oh, more than 15? More than 30? Yep. That's what we call lying kids!

Frankly, some of you are acting like this is quantum mechanics... it isn't, this is general relativity in the realm of comparison. This is a simple situation with an even simpler solution. If you don't understand the methods of computer aided algorithms, then you probably shouldn't try to claim there could or couldn't be inaccuracies.

Hypothesised probabilities mean nothing, so don't use them as a basis for logic.

Lord Sojar · May 31, 2010, 11:06 AM // 11:06

Quote:

Originally Posted by Ashius

But you seem to be missing the point that it is highly unlikely that A-Net's methods utilised a completely computer-based method for detecting these bots, and the complexity of making judgements on individual cases cannot be left for a computer to decide. So the two options are to use a system that is completely computer-based and cannot differentiate case-by-case examples, or use a combination of both computer and human based operations. As it is likely the latter is true, then human error is a highly probable possibility.

And while we're quoting titles here,
<PhD in Organisational Psychology.

Hypothesised probabilities mean nothing, so don't use them as a basis for logic.

You don't think a multi million dollar company used a computer based algorithm for detecting bots that used dll injection and packet manipulation? LOL. Oh ok... that's a fairly naive view, not going to lie....

The system for detection was automated. The results would be sent for further review and litmus testing. If after 1-3 more litmus tests, the account was still found to have the flags the algorithm was intended to find, it would be added to a list of a mass banning. This isn't rocket science, it's actually rather simple.

1= guilty
0= not guilty.

COMPUTER SAYS... 1... Support tech confirms 1 from records. Banned.

COMPUTER SAYS... 0... Account is passed and not subject to further investigation. Not banned.

To your second point: Rationality can be used in place of logic when standards exist, as they do in this case. Rationality + logic = a pretty air tight argument. What side of this are the rest of you arguing exactly? I fail to see where you could possibly have ANY argument here.

Ashius · May 31, 2010, 11:12 AM // 11:12

Quote:

Originally Posted by Rahja the Thief

You don't think a multi million dollar company used a computer based algorithm for detecting bots that used dll injection and packet manipulation? LOL. Oh ok... that's a fairly naive view, not going to lie....

The system for detection was automated. The results would be sent for further review and litmus testing. If after 1-3 more litmus tests, the account was still found to have the flags the algorithm was intended to find, it would be added to a list of a mass banning. This isn't rocket science, it's actually rather simple.

1= guilty
0= not guilty.

COMPUTER SAYS... 1... Support tech confirms 1 from records. Banned.

COMPUTER SAYS... 0... Account is passed and not subject to further investigation. Not banned.

To your second point: Rationality can be used in place of logic when standards exist, as they do in this case. Rationality + logic = a pretty air tight argument. What side of this are the rest of you arguing exactly? I fail to see where you could possibly have ANY argument here.

Sorry, you caught me out on a bad phrasing (i meant detected and action on). You were arguing a couple of posts ago that the whole process is based on binary operation, and as such there is no error in their judgement. But now you're arguing that they do infact use human input? One thing I have discovered is that whenever there is human input, there is the chance for human error.

Lord Sojar · May 31, 2010, 11:20 AM // 11:20

Quote:

Originally Posted by Ashius

Sorry, you caught me out on a bad phrasing (i meant detected and action on). You were arguing a couple of posts ago that the whole process is based on binary operation, and as such there is no error in their judgement. But now you're arguing that they do infact use human input? One thing I have discovered is that whenever there is human input, there is the chance for human error.

It is doubtful that the algorithm was written in binary (by doubtful, I mean it wasn't, period.)

The language doesn't matter....

The fact is, a monkey could write a search algorithm to search for the values they were searching for.

Method:

Test programs to find value mutations/changes

Bot programs 1-6 change value A to value B.
Bot programs 7-11 change value A to value C.
Normal operation changes value A to value D.
Client side operations change value A to value D.
Bot programs 12-14 change value A to E (this 'bot' or mod has no effect on game play and is therefore ignored)

Run search algorithm for value B and C.

If B and/or C are detected, flag account.

Review further instances of value change on flagged account. If positive... BAN.

I'm sorry, did I miss something here?

Are you guys seriously suggesting that something in normal operation or something that is PURELY client side can manipulate the values they were looking at? If you are seriously suggesting that... LOL @ YOU. It helps to know something about the topic you are discussing... typically at least.

This is, again, cut and dry. They didn't make a mistake here... they checked this out, litmus tested it. It's infallible.

Faer · May 31, 2010, 11:26 AM // 11:26

Quote:

Originally Posted by Rahja the Thief

So... care to point out to me where she mentioned these "false positives"?

If you are too lazy to read what has been posted in this very thread, then I am far too lazy to repeat what has already been said for your benefit.

Ashius · May 31, 2010, 11:29 AM // 11:29

Quote:

Originally Posted by Rahja the Thief

It is doubtful that the algorithm was written in binary (by doubtful, I mean it wasn't, period.)

The language doesn't matter....

The fact is, a monkey could write a search algorithm to search for the values they were searching for.

Method:

Test programs to find value mutations/changes

Bot programs 1-6 change value A to value B.
Bot programs 7-11 change value A to value C.
Normal operation changes value A to value D.
Client side operations change value A to value D.
Bot programs 12-14 change value A to E (this 'bot' or mod has no effect on game play and is therefore ignored)

Run search algorithm for value B and C.

If B and/or C are detected, flag account.

Review further instances of value change on flagged account. If positive... BAN.

I'm sorry, did I miss something here?

Are you guys seriously suggesting that something in normal operation or something that is PURELY client side can manipulate the values they were looking at? If you are seriously suggesting that... LOL @ YOU. It helps to know something about the topic you are discussing... typically at least.

This is, again, cut and dry. They didn't make a mistake here... they checked this out, litmus tested it. It's infallible.

I've bolded, underlined and italicised the factor which can fail, and it is due to what I have been arguing this whole time, human error. And I used binary as your original explanation used this as an example, but this has no real merit in this discussion because this process still produces empirical evidence which must be analysed. Scientific research 101. And when analysis is performed, there is no guarentee that the analysis will be flawless.

Lord Sojar · May 31, 2010, 11:36 AM // 11:36

Quote:

Originally Posted by Ashius

I've bolded, underlined and italicised the factor which can fail, and it is due to what I have been arguing this whole time, human error. And I used binary as your original explanation used this as an example, but this has no real merit in this discussion because this process still produces empirical evidence which must be analysed. Scientific research 101. And when analysis is performed, there is no guarentee that the analysis will be flawless.

The review is really only a human confirmation of a computer certainty at this point. It simple serves as a last check. The computer algorithm is already correct, and that is the point people are failing to realize. This is a black and white issue, and people are trying to treat computer algorithms like a gray area. Computers aren't gray guys and gals, sorry to say. The review was for the benefit of those identified as cheaters/botters, not for their determent. ANET didn't even have to review the flagged accounts, they could have done it in a totally automated manner, and it likely would have yielded the same results.

If anything, the second review may have let a botter or 2 slide by, because the human judgment overruled the computer detected value changes I elaborated on above. So... that's game, set, and match.

Quote:

Originally Posted by Theocrat

If you are too lazy to read what has been posted in this very thread, then I am far too lazy to repeat what has already been said for your benefit.

To suggest I am lazy after posting an explanation and quoting her on her most RECENT statements is rather bold. But, that's your choice Theo. I'm not going to publicly argue with another mod over such a black and white issue, where there is no gray area.

If an account was banned wrongfully, it was a fluke, a random error, and not the norm. When you submit a ticket to support regarding this, they will do yet another check on your account, as it doesn't take long. If this additional check corrects a very, very rare error, so be it. Likelihood of that...? Slim to none. They are giving you automated responses if their ADDITIONAL review has still found you to be guilty. Anyone arguing past that point is simply lying.

Ashius · May 31, 2010, 11:45 AM // 11:45

Quote:

Originally Posted by Rahja the Thief

The review is really only a human confirmation of a computer certainty at this point. It simple serves as a last check. The computer algorithm is already correct, and that is the point people are failing to realize. This is a black and white issue, and people are trying to treat computer algorithms like a gray area. Computers aren't gray guys and gals, sorry to say. The review was for the benefit of those identified as cheaters/botters, not for their determent. ANET didn't even have to review the flagged accounts, they could have done it in a totally automated manner, and it likely would have yielded the same results.

If anything, the second review may have let a botter or 2 slide by, because the human judgment overruled the computer detected value changes I elaborated on above. So... that's game, set, and match.

As you do not have access to the methods that A-Net use for detection, then what you are arguing is only a hypothesis. A-Net could have simply set a quota for .dll injections, and used that as their evidence to ban. The method you have stated would have been a great method to use, but A-Net do not exactly have a track record for selecting the best possible tool for the situation.

Lord Sojar · May 31, 2010, 11:48 AM // 11:48

Quote:

Originally Posted by Ashius

As you do not have access to the methods that A-Net use for detection, then what you are arguing is only a hypothesis. A-Net could have simply set a quota for .dll injections, and used that as their evidence to ban. The method you have stated would have been a great method to use, but A-Net do not exactly have a track record for selecting the best possible tool for the situation.

Going off the fact that this was done by a multi million dollar company with stock holders, it would be assumed they used a fairly safe, reasonable method that was thought out enough to be effective, without being to broad, as to ban those that didn't deserve it. My point stands, hypothesis or not. The vast majority of these bans were legitimate, and those that were wrongfully banned will be unbanned.

GoF · May 31, 2010, 11:51 AM // 11:51

I highly doubt they let the Asurans take care of the algorithms, so we cannot be sure that the algorithms are flawless, right?

Ashius · May 31, 2010, 11:54 AM // 11:54

Quote:

Originally Posted by Rahja the Thief

Going off the fact that this was done by a multi million dollar company with stock holders, it would be assumed they used a fairly safe, reasonable method that was thought out enough to be effective, without being to broad, as to ban those that didn't deserve it. My point stands, hypothesis or not. The vast majority of these bans were legitimate, and those that were wrongfully banned will be unbanned.

Huh? So your concluding statement actually agrees with me that human error existed in their judgement? I'm now a little confused as to what you were arguing now...

Lord Sojar · May 31, 2010, 11:56 AM // 11:56

Quote:

Originally Posted by GoF

I highly doubt they let the Asurans take care of the algorithms, so we cannot be sure that the algorithms are flawless, right?

Judging from the amount of bans that were confirmed as valid from the people admitting the use of bots both here on Guru and on the Wiki, it is safe to assume the algorithm was dead on. Again, this type of algorithm would be extraordinarily easy to create, and litmus testing would ensure its accuracy. They did a press release on this matter, and stock investors would shit bricks if this wasn't done perfectly, by the books, and in the best interest of the company.

So... you can bet this was done with extreme scrutiny and in a very controlled, logical manner. That, again, isn't to suggest ANET/NCSoft aren't infallible, but it is suggesting that in this case, 99%+ of the bans were legitimate. That is an excellent value to aim for, and we have seen far more than 1% of the banned people claiming innocence. It's far easier to assume those people are lying than claiming computer error in an investigation that has spanned several MONTHS.

Quote:

Originally Posted by Ashius

Huh? So your concluding statement actually agrees with me that human error existed in their judgement? I'm now a little confused as to what you were arguing now...

No, I'm not agreeing with you, at least not in the sense you feel I am. The human error I point to would simply be a bookkeeping error. The algorithm for this sort of detection is rather simple, it must might contain many variables. It is possible, however unlikely, that the algorithm was flawed, but the likelihood of someone lying is far, far higher than the algorithm being fundamentally flawed. You supposedly have a PhD in psychology, you should know that...

A public posting of innocence screams, "I'm trying to get people to believe me so that I look better to ANET, or I can start some kind of public outcry to make it seem like reversing this decision will get good press!", instead of that user just sending an email to support and having them triple check things yet again. These automated responses are only sent after they RECHECK it.

Quote:

Originally Posted by Ashius

As you do not have access to the methods that A-Net use for detection, then what you are arguing is only a hypothesis. A-Net could have simply set a quota for .dll injections, and used that as their evidence to ban. The method you have stated would have been a great method to use, but A-Net do not exactly have a track record for selecting the best possible tool for the situation.

The "evidence" would argue the quota hypothesis. My hypothesis is the most logical and rational based on all available data. This takes into account those that were banned for excessive bots and those that simply tried the bot to see it work, perhaps only using it for a few minutes on one occasion. They were all banned, without differentiation. That implies, simply, that it was a return value based search, most likely looking at packet data. Specific values were IDed, and targeted for a search through the database, which according to Gaile, went months back.

Ashius · May 31, 2010, 12:11 PM // 12:11

Quote:

Originally Posted by Rahja the Thief

No, I'm not agreeing with you, at least not in the sense you feel I am. The human error I point to would simply be a bookkeeping error. The algorithm for this sort of detection is rather simple, it must might contain many variables. It is possible, however unlikely, that the algorithm was flawed, but the likelihood of someone lying is far, far higher than the algorithm being fundamentally flawed. You supposedly have a PhD in psychology, you should know that...

A public posting of innocence screams, "I'm trying to get people to believe me so that I look better to ANET, or I can start some kind of public outcry to make it seem like reversing this decision will get good press!", instead of that user just sending an email to support and having them triple check things yet again. These automated responses are only sent after they RECHECK it.

This isn't an argument with respects to the likelihood of lying compared to the results of a computer algorithm, it is an argument about the presence of human error, which you were arguing that there was none of (at least in the beginning).

And personal attacks are a characteristic of a weak argument, attack my argument, not my person. But surely one who also a member of the world of academia understands this?

Lord Sojar · May 31, 2010, 12:16 PM // 12:16

Quote:

Originally Posted by Ashius

This isn't an argument with respects to the likelihood of lying compared to the results of a computer algorithm, it is an argument about the presence of human error, which you were arguing that there was none of (at least in the beginning).

And personal attacks are a characteristic of a weak argument, attack my argument, not my person. But surely one who also a member of the world of academia understands this?

Tongue in cheek. I wasn't attacking you personally, simply pointing out that psychology would strongly suggest my point.

I never said there wasn't human error, I said there was no computer error. Any human error would be eliminated by an ADDITIONAL check through the data collected, and checking the determination. If there is one poor soul, who after an additional review, was truly innocent, and their ban wasn't lifted, I do feel sorry for them, but the chances are about as high as a human being able to lift a full size Blue Whale with one finger. Human error, in this case, would be very small, and would not account for the number of reports of innocence we have seen, on Guru, let alone other sites.

I'm going to bed now, having made all the points I could make on this topic. I will agree to disagree with anyone who would argue against the overall accuracy of this ban given the available data.

Goodnight.

Ashius · May 31, 2010, 12:20 PM // 12:20

Quote:

Originally Posted by Rahja the Thief

Tongue in cheek. I wasn't attacking you personally, simply pointing out that psychology would strongly suggest my point.

I never said there wasn't human error, I said there was no computer error. Human error, in this case, would be very small, and would not account for the number of reports of innocence we have seen, on Guru, let alone other sites. I'm going to bed now, having made all the points I could make on this topic. I will agree to disagree with anyone who would argue against the overall accuracy of this ban given the available data.

Goodnight.

This little exercise was fun ^^ It is good to know that there are other intelligent individuals out there, who have the ability to present a valid argument. Take care.

JoeKnowMo · May 31, 2010, 12:48 PM // 12:48

Quote:

Originally Posted by Rahja the Thief

If there is one poor soul, who after an additional review, was truly innocent, and their ban wasn't lifted, I do feel sorry for them, but the chances are about as high as a human being able to lift a full size Blue Whale with one finger. Human error, in this case, would be very small, and would not account for the number of reports of innocence we have seen, on Guru, let alone other sites.

Most of the claims of 'innocence' are by those who claim they tried the bots out because they were curious. I'm inclined to believe that the human error rate was very low given the process you outlined.

GreenX · May 31, 2010, 01:54 PM // 13:54

rahja could u plz stop bsing;

tell me what if legal program 15 changed value A to B?
false positive?

who would check all 3700 accounts for false positives reviewing logs and such?

indeed they wouldn't they would just ban all, and w8 for people who think they got banned unjustly to ask for revision!

and this last point is what people are discontent about as customer support is failing to give a **** about them, and anet is simply using a bot to reply back to them.

furthermore there are bots who work through pixel detection (farming bots) or are simply macros (spam bots) which don't change any values and as you setup in ur hypothetical situation value A will be changed to value D and thus these bot will and are going below the radar.

Gill Halendt · May 31, 2010, 02:07 PM // 14:07

Quote:

Originally Posted by GreenX

tell me what if legal program 15 changed value A to B?
false positive?

You know what "professional" means, do you?

Are you suggesting that prople working at this system are just amateurs?

They've set criteria to detect bots in their system and obviously NO "legal program" would have changed value A to B. They know their code better than you and they surely know how to check their code better than you as well.

If anything, if it was so easy to have "false positives", the consequent ban slaughter would have been massive and clearly evident.

Please, get serious.

Ka Tet · May 31, 2010, 02:48 PM // 14:48

Quote:

Originally Posted by Rahja the Thief

This is, again, cut and dry. They didn't make a mistake here... they checked this out, litmus tested it. It's infallible.

It is only cut and dry if you are willing to assume a certain set of assumptions.

Zarion Silverarrow · May 31, 2010, 03:24 PM // 15:24

I don't care if Google,Microsoft,Apple,Sony,and Nintendo made a company together to make this detector. It can still go wrong,because humans made it.

1. If it was tested,I doubt (repeat doubt) that it was tested 3.7k times

2. I know they went wrong on something,because a friend I know IRL was banned,and I know he has no knowlage of this things. You won't find a more casual gamer than him,Jesus Christ he just recently found out about Fallout 3 and other games having console commands. He did not use any third party programs,or buy any money from GW. He hasn't broken any rule other than language (and I'm just puting in the language thing to say something)

Microsoft,one of the most known software companies,has errors on ALL of their OS's (see Windows Vista)...why it has errors? humans made it. Same with Apple...hence all programs can f*ck up.

I know it sounds cheesy,but it is the truth.