Thogard wrote: »

Theignson wrote: »

Thogard my main is an AR46 stam DK, my second is AR39 stam warden, I have been around, although not as much as you and others. I've Pvped virtually every day in the last year, exclusively in cyrodil. I have even fought your group a few times in cyrodil.

I shouldn't have said "every time" because, as I indicated, the targeted player can move , and as you noted, break free etc.

But your posts are informed enough that you should assess what I said about statistics. I assume you have some knowledge of statistical analysis, standard deviation, etc.

I have logged players who have landed the exact same combo on me (well on my squishy 3rd, a stamblade) with almost identical .002 second delays between their light attacks and the abilities in every sequence. Consistently doing this, I argue is statistically impossible given human physiology.

Maybe you can reproduce this, I'd like to know. I do admire you for showing your keystrokes.

I don't think its common, but on the other hand I haven't looked much.

If you think about it, people could just assign a light attack macro (as described in my last post) for each of their 5 abilities and use a multi=key mouse. Then they would have perfect weaving each time they used an ability.

If you’re logging it on your end then that means they’re getting no variance in their route to the server and then no server processing time variance (lol) and finally no variance on the servers route to you. Or if logs are server side then only the first two...

Regardless, your logs are meaningless.

Only client side logs from the person accused would be relevant

When the receiver of the attacks consistently sees a 0.002 second delay between them, it is safe to assume the sender actually sent them with 0.002 seconds delay between them.

While internet can mess up results, it generally increases randomness, not decreases it.

Thogard wrote: »

Sharee wrote: »

Thogard wrote: »

Theignson wrote: »

Thogard my main is an AR46 stam DK, my second is AR39 stam warden, I have been around, although not as much as you and others. I've Pvped virtually every day in the last year, exclusively in cyrodil. I have even fought your group a few times in cyrodil.

I shouldn't have said "every time" because, as I indicated, the targeted player can move , and as you noted, break free etc.

But your posts are informed enough that you should assess what I said about statistics. I assume you have some knowledge of statistical analysis, standard deviation, etc.

I have logged players who have landed the exact same combo on me (well on my squishy 3rd, a stamblade) with almost identical .002 second delays between their light attacks and the abilities in every sequence. Consistently doing this, I argue is statistically impossible given human physiology.

Maybe you can reproduce this, I'd like to know. I do admire you for showing your keystrokes.

I don't think its common, but on the other hand I haven't looked much.

If you think about it, people could just assign a light attack macro (as described in my last post) for each of their 5 abilities and use a multi=key mouse. Then they would have perfect weaving each time they used an ability.

If you’re logging it on your end then that means they’re getting no variance in their route to the server and then no server processing time variance (lol) and finally no variance on the servers route to you. Or if logs are server side then only the first two...

Regardless, your logs are meaningless.

Only client side logs from the person accused would be relevant

When the receiver of the attacks consistently sees a 0.002 second delay between them, it is safe to assume the sender actually sent them with 0.002 seconds delay between them.

While internet can mess up results, it generally increases randomness, not decreases it.

Honestly that’s a safer assumption than assuming the internet connection and the server processing had absolutely 0 variability. The latter is pretty much the only factor we can rule out outright... the others are possible albeit unlikely.

While internet is far from reliable, the variability of the unreliability is not so bad that it would cause three attacks done virtually at the same time to have wildly variable delays between them when they arrive.

What i mean: the internet can cause the whole batch of three attacks to arrive in 100 ms (if the weather is good) or in 500 ms(if the weather is bad), but it will not cause the delay between the first attack and the second to be wildly different from the delay between the second attack and the third.

For that to happen, the "weather" would have to change in the time it takes for the whole batch to be sent: you have good connection at T+0, then a bad connection at T+0.002s, and then a good connection again at T+0.004s. Internet is not that volatile.

For this reason, if you see incoming attacks (sent within a very short timeframe) to have consistently the same delay between them when they arrive, they most likely also had the same delay between them on the sender's side.

Sharee wrote: »

Thogard wrote: »

Sharee wrote: »

Thogard wrote: »

Theignson wrote: »

Thogard my main is an AR46 stam DK, my second is AR39 stam warden, I have been around, although not as much as you and others. I've Pvped virtually every day in the last year, exclusively in cyrodil. I have even fought your group a few times in cyrodil.

I shouldn't have said "every time" because, as I indicated, the targeted player can move , and as you noted, break free etc.

But your posts are informed enough that you should assess what I said about statistics. I assume you have some knowledge of statistical analysis, standard deviation, etc.

I have logged players who have landed the exact same combo on me (well on my squishy 3rd, a stamblade) with almost identical .002 second delays between their light attacks and the abilities in every sequence. Consistently doing this, I argue is statistically impossible given human physiology.

Maybe you can reproduce this, I'd like to know. I do admire you for showing your keystrokes.

I don't think its common, but on the other hand I haven't looked much.

If you think about it, people could just assign a light attack macro (as described in my last post) for each of their 5 abilities and use a multi=key mouse. Then they would have perfect weaving each time they used an ability.

If you’re logging it on your end then that means they’re getting no variance in their route to the server and then no server processing time variance (lol) and finally no variance on the servers route to you. Or if logs are server side then only the first two...

Regardless, your logs are meaningless.

Only client side logs from the person accused would be relevant

When the receiver of the attacks consistently sees a 0.002 second delay between them, it is safe to assume the sender actually sent them with 0.002 seconds delay between them.

While internet can mess up results, it generally increases randomness, not decreases it.

Honestly that’s a safer assumption than assuming the internet connection and the server processing had absolutely 0 variability. The latter is pretty much the only factor we can rule out outright... the others are possible albeit unlikely.

While internet is far from reliable, the variability of the unreliability is not so bad that it would cause three attacks done virtually at the same time to have wildly variable delays between them when they arrive.

What i mean: the internet can cause the whole batch of three attacks to arrive in 100 ms (if the weather is good) or in 500 ms(if the weather is bad), but it will not cause the delay between the first attack and the second to be wildly different from the delay between the second attack and the third.

For that to happen, the "weather" would have to change in the time it takes for the whole batch to be sent: you have good connection at T+0, then a bad connection at T+0.002s, and then a good connection again at T+0.004s. Internet is not that volatile.

For this reason, if you see incoming attacks (sent within a very short timeframe) to have consistently the same delay between them when they arrive, they most likely also had the same delay between them on the sender's side.

I disagree but Post logs

Theignson wrote: »

Sharee is correct. Nerd alert:

The statistical concept is that of "expected value". What is the expected value of the time difference between the light attack and the ability, if 1) a computer performs the attacks (macro) or 2) a human? With a macro, the expected value is.002 seconds. With a human, it is longer; you can test this on a dummy by just casting 20 light attack/ablity combos then looking at your log.In my case, I don't usually cancel the light attack "perfectly", so sometimes the gap is .2 seconds, sometimes even longer. Lets say it is a mean of 0.2 seconds difference.

That is a 100 fold difference in expected value (or, "expected mean") between the macro and the player! It should be even bigger in cyrodil where the opponent can move, roll etc.

Thogard seems to be arguing that because of systemic variance (eg lag) we cant detect a 100 fold difference in expected value. However, the variance is presumably the same between the experimental and control groups (as they would be described in science).Ie there is no reason to think that the lag when a macro is performing attacks differs from that when a human does.

Or, maybe some people would argue that a "good" player could match the expected value of a computer. No, not possible over time. Humans get tired, humans get distracted, and most importantly, humans physiologically can't sense time intervals to the thousandth of a second. If you think a human could match the timing of a computer, no point in discussing this further.

If Thogard was correct, then vast realms of science could not function. The null hypothesis is routinely rejected in science with differences in expected mean of only .2 fold (not 100 fold!) (ie biologists, physicists and chemists routinely detect statistically significant difference between means with much less than 2 fold differences).

It is all about detecting signal in the noise. This is routine stuff for modern science/statistics. Therefore it would be possible to screen for macro use on a statistical basis and detect cheaters with a p value of approximately .001 (ie a 1 in 1000 chance of wrongly suspecting a human of using a macro) which would improve with a larger sample size.

You do not have any idea what I am referring to. You are discussing statistics without understanding that my argument was that there was a sampling error. If you wish to refute my argument you must discuss why you believe there is no sampling error, rather than rambling about freshman statistic truisms.

To reiterate my point - the logs being referred to are client-side for the player being attacked.

If you believe that the person being attacked CONSISTENTLY receives those attacks with the exact same time difference as the time difference between the key presses of the attacker, then Please let me know what server you’re on.

Try bash weaving in vivec. Then do it in shor. Then do it in BGs.

There are so many variables in between the attackers button presses and when those skills register on the recipients combat logs that the arguments you people are proposing should be thrown out outright. Statistically irrelevant. That’s why I’m asking for the logs, so that I can see the sample size. I’m willing to bet it’s really really small.

I also want to point out that in no point in your “nerd alert” did you actually use your statistics correctly. Without the sample size of this macro user’s weaves we can’t even talk statistics. Might I suggest learning how to apply stat in practical purposes rather than listing definitions for concepts that have not yet been determined to be relevant.

Also you aren’t even looking at the right variable. Please just stop. Your average weave delay has literally nothing to do with the conversation and never would, even if we had a large sample size. For example, with your flawed logic we could just set up a macro with a 0.2s delay and then according to you that wouldn’t be a macro? You are out of your element.

For anyone that’s paying attention and actually understands statistics, allow me to recap:

The difference between a player and a macro is the variability / standard deviation in the delay between light attacks and abilities. A macro would have a consistent delay, but a player would have a variable delay. The fact that @Theignson referenced his own mean / expected delay rather than the standard deviation of his own results is why we can safely conclude that he isn’t qualified to weigh in on this issue.

the only things relevant here are 1. the standard deviation in the delay between light attack and the ability (which the poster claims is 0) and 2. The sample size

If we can confirm that the attacker had a consistent fixed delay between LA and ability (regardless of what that delay is... what we want to determine is a consistency in the delay AKA a std dev of 0) and a statistically significant sample size then we could confirm that the attacker is macroing.

My point previously was that due to the inherent variability in connection speed, using the logs from the victims side rather than the attackers side would be flawed ... in other words the variability of the attacks from connection speed issues could make a macro user look like a player. How to attribute it to one and not the other? The variable has not been isolated.

But if we see a significant sample size then we could still draw a conclusion. But I doubt we’ll see that because those logs are next to impossible due to variability from server lag.

So please post logs. I can admit when I’m wrong. This just isn’t one of these instances, and it certainly isn’t for any of the farcical reasons quoted.

Theignson wrote: »

Thogard wrote: »

Theignson wrote: »

Sharee is correct. Nerd alert:

The statistical concept is that of "expected value". What is the expected value of the time difference between the light attack and the ability, if 1) a computer performs the attacks (macro) or 2) a human? With a macro, the expected value is.002 seconds. With a human, it is longer; you can test this on a dummy by just casting 20 light attack/ablity combos then looking at your log.In my case, I don't usually cancel the light attack "perfectly", so sometimes the gap is .2 seconds, sometimes even longer. Lets say it is a mean of 0.2 seconds difference.

That is a 100 fold difference in expected value (or, "expected mean") between the macro and the player! It should be even bigger in cyrodil where the opponent can move, roll etc.

Thogard seems to be arguing that because of systemic variance (eg lag) we cant detect a 100 fold difference in expected value. However, the variance is presumably the same between the experimental and control groups (as they would be described in science).Ie there is no reason to think that the lag when a macro is performing attacks differs from that when a human does.

Or, maybe some people would argue that a "good" player could match the expected value of a computer. No, not possible over time. Humans get tired, humans get distracted, and most importantly, humans physiologically can't sense time intervals to the thousandth of a second. If you think a human could match the timing of a computer, no point in discussing this further.

If Thogard was correct, then vast realms of science could not function. The null hypothesis is routinely rejected in science with differences in expected mean of only .2 fold (not 100 fold!) (ie biologists, physicists and chemists routinely detect statistically significant difference between means with much less than 2 fold differences).

It is all about detecting signal in the noise. This is routine stuff for modern science/statistics. Therefore it would be possible to screen for macro use on a statistical basis and detect cheaters with a p value of approximately .001 (ie a 1 in 1000 chance of wrongly suspecting a human of using a macro) which would improve with a larger sample size.

You do not have any idea what I am referring to. You are discussing statistics without understanding that my argument was that there was a sampling error. If you wish to refute my argument you must discuss why you believe there is no sampling error, rather than rambling about freshman statistic truisms.

To reiterate my point - the logs being referred to are client-side for the player being attacked.

If you believe that the person being attacked CONSISTENTLY receives those attacks with the exact same time difference as the time difference between the key presses of the attacker, then Please let me know what server you’re on.

Try bash weaving in vivec. Then do it in shor. Then do it in BGs.

There are so many variables in between the attackers button presses and when those skills register on the recipients combat logs that the arguments you people are proposing should be thrown out outright. Statistically irrelevant. That’s why I’m asking for the logs, so that I can see the sample size. I’m willing to bet it’s really really small.

Lol. Do you have any actual experience or training in statistics or science? Don't you understand that the main use of statistics in science is to compensate for sampling error? Have you ever even analyzed any data using statistics?

Do you think that in the lab there is no sampling error, no source of noise, no variance between different people performing the experiments? How do you think scientists address this?
OF COURSE there is sampling error-- that is why we use statistics--

The client side log reflects error introduced BOTH when a macro is used and when a player is hitting the keys.

OF COURSE they don't "receive those attacks with the exact same time difference as the time difference between the key presses of the attacker". There is variance introduced by lag at all the levels that you mention. The key point: this variance is the SAME whether a macro is used or a player is hitting keys. Are you arguing that the lag is different, and therefore the variance on the client side logs is different, when a player hits the keys or when a macro is used? Of course not. Thus the error introduced in both cases is the same. If the variance is similar but the means are different, then we can detect a statistically significant difference between the means (of the time difference between the light attack and the ability) when a macro is used vs. when a player does it.

I don't know how to explain this to you any more clearly

Instead of assuming someone else is ignorant, re-read what I've written. Not everyone who is playing this game is uneducated, 22 years old, or even just beginning a professional career.

“OF COURSE they don't "receive those attacks with the exact same time difference as the time difference between the key presses of the attacker"-Theigson

The person who you’re saying is right has explicitly stated otherwise. You are agreeing with me, not them. My whole point was that this means their analysis was flawed.

“this variance is the SAME whether a macro is used or a player is hitting keys.”-Theigson

Umm no. That’s the only way of detecting a macro.

Maybe you know stats maybe not, but it doesn’t seem you know how macros work??

Your points are all over the place.

Fact: Macros can have the same mean delay as a player.

Fact: Players have higher std dev in delays than macros. These facts had already been established...

Your statistical analysis is wrong, and your understand of the game mechanics is wrong. And yes, I work with statistics literally every day. I’m not sure if your statistics are wrong due to poor understanding of how the game / macros work or of how statistics work. Regardless, I’m not going to reply to you anymore as I do not view you qualified to have an opinion on this issue for the reasons already established.

Thogard wrote: »

I’ve never seen any real hacking... not in at least two years.

[...]

I don’t think we’ll ever get rid of exploiters. But I’m glad that we don’t have to deal with hackers.

We have to deal with hackers. About three weeks ago, there was a rather prominent forum post (that ZOS removed very quickly) protesting against the increasing amount of cheaters in PVP and referring to a russian hacksite (rfcheats) selling the current hacks for ESO. That site was showcasing various videos how the hack works in-game. This was real-game material, not something made up on the PTS, as I knew several of the vicitim-players that were visible in the video - sometimes up to 20.

And just four days ago I saw (and videoed) two guys lifting themselves vertically off the ground in Bleakers and standing in air to spam healing springs for their ballgroup without being touched by any ground damage. They did exactly what was showcased the the 'promotional video'.

So yes, cheaters are a reality although only very few are so mega-stupid as the two I saw at Bleakers. Just tuning up your gaming speed or your regen a bit is as effective as to walk through walls or fly throug air (both possible with current hacks), but ppl will always be uncertain whether a specific player is cheating or just incredibly good.