Thread: Monster resistance rates in Book 1

I havn't read the entire thread but the noobs claiming you need a sample size of 10,000 to test a simple 10% probability please just shut up. If you sample a 100 or so and are sitting at 50% something is up. It may not be bugged but for some reason its not 10% in that case. The chances of hitting even 50/100 with 10% chances are just silly.

Originally Posted by SaintBass

There is some other wonkyness I noticed yesterday between Blinding Flash and Stuns... not sure if it is the case with all Mez's and Stuns, but I did notice that there were instances of a mez not taking if it was applied towards the end of a stun. Couldn't really pinpoint if it had to do with the shake off animation or not, but it is worth mentioning as somewhat related.

The status effect mobs get during the shake off animation right after a stun or mez wears off, the one with a yellow icon, functions as an immunity to all stuns and mezzes. I am not sure if that is intended, but it's really annoying either way. You need to wait to mez until after that effect goes away.

Gotcha... been a long time since I played my LM so I am not totally up to speed yet. I will try to snag the tool-tip of the icon to see if there is a description there.

Originally Posted by Telcharan

The status effect mobs get during the shake off animation right after a stun or mez wears off, the one with a yellow icon, functions as an immunity to all stuns and mezzes. I am not sure if that is intended, but it's really annoying either way. You need to wait to mez until after that effect goes away.

I've been burned by that shake-off animation many times. It'd be great to get some sort of confirmation from the devs whether or not it's WAI that the mobs are immune to re-mez for that ~1s period right after the mez expires, but I suppose it's a bit off-topic for this thread

Originally Posted by Telcharan

The status effect mobs get during the shake off animation right after a stun or mez wears off, the one with a yellow icon, functions as an immunity to all stuns and mezzes. I am not sure if that is intended, but it's really annoying either way. You need to wait to mez until after that effect goes away.

ah-ha! I guess that's what happened to me a couple of times when I was running through the level 15 class quest on my LM. good to know.

Originally Posted by xadoor

I havn't read the entire thread but the noobs claiming you need a sample size of 10,000 to test a simple 10% probability please just shut up. If you sample a 100 or so and are sitting at 50% something is up. It may not be bugged but for some reason its not 10% in that case. The chances of hitting even 50/100 with 10% chances are just silly.

There is nothing right about this post. Sample Size is incredibly important in quantifying things like this. The entire idea of using a percentile role (or variation thereof) is based on the Law of Large Numbers. In the big picture, the developers have already determined our collective success rate and if you want to show that the collective rate of failure is off, you need a LARGE sample size. Add to this that RNG's are notably streaky and it is entirely plausible to get a 50% failure rate with a sample size of 100. Case in point, in the days of SoA crafting, I had a 90% failure rate when using 1 shot recipes, but I only ever used 50 single use recipes. There were EPIC threads from SoA about crafting woes because the system was geared so that players really only ever saw the small sample size failures. Almost as notable were the people who would knock out crit after crit with single use recipes.

Another thing that may need to be considered, is that we don't know whether or not the 'random' seed number is global, local, or individual. If there's a global seed being used by everyone at any given moment, then I would think that would invalidate any attempt are determining true % rates.

Originally Posted by auximenes

Another thing that may need to be considered, is that we don't know whether or not the 'random' seed number is global, local, or individual. If there's a global seed being used by everyone at any given moment, then I would think that would invalidate any attempt are determining true % rates.

I would tend to think that the seed number is individual, otherwise, we would probably see many more instances of people missing or b/p/e'ing all at once. Multi-boxers in particular would have taken note of it since it would be peculiar if all your skills from the same class missed all at the same time.

Originally Posted by SaintBass

I would tend to think that the seed number is individual, otherwise, we would probably see many more instances of people missing or b/p/e'ing all at once. Multi-boxers in particular would have taken note of it since it would be peculiar if all your skills from the same class missed all at the same time.

Actually, I think the opposite is true: it's global. In fact, I thought we got a dev statement to that effect back when there was lots of outrage over crafting crit chances in early SoA. That's why we DO see frequent resists in a row (the very source of these threads). If you've ever played a Champion, you are quite used to seeing Blocked, Evaded, Parried, Missed, Missed all at one time.

Oh yay! Now we're back to disagreeing!

Originally Posted by gildhur

Oh yay! Now we're back to disagreeing!

Natural order has been restored.

Originally Posted by gildhur

Actually, I think the opposite is true: it's global. In fact, I thought we got a dev statement to that effect back when there was lots of outrage over crafting crit chances in early SoA. That's why we DO see frequent resists in a row (the very source of these threads).

Yeah, if the seed is global, it means that your previous 'roll' on whatever has absolutely no bearing on the probability of anything on your next 'roll'.

Originally Posted by gildhur

Actually, I think the opposite is true: it's global. In fact, I thought we got a dev statement to that effect back when there was lots of outrage over crafting crit chances in early SoA. That's why we DO see frequent resists in a row (the very source of these threads). If you've ever played a Champion, you are quite used to seeing Blocked, Evaded, Parried, Missed, Missed all at one time.

Oh yay! Now we're back to disagreeing!

Pfft, we are only disagreeing because you are wrong again. XD

I was more referring to seeing a bunch of B/P/E's or misses all at once from multiple simultaneous strikes from a variety of sources. As an example, you have the RNG just spitting out numbers that the various attack and defense rolls pull from, it would (from what I know) be impractical for the RNG to operate at top speed since PC attack rate is standardized and you just don't need that much processing speed devoted to the RNG... so you really only need to generate (spit-balling) 40 numbers per second. If there are attacks that execute at the same time (or have multiple outputs) I would expect to see a lot of the same values cropping up in groups when I view everybody's damage output. I also find it hard to believe that if I am playing my LM, and I have a mez resist, that the same number would be up in the time it takes me to hit Call to the Valar and mez again to see another resist. That seems more indicative of the streakiness of the RNG than pulling the same number twice in one second.

In terms of multi-output attacks, I have definitely seen Brutal Strikes get "Evade-Evade-Evade" and the like, but even that, I still see as streakiness rather than all three attacks pulling from the same number since I have also seen many different combinations of B/P/E miss and hit.

Disclaimer: Not saying that I am totally right, since I am making some assumptions, but they are pretty well reasoned assumptions based on what little I know about computing.

Originally Posted by SaintBass

I would tend to think that the seed number is individual, otherwise, we would probably see many more instances of people missing or b/p/e'ing all at once.

That doesn't add up. The random numbers are not "win" or "lose", they're large integers or floating point numbers that are then converted into wins or losses after applying a formula. When a RNG is bad it is not because you get a long streak of low numbers or high numbers (because any two bit pseudo RNG will avoid that), but because an underlying pattern is exposed. Ie, it looks very uniform until you plot it out, or until you look at the numbers modulo another number, or until you look at every Nth result, etc.

If the RNG is global, you will tend to see fewer patterns in the results, because the uses of the numbers will vary tremendously. Ie, the first number is used for crafting critical chance, the next number generated will be used to see of an orc evades and attack, the third number determines how much money is present on the goblin someone just killed, the fourth is used to see what angle a shrew will turn in its random walk, etc. The randomness of the use of the numbers would mask the lack of randomness in the generator. If it was global just for crafting, then you have to take into account all people crafting; ie, between your two crafting attempts there may have been 20 other people who also made crafting attempts. However if you had your own RNG used only for your own crafting and nothing else, then any flaws in the RNG would become more apparent.

It won't be global anyway. At the very least there will be one RNG per task/computer/server/zone/etc. That is, when the request for a result is needed it would be faster for whichever processor is handling that task to generate the number than to spend time consulting a global RNG processor.

Originally Posted by SaintBass

As an example, you have the RNG just spitting out numbers that the various attack and defense rolls pull from, it would (from what I know) be impractical for the RNG to operate at top speed since PC attack rate is standardized and you just don't need that much processing speed devoted to the RNG... so you really only need to generate (spit-balling) 40 numbers per second.

RNG algorithms can be made that are very strong (good distribution, low auto-correlation, etc.) that execute in only a small number of integer level instructions, so generating thousands or even millions, or pseudo-random values per second isn't usually going to require a lot of CPU overhead. More than likely the RNG library function is the same for each server machine (e.g., each zone within a server "world") as well, so given suitably distributed initial seeds (easily done), we shouldn't expect any artificial RNG correlations, etc. Of course, thats assuming that Turbine has even implemented the RNG in software

Keep in mind that in most cases, with a so-close-to-truly-random-as-makes-no-difference RNG, streakiness is a "human" mis-perception (observation bias) more than anything else. That isn't to say that there aren't correlating effects possible due to dynamic or short-lived changes to the probability tables from which effects are determined (e.g., a hidden if-you-resist-at-all-you-get-a-3-second-resist-buff).

As others have stated previously, seeing a string of resists or crits for something that should be low probability isn't in and of itself an indicator that something is out of whack (because you could be just "unlucky", or you've subconciously "cherry picked" a particularly noticeable/annoying sequence out of a normal sequence). However, if you keep seeing it, over and over, then that starts to become statistically significant.

Case in point, just for a lark, on my level 21 LM, I went and found a level 21 mob (a boar in north downs) and cast SoP:C (tactical resist) on it 100 times, which I easily counted via the trait I'm still working on, while keeping the animal mezzed with BF, and counted the resists I saw visually (so there's possibility for error right there). In my one-and-only-sample, I saw 6 resists for SoP:C, with 2 of them in a row at one point. Does this mean that the on-level tactical resist rate of boars at level 21 is 6% and it is streaky? Well, no, not in and of itself. What I need to do (and will do) is run the same test a number of times over a period of time (at least 10 would be a start, but I'm not commiting to do that many that right now since this is at level 21 ). That will give me a "sample set" that I can use to draw suitable obserations from. (And of course this will still only tell me about one skill used on one type of boar in one place).

It'd be ideal if others could attempt similar sample sets (multiple runs of testing, with as many variables kept the same as possible on each run) at various character levels (particularly above level 43) and locations so we can see what, if any, trends/patterns apply, and raise any issues to Turbine with some level of surety about what the nature of the problem is.

Originally Posted by Lohi

That doesn't add up. The random numbers are not "win" or "lose", they're large integers or floating point numbers that are then converted into wins or losses after applying a formula. When a RNG is bad it is not because you get a long streak of low numbers or high numbers (because any two bit pseudo RNG will avoid that), but because an underlying pattern is exposed. Ie, it looks very uniform until you plot it out, or until you look at the numbers modulo another number, or until you look at every Nth result, etc.

If the RNG is global, you will tend to see fewer patterns in the results, because the uses of the numbers will vary tremendously. Ie, the first number is used for crafting critical chance, the next number generated will be used to see of an orc evades and attack, the third number determines how much money is present on the goblin someone just killed, the fourth is used to see what angle a shrew will turn in its random walk, etc. The randomness of the use of the numbers would mask the lack of randomness in the generator. If it was global just for crafting, then you have to take into account all people crafting; ie, between your two crafting attempts there may have been 20 other people who also made crafting attempts. However if you had your own RNG used only for your own crafting and nothing else, then any flaws in the RNG would become more apparent.

It won't be global anyway. At the very least there will be one RNG per task/computer/server/zone/etc. That is, when the request for a result is needed it would be faster for whichever processor is handling that task to generate the number than to spend time consulting a global RNG processor.

Just for clarification: are you saying that the RNG is neither individual or global? Saint Bass was arguing in the favor of an individual RNG, Gildhur's stance is that RNG may be global. I'm only curious because you quoted Saint Bass, then somewhat agreed with him on the principle issue.

Originally Posted by cwswim03

But anyway, I'm sure you put most people who disagree with you on ignore.

That would be entirely correct, he does do that.

Originally Posted by SaintBass

blah, blah, blah

Grow up, and learn to deal with people who disagree with you.

Originally Posted by Lohi

That doesn't add up. The random numbers are not "win" or "lose", they're large integers or floating point numbers that are then converted into wins or losses after applying a formula. When a RNG is bad it is not because you get a long streak of low numbers or high numbers (because any two bit pseudo RNG will avoid that), but because an underlying pattern is exposed. Ie, it looks very uniform until you plot it out, or until you look at the numbers modulo another number, or until you look at every Nth result, etc.

If the RNG is global, you will tend to see fewer patterns in the results, because the uses of the numbers will vary tremendously. Ie, the first number is used for crafting critical chance, the next number generated will be used to see of an orc evades and attack, the third number determines how much money is present on the goblin someone just killed, the fourth is used to see what angle a shrew will turn in its random walk, etc. The randomness of the use of the numbers would mask the lack of randomness in the generator. If it was global just for crafting, then you have to take into account all people crafting; ie, between your two crafting attempts there may have been 20 other people who also made crafting attempts. However if you had your own RNG used only for your own crafting and nothing else, then any flaws in the RNG would become more apparent.

It won't be global anyway. At the very least there will be one RNG per task/computer/server/zone/etc. That is, when the request for a result is needed it would be faster for whichever processor is handling that task to generate the number than to spend time consulting a global RNG processor.

I know that the numbers themselves are not win or lose, however, there is a range of values and the result is then modified pretty consistently, so you can reasonably predict that there is a set of values within that range that is pretty consistently win while another set of values that is pretty consistently fail. The numbers themselves bear no significance until you attach modifiers and use them to resolve situations by comparing the resulting value to a static value or another RNG modified number.

Pretty sure we just said the exact same thing in different ways. My point is that in situations you are using the same number for multiple resolutions, you will see remarkably similar results among those resolutions, since the character modifiers that get plugged into the RNG are fairly similar across the board. If it were the case, I posit that we would have seen situations in groups and with multi-boxers where multiple actions that were resolved with the same RNG product would all have largely similar if not identical results. We would have been able to observe a second or so where every or nearly every attack in a raid would have whiffed.

Now, I may be totally off track in some of my assumptions here, I am not super familiar with the intricacies of RNGs and there are still elements of the combat system that we know nothing about, but from what I do know about action resolution systems that rely on a random element added to positive and negative modifiers. I only suggested that the random seed be generated individually and on a per action basis, because the more results that are used communally the more the community will be able to pick up on the pattern... especially since human thinking is prone to seeking out and identifying patterns of activity.

Originally Posted by Deyjarl

Are you trying to pick a fight?

Originally Posted by erikthe10man

Just for clarification: are you saying that the RNG is neither individual or global?

Yes, it's most likely one RNG seed per processor, rather than one per player, or one per "server".

Originally Posted by SaintBass

My point is that in situations you are using the same number for multiple resolutions, you will see remarkably similar results among those resolutions, since the character modifiers that get plugged into the RNG are fairly similar across the board. If it were the case, I posit that we would have seen situations in groups and with multi-boxers where multiple actions that were resolved with the same RNG product would all have largely similar if not identical results. We would have been able to observe a second or so where every or nearly every attack in a raid would have whiffed.

That's what doesn't make sense really. Players may share the same random number generator, but they will not share the same random number! Each and every action will generate a new random number. Each member of the raid will generate a new random number (global means each is generated sequentially from the same sequence, individual would mean each player has their own sequence). They're so incredibly fast to generate that there's no need to share the resulting number.

In other words, we probably have either of two metaphorical situations. First is every player rolling their own dice, and second is a game master rolling the dice for them when it's every player's turn. I don't think the GM would roll the dice and then let every player share that result.

Originally Posted by Lohi

Yes, it's most likely one RNG seed per processor, rather than one per player, or one per "server".


That's what doesn't make sense really. Players may share the same random number generator, but they will not share the same random number! Each and every action will generate a new random number. Each member of the raid will generate a new random number (global means each is generated sequentially from the same sequence, individual would mean each player has their own sequence). They're so incredibly fast to generate that there's no need to share the resulting number.

In other words, we probably have either of two metaphorical situations. First is every player rolling their own dice, and second is a game master rolling the dice for them when it's every player's turn. I don't think the GM would roll the dice and then let every player share that result.

OK cool, we are actually in "violent agreement." Only one RNG is running (per processor) but it never gives out the same result for multiple actions. I was protesting the implication that it did/could.

Originally Posted by auximenes

If there's a global seed being used by everyone at any given moment, then I would think that would invalidate any attempt are determining true % rates.

IMO it's inconceivable the RNG is server-based. There are just too many random events happening in the world every second to make client/server communication plausible for every one. I'm sure we'd see a LOT more in-game stutter if clients were waiting to talk to the server to generate a random outcome.

Originally Posted by Kraggy

IMO it's inconceivable the RNG is server-based. There are just too many random events happening in the world every second to make client/server communication plausible for every one. I'm sure we'd see a LOT more in-game stutter if clients were waiting to talk to the server to generate a random outcome.

Actually Kraggy, the servers pretty much do everything you see (calculate outcomes, move NPCs, etc.). The only thing the client on your PC does really is (i) send your movements/button presses to the server, and (ii) render the world according to how the server defines it. In every other respect the game actually "plays" on the server.

Doing it any other way (e.g., allowing the PC to calculate outcomes, etc.) opens the door for clients to "spoof" the results (e.g., tell the server you rolled a perfect 100 every time) and would make it far more difficult to keep other players in sync (e.g., if you have 12 people in a raid, who would be calculating what, and how would you tell everyone else?).

Lag effects due to network latencies become quite noticeable if you play from far away (like over here in Australia, at 250msec+ latency to the US based servers ).

Last night my 65 warden was helping my hubby's lower level RK finish the Riddle quest, in the cave with the crawlers. I ambushed and stunned the main crawler in the back. All the other crawlers appeared and charge forward, so I threw my nifty multiple javelin legendary skill.
EVERY SINGLE crawler resisted it! Those crawlers were waaay below my warden's level, yet every one of them resisted her attack at the same time.

On the positive side the elite drakes in Giant Valley seem to be fixed. A bit after Mirkwood, I was testing to see if my then 62 warden could bring my hubby's level 32 warden in there and do the woodworker quest. Fought one of the drakes solo. Got it down to 3/4th morale and it had already gotten my warden, in maxed out LI traited determination, down to less than 1/4th her morale. I realized that she would not be able to handle tanking two of them and keep him alive if a third mob aggroed on him. This time, with the same gear and just 3 levels higher, same stats as those didn't go up, she took a drake down easily and fast. So whatever was messed up about those drakes did get fixed.

Originally Posted by redwoodtreesprite

So whatever was messed up about those drakes did get fixed.

Maybe because they're sigs, not elites.

Originally Posted by xadoor

I havn't read the entire thread but the noobs claiming you need a sample size of 10,000 to test a simple 10% probability please just shut up. If you sample a 100 or so and are sitting at 50% something is up. It may not be bugged but for some reason its not 10% in that case. The chances of hitting even 50/100 with 10% chances are just silly.

You'll understand it a bit better when you get to your first lesson on statistics in high school.

Originally Posted by DougHillman

You'll understand it a bit better when you get to your first lesson on statistics in high school.

Actually he is right (speaking from a vantage of 6 years of college level math). If you are depending on a RNG, and the RNG returns results that are significantly skewed after a sample size of 100, it is a bad RNG. No data set should ever require 10,000 samples to prove a trend.

Originally Posted by DougHillman

Originally Posted by xadoor
I havn't read the entire thread but the noobs claiming you need a sample size of 10,000 to test a simple 10% probability please just shut up. If you sample a 100 or so and are sitting at 50% something is up. It may not be bugged but for some reason its not 10% in that case. The chances of hitting even 50/100 with 10% chances are just silly.
You'll understand it a bit better when you get to your first lesson on statistics in high school.You'll understand it a bit better when you get to your first lesson on statistics in high school.

/signed. if all we needed was 100, then the devs would have a different story to tell by now. modified RNG are funny things...

Originally Posted by Barruktp

Actually he is right (speaking from a vantage of 6 years of college level math). If you are depending on a RNG, and the RNG returns results that are significantly skewed after a sample size of 100, it is a bad RNG. No data set should ever require 10,000 samples to prove a trend.

Just because the numbers are skewed in a small sample set does not dictate that the RNG is broken. If the 10% chance is fixed absolutely, it is no a "random number generator". A fixed 10% would mean after "X" amount of failures there would be a guaranteed amount of successful "hits" until the 10% number is achieved, which in itself would be ludicrous.

Just because something is improbable, or less than likely, does not mean it's "impossible". It's like the winning the lottery/struck by lightening theory - it may be hard to achieve getting either, but it happens all the time. A larger sample size would be a better indicator of a trend than such a small, isolated incident (or small groups of incidents).

Originally Posted by Barruktp

Actually he is right (speaking from a vantage of 6 years of college level math). If you are depending on a RNG, and the RNG returns results that are significantly skewed after a sample size of 100, it is a bad RNG. No data set should ever require 10,000 samples to prove a trend.

I hope you didn't pay much for those 6 years.


That being said, I should have gone on and said that I'd think that seeing something that skewed in a sample size of 100 might be an indicator that it should be looked at with a larger sample size. But it no way proves that there's an issue.

Originally Posted by DougHillman

I hope you didn't pay much for those 6 years.

If I could come up with zingy quips like that on the fly, I would be a politician, not an engineeer.

However I think you and I agree in principle on statistics. My words were "significantly skewed". The term "significance" has special meaning in statistics as relates to confidence values. My point being that if you have a sample data set of 100 points, and your calculated confidence value is low, then the set itself contains bad data. This can be proven mathematically.

Now if you want to confirm a 10% probability to 0.01% accuracy, yes you would need 10,000 or more samples. But that is not significant for our purposes.

Originally Posted by Barruktp

Actually he is right (speaking from a vantage of 6 years of college level math). If you are depending on a RNG, and the RNG returns results that are significantly skewed after a sample size of 100, it is a bad RNG. No data set should ever require 10,000 samples to prove a trend.

Umm, I think you might be confusing statistical behaviour here (it is possible for a truly random system to have bursts which significantly skew any "single" observation of that system) with desired behaviour (users shouldn't ever see unreasonably/significantly skewed results in any single sample).

As myself and others have posted, a single sample (of 100 events) is just too small to draw any significant conclusions from statisically for a nominal 10% occurrence rate - you need multiple samples to be able to do that. E.g., as I previously posted, you'd probably want at least 10 samples (of 100 events each) to be able to subject it to any sort of statistical analysis.

Case in point - as per my previous post/suggestion, I repeated my 100 casts of SoP:C on an on-level (level 21) mob with my level 21 LM:
- first time, 100 casts saw 6 resists, 2 of them in a row.
- second time, same class/level of mob, 100 casts saw 1 resist.

If I'd stopped at one single sample (first run of 100), I'd have said "wow, 6% resist rate at level 21, no way is that WAI, quick, to the forums!". But as you can see, from two samples (which is still not enough) the rate looks closer to 3.5%, and it tells me that there is significant variation between two samples of length 100. What will I see when (if ) I get to 10 samples?

Now, intuitively, if I'd seen 50 resists in that first 100 sample, I'd also want to say "OMG, no way is that only 10%". But even then, statistically, I could just have been the most unlucky SOB on the server/planet, and my correct response should be "OMG, you won't believe how unlucky I was today". In other words, I'd have to see my unfortunate experience repeated multiple times to be able to definitely say there was something wrong.

From a game-play view point however, its arguable that no single player should ever experience an unreasonable worst-case, irrespective of its true probability, which actually then suggests that a deliberately non-random RNG mechanism is required. (Which I'd hate to have to design on the scale of a MMO, ouch).

(PS:I type too slow, lol)

The important thing here is really not whether or not we can come up with a truely significant set of data. It's whether or not the results we've seen are enough to raise doubt. If Turbine feels that there have been enough unreasonable results, then they will check their code (and probably are). They can more easily generate a significant data set and verify their RNG. No single player should have to come up with that kind of data.

Of course, no player should say the system is definitely broken unless they do capture a large data set. All they can really say is that it doesn't seem right.

Originally Posted by Gimlet008

From a game-play view point however, its arguable that no single player should ever experience an unreasonable worst-case, irrespective of its true probability, which actually then suggests that a deliberately non-random RNG mechanism is required. (Which I'd hate to have to design on the scale of a MMO, ouch).

I've made essentially this same point in the crafting forums before. I could care less if the RNG is working spot on. The fact that it can be incredibly frustrating for some players is BAD. I have been, over the course of the game since before Day 1, ridiculously lucky with regards to crits on single use recipes. I kept track for a while and my pecentage was over 75% back before the day of Crafting Books that could get your one-shot attempt anywhere NEAR that. Hooray for me. Unfortunately, some other poor sap never critted on single-use recipes.

I understand the mechanics just fine. I know that just because my number says 50%, I'm not guaranteed to crit 2 if I make 4. I donno that I have any suggestion for a better mechanic, but a straight RNG that can leave some people as princes and some as paupers is ******.

Originally Posted by DougHillman

I've made essentially this same point in the crafting forums before. I could care less if the RNG is working spot on. The fact that it can be [b]incredibly/b] frustrating for some players is BAD. I have been, over the course of the game since before Day 1, ridiculously lucky with regards to crits on single use recipes. I kept track for a while and my pecentage was over 75% back before the day of Crafting Books that could get your one-shot attempt anywhere NEAR that. Hooray for me. Unfortunately, some other poor sap never critted on single-use recipes.

/raisehand

That'd be me. My lifetime crit rate on one-off recipes at a 33%-35% chance is, at a guesstimate, well under 15%. At one point in SoA I failed 19 jeweler one-offs in a row. The chances of that are astronomically low. That's all well and good if I'm the poor sap who's just that unlucky. However, when a former kinmate had a streak of 22 or 23 fails in a row on tailor one-offs at 35%, I started to wonder. Is it possible for two people that phenomenally unlucky to not only be on the same server but in the same kinship? Yeah, I suppose. It's possible for me to be dealt 5 straight royal flushes in poker, too.

Originally Posted by Solien

/raisehand

That'd be me. My lifetime crit rate on one-off recipes at a 33%-35% chance is, at a guesstimate, well under 15%. At one point in SoA I failed 19 jeweler one-offs in a row. The chances of that are astronomically low. That's all well and good if I'm the poor sap who's just that unlucky. However, when a former kinmate had a streak of 22 or 23 fails in a row on tailor one-offs at 35%, I started to wonder. Is it possible for two people that phenomenally unlucky to not only be on the same server but in the same kinship? Yeah, I suppose. It's possible for me to be dealt 5 straight royal flushes in poker, too.

Turbine hates Hall of Fire even more than Hunters.

Originally Posted by gildhur

Turbine hates Hall of Fire even more than Hunters.

Most of the Hunters that Turbine hates have long since ragequit... now they just view Hunters with apprehension, but no bile driven gag reflex.

Originally Posted by Ayrolen

Maybe because they're sigs, not elites.

What's in a name? The drake had been doing way too much damage on my warden before, and she way too little. I was trying to say something positive, OK? Some things were fixed at the same time that resists seemed to get broke.

Originally Posted by redwoodtreesprite

What's in a name? The drake had been doing way too much damage on my warden before, and she way too little. I was trying to say something positive, OK? Some things were fixed at the same time that resists seemed to get broke.

I think their point was that reducing the mob to a sig would have the effect you noticed. They were trying to point out what changed in support of what you said.

Originally Posted by musicman2000

I think their point was that reducing the mob to a sig would have the effect you noticed. They were trying to point out what changed in support of what you said.

Guess I misunderstood as there was the wink. It really was frustrating for me, having so much trouble with a mob more than 20 levels below me, and one that was standard for that area as well. As for those wood trolls, shudder. Who ever heard of cowardly wood trolls that run away when they take some damage. Only purpose is to pull a swarm of nearby nasty trolls. Absurd, and a perfect example why that is such a nasty valley. Just there to be a griefer area for cooks and woodworkers. Never have done the book quest in there. Must be a dreadful quest to finish.