The True Rates of Refining

Fissile - Archosaur

I will quote my previous post on this subject here followed by my proposal on accurately characterizing refinement within the game. Please relegate unrelated discussion to the many other threads on refinement.

Fissile - Archosaur wrote: »

No one has asserted any credible argument regarding the true refine rates in this thread or any I have seen on the topic. The fact of the matter is the true mechanics of refining are hidden from the user. The refining guides posted here and in the wiki are essentially rough estimates provided with no factual basis.

Given a reasonable amount of data, one could conceivable perform a rigorous statistical analysis of refining and draw real conclusions based on observations.

If all these people intent on spending thousands of celestones refining a single item had the wherewithal to record a video of the attempts, maybe we could support some of the arguments people have been making. However, in the absence of supporting evidence, all arguments are equally invalid. And no, a claim does not constitute evidence without an actual record of what exactly happened.

Since the topic seems to be of great interest to many, I am going to start a thread with the purpose of statistically describing the actual refining process once and for all.

What I suggest is simple. When you get your brand new super-awesome but as-yet unrefined gear, before you head to your elder or preferred refining location and start spamming the "Refine" button, turn on a screen capture tool and record the entire refinement process from beginning until you decide to stop. Post the video somewhere accessible to the public and link it here.

I will turn the video into something that can be used for statistical analysis (I haven't decided yet how exactly I will do that). Depending on the number of submissions that get posted here, I may ask for help with that. I will apply my (admittedly limited, but above average) knowledge of statistics and the popular statistical analysis program R and see what we can conclude about refinement that is actually supported by evidence.

PotatoHeadQR - Dreamweaver

+1

/10chars

Miugre - Heavens Tear

...as I said to you in the other thread, this is a nice experiment but it ignores basic statistical principles, not to mention the existing evidence posted in the other thread.

1. A very small portion of the players visit this forum, a small percentage of those will read this thread and bother replying, an even smaller percentage of those will know how to make a video and bother uploading it. And even of those who match that extremely restrictive criteria, a very large percentage of them will only go to the trouble because they're motivated to QQ about their horrible luck in refining. You've essentially constructed an experiment where all semblance of a reasonable sample size and lack of bias are already forfeit.

2. SylenThunder named the file where the refining rates are handled, and mentioned that he'd already compared today's version against the one from beta. I've actually just sent him a PM asking to post the hard data from that file. Would that be enough evidence for you? -_-

The__Sun - Dreamweaver

only problem is the most info you would get is probably up to +7 and maybe some unreliable data at +8 cause not many go that far with mirages/tisha/tienkang

Tsyn - Raging Tide

I would love to see this. Imagine.
A group of players completely taking down the false information provided by PWE.

That is, unless a GM closes this down before we succeed, or maybe, just MAYBE, our CM actually tries to get us some info. XD

Fissile - Archosaur

Miugre - Heavens Tear wrote: »

...as I said to you in the other thread, this is a nice experiment but it ignores basic statistical principles, not to mention the existing evidence posted in the other thread.

1. A very small portion of the players visit this forum, a small percentage of those will read this thread and bother replying, an even smaller percentage of those will know how to make a video and bother uploading it. And even of those who match that extremely restrictive criteria, a very large percentage of them will only go to the trouble because they're motivated to QQ about their horrible luck in refining. You've essentially constructed an experiment where all semblance of a reasonable sample size and lack of bias are already forfeit. There is no shortage of players asserting that they spent thousands or celestones trying to refine a single item with no luck it would only take a very small number of such sessions to establish statistically significant results.

2. SylenThunder named the file where the refining rates are handled, and mentioned that he'd already compared today's version against the one from beta. I've actually just sent him a PM asking to post the hard data from that file. Would that be enough evidence for you? -_- If this is correct then you should be vindicated by the results of the thread and you can post later and say "haha I told you so." Others have asserted certain factors (like where you refine) could have an impact on refining rates. Unless you are privy to the actual source code you don't know any better than I do how those values are actually used.

If you don't agree with the point of the thread then please don't argue with me here. You can PM me or talk about it in the other thread. Whether or not this reveals any new information, I still consider it a worthwhile exercise.

The__Sun - Dreamweaver wrote: »

only problem is the most info you would get is probably up to +7 and maybe some unreliable data at +8 cause not many go that far with mirages/tisha/tienkang

There is information to be gleaned from every refine attempt, with or without refining aids. Without any samples at all, there is nothing to be learned.

HrunsPanda - Archosaur

I totally agree with miugre. And i therefore i doubt it is worthwhile because getting biased results is only detrimental. It is not even a case of "why not give it a try", you run the risk of producing false information that will readilly be absorbed by the QQing crowds and before we know, its accepted as fact.

A more unbiased result would come from asking those who offer refining services for their results. Assuming they have enough customers, they can provide a reasonable sample size by themselves and should be above the good luck bad luck bias since that is why they do their bussiness. Only issue is they might not want to share their results out of bussiness interests, espescially if the results show higher succes chances than what is now the common knowledge.

Shardik - Lost City

Guys, Hey you! just answered this

in this thread last post.


TT_TT

Descendus - Lost City

v4liance wrote: »

Heeeeyyyy guys!

It's random.

Closing the thread!

- Val

(^ From the last Refinement thread)

Hmmm, Something smells fishy~ Oh wait...

XScootaloo - Dreamweaver

of course its random. with it being random players are practically forced to spend money on dragon orbs to get that R9 third cast gear to +12. if there was a set formula to it players would figure out how to refine using as few orbs as possible and that wouldn't be making as much money as it would if it was just random.

why would you think there is a set formula? f:ponder

Xdmghyper - Lost City

It maybe random but it has been nerfed, spend 100mil on refine aids and mirages in the last 2 days and item remains +0 - +2 ... sad lol b:surrender

HrunsPanda - Archosaur

Xdmghyper - Lost City wrote: »

It maybe random but it has been nerfed, spend 100mil on refine aids and mirages in the last 2 days and item remains +0 - +2 ... sad lol b:surrender

what a completely meaningless statement. Did you try to get to +10 with mirages or are you trying to make us believe you didnt manage to reach +3 ?

soundslegit

I'm still waiting for the PWI Loyalists to explain why client side data files would describe server side operations. Explain why PWI could not have changed refining chances server side while leaving your precious tasks.data file on the client unchanged.

Xdmghyper - Lost City

HrunsPanda - Archosaur wrote: »

what a completely meaningless statement. Did you try to get to +10 with mirages or are you trying to make us believe you didnt manage to reach +3 ?

just trying to get 1 +7

Asterelle - Sanctuary_1381265973

soundslegit wrote: »

I'm still waiting for the PWI Loyalists to explain why client side data files would describe server side operations. Explain why PWI could not have changed refining chances server side while leaving your precious tasks.data file on the client unchanged.

Because then the client and server files will be out of sync.
Part of the logging in process is having the client compare the version of its copy of those files to the version on the server. If those version numbers are different you get a message saying 'client version is too low', which prompts you to run the patcher which syncs up the files.

As to why the client would need a copy of those files to begin with... it's too minimize the traffic between the client and the server. Consider the act of accepting a quest from an NPC. If all that data was server side and nothing at the client then to just look at a quest would require the server to send over the appropriate text, quest requirements, and other info whenever someone tries to do anything. What really happens is a very simple message between the client and the server that contains nothing but a 'AcceptQuest' action with the quest id. That message is 22 bytes in length.

By contrast this message that I am typing right now is 1312 bytes in length.

SylenThunder - Twilight Temple

Asterelle - Sanctuary wrote: »

Because then the client and server files will be out of sync.
Part of the logging in process is having the client compare the version of its copy of those files to the version on the server. If those version numbers are different you get a message saying 'client version is too low', which prompts you to run the patcher which syncs up the files.

As to why the client would need a copy of those files to begin with... it's too minimize the traffic between the client and the server. Consider the act of accepting a quest from an NPC. If all that data was server side and nothing at the client then to just look at a quest would require the server to send over the appropriate text, quest requirements, and other info whenever someone tries to do anything. What really happens is a very simple message between the client and the server that contains nothing but a 'AcceptQuest' action with the quest id. That message is 22 bytes in length.

By contrast this message that I am typing right now is 1312 bytes in length.

To even further expand on this, take the Boutique for an example.
The current main item boutique is about 2.6MB in size.
Imagine if you had to wait for that to download to your PC every time you opened it.

Now think about this....
Tasks.data, (where the data for our current subject of conversation sits), is 69.2MB in size.
Now imagine what would happen if your client had to reference data in that file from the server every time you did something.

When you load the client, it checks the dates, internal timestamps, and MD5 checksums of each of the files in the data folder. If they don't match exactly with what's on the server, you'll get an error that you aren't running the right version.

I didn't get a chance to compare the old tasks file with the current one last night, but I'll work on it when I can. I recall refining rates for each item used to be on PWDatabase, but I was unable to locate the data when I looked.

Tide_Surfer - Archosaur

Let Fissile do this if she wishes, no one has the right to stop her or to create an argument about it with her. I find it to be quite an interesting experiment, to see the very rough refining chances in a population of players.

Miugre - Heavens Tear

Fissile - Archosaur wrote:

There is no shortage of players asserting that they spent thousands or celestones trying to refine a single item with no luck it would only take a very small number of such sessions to establish statistically significant results.

The first part of this is exactly my point! You can't claim neutrality if your entire sample size consists of opt-in people motivated by their own QQing. Hell, who's to say their video(s) won't just show their terrible refine attempts and omit the good ones? Are you asking me to trust refining QQers to be objective about refining?

Fissile - Archosaur wrote:

If this is correct then you should be vindicated by the results of the thread and you can post later and say "haha I told you so." Others have asserted certain factors (like where you refine) could have an impact on refining rates. Unless you are privy to the actual source code you don't know any better than I do how those values are actually used.

Maybe not, but I do have an open mind and am willing to listen to those who do have access to the relevant files. Two such people have already posted in this thread, and you have yet to refute either of their arguments. At this point the burden of proof is on you, and sadly, you can't provide unbiased proof with a smattering of QQ videos.

Tide_Surfer - Archosaur wrote: »

Let Fissile do this if she wishes, no one has the right to stop her or to create an argument about it with her. I find it to be quite an interesting experiment, to see the very rough refining chances in a population of players.

The problem is how she's presenting it. She's going into this with an inherent bias towards believing that PWE screwed us on the refining rates. Hence the whole "experiment" is just an ignorant attempt to create misinformation - whether that's her stated intent or not, that's going to be the result. The very title of the thread betrays her neutrality, implying that PWE lied to us about the refine rates.

There are conclusive arguments and proof in this thread by multiple people as well as the last one that the refining rates have not changed; therefore no experiment needs to be made and there's no point in having one. You'd accomplish about as much by trying to disprove the theory of gravity.



I'm sorry to be sounding so harsh. I just have a huge problem with the spreading of misinformation (even if the intent was noble). If Fissile wants to challenge the conventional belief about refining rates, s/he can start by refuting the existing evidence which supports that belief. Until s/he can do that, this thread just looks like a QQer requesting backup. :-/

PotatoHeadQR - Dreamweaver

while the experiment relies upon the good will of the community (send all vids, with good or bad results) it is not so heavily influenced; you cannot submit a vid of "that one time when i lost 1000 mirages trying for +1". Also, even if there is an evil conspirancy, we can get pretty solid results (and calculate how certain we can be) for +1+2+3 given a large amount of data, unless we get a ton of vids that fail to refine to +1 and quit.

correct me if I'm wrong, but, even if we accept that the rate didn't change based on the fact that the file did not change, what proof do we have that the initial values are correct? If you look the history of the wiki page you get two versions, both of them stating research/players as source:

2009 (Tiaque - Sanctuary)
+1 95%
+2 50%
+3 25%
+4 10%
+5 5%
+6 1%
+7 0.5%
+8 0.25%
+9 0.1%
+10 0.05%
+11 0.025%
+12 0.01%

2010 (Updated the actual refining chances using latest researched results)
+1 50%
+2 30%
+3 30%
+4 30%
+5 30%
+6 30%
+7 30%
+8 25%
+9 20%
+10 15%
+11 10%
+12 5%

So where do those rates come from anyway? And if they come from source files, can we check again?

Last, about misinformation... yeah, a positive result will "empower" them, perhaps erroneously. yet, I believe it will be but a tiny drop in the rainstorm of qq reasons

soundslegit

Aster and sylent have not answered my question. I already understood how hash checking works and that the server verifies the validity of the client being used. Neither of these issues prove that the server's version of this data is identical to what is found on the client side.

Jargony exposition of well known facts are not an argument. How do any of you really know that the probabilities used on the server side are the same as found in the client side data files? An actual answer would help.

If any of you think that you can fake your way through this your wrong. A fact dump is not an argument. Please make an argument as to how we know the probabilities used on the server side are those found in the client side data files.

Apostasy - Raging Tide

I find the most amusing thing about the word "random" being applied to anything manmade, while everything including computer programs (being even the programs we design are finite) must be finite (look up kurt godel's incompleteness theorem)... thus, even the "random" generating programs can not be entirely random.

Even in quantum mechanics, the randomness in nature still has very definite probabilities for phenomena to occur, which can determine end results of entire sets or systems.

Even the lottery, if it were to use the exact same balls, etc... even it and all of its parts must be flawed in some way whereas patterns most definitely emerge over a great enough data set.

Even the r8r wrists and oht neck -6 channeling x3 example... follows along the lines of old programming from many other games. How many games have you played that a "rare" was no where to be found for a long period of playtime? Then you obtained a single "rare", and behold! What is this? Another "rare", if not the exact same "rare"?

To support that this version of "randomness" is indeed found in pwi as well... how many times have you seen 2 of the exact same mold during 2x drops from a single boss kill? Thus, whereas it may likely be "rare' to obtain one's first -6 channeling on the presented gear pieces, the second one likely becomes easier to obtain once achieving the first. I would think this more so applies toward the r8r gears than the oht however, due to the fact you do not reroll the oht gears the same way. Unless the program can be "persuaded" to make a 2x -chan piece by holding onto the previous 1x. And then, further persuaded to make that 3x, by holding onto the 2x while crafting gears.

Let Fissile perform her experiment, and try not to say you use logic if you have no actual evidence on which to found "your logic". Logic says... let the data speak for itself...

Yin - Momaganon

If people would post all refines it would do no harm but mostlikely bad attempts would be posted more oftehn and therefore the result would show much worse than it really is,

Mayfly - Dreamweaver

While indeed the "random" number generator is deterministic, as I understand it there's just one seed that's shared by all player processes on the server, so player action introduces a non-deterministic element into your results.

Miugre - Heavens Tear

soundslegit wrote: »

Aster and sylent have not answered my question. I already understood how hash checking works and that the server verifies the validity of the client being used. Neither of these issues prove that the server's version of this data is identical to what is found on the client side.

Jargony exposition of well known facts are not an argument. How do any of you really know that the probabilities used on the server side are the same as found in the client side data files? An actual answer would help.

If any of you think that you can fake your way through this your wrong. A fact dump is not an argument. Please make an argument as to how we know the probabilities used on the server side are those found in the client side data files.

What Ast said is that it checks tasks.data (and other files) against the server's versions before even allowing you to login. Therefore whatever you find on your client machine is identical to to server's copy at any given time (unless you're behind on your client updates). What part of that was hard to understand? o.O

PotatoHeadQR - Dreamweaver wrote:

2009 (Tiaque - Sanctuary)
+1 95%
+2 50%
+3 25%
+4 10%
+5 5%
+6 1%
+7 0.5%
+8 0.25%
+9 0.1%
+10 0.05%
+11 0.025%
+12 0.01%

If anyone thought this was the real refine rate at any point in the game, they've never refined anything at all. Posting this just gives QQers more (inaccurate) ammo.

And hey, I think a refine system like this would be great. Maybe if you did this and then removed Dragon Orbs entirely, we'd have no one above +5 and PVP would be way more balanced. But as long as we're not working in our own personal dream-lands here, we have to ask ourselves if we've ever refined 20 items in a row to +1 and 19 of them came out successful.

PotatoHeadQR - Dreamweaver

Mayfly - Dreamweaver wrote: »

While indeed the "random" number generator is deterministic, as I understand it there's just one seed that's shared by all player processes on the server, so player action introduces a non-deterministic element into your results.

in theory they could be getting 'truly' random numbers from real-world collectors too

Miugre - Heavens Tear wrote: »

What Ast said is that it checks tasks.data (and other files) against the server's versions before even allowing you to login. Therefore whatever you find on your client machine is identical to to server's copy at any given time (unless you're behind on your client updates). What part of that was hard to understand? o.O

If the final desision for refining is made server-side the tasks.data consistency doesn't really make any sense. Sure, it is the same with a file in the server but who guarantees that that file is the one used by the server for the refining?

Miugre - Heavens Tear wrote: »

Posting this just gives QQers more (inaccurate) ammo.

all I'm saying is that, while it appears that the data used by warren are pretty close to reality for me, unless I missed something, we don't know where they came from. and I don't really care about the ammo of qqers as long as they cant use it in NW.

Apostasy - Raging Tide

Is it possible, that even with "tasks.data" on our end and the server end matching, that the server end can simply be linked to numbers / probabilities which are "tweaked" behind the scenes? I would think programming can be worked to this end.

Regardless... my long term "luck" up to the summer of 2012, to just after that summer, changed dramatically. Whereas +4 used to be fairly decent to achieve via mirages alone... that just has not been the case since. I'm interested in seeing actual data though.

Miugre - Heavens Tear

PotatoHeadQR - Dreamweaver wrote: »

If the final desision for refining is made server-side the tasks.data consistency doesn't really make any sense. Sure, it is the same with a file in the server but who guarantees that that file is the one used by the server for the refining?

Apostasy - Raging Tide wrote:

Is it possible, that even with "tasks.data" on our end and the server end matching, that the server end can simply be linked to numbers / probabilities which are "tweaked" behind the scenes? I would think programming can be worked to this end.

In general I would defer to Ast and Sylen at this point (particularly Sylen) since they'd have better insight into the code itself. However, basic programming principles would strongly suggest against this.

The actual comparison would still need to be done in tasks.data, because if it were done in another file server-side you'd probably find a reference to that file in tasks.data. And we know that the server checks for parity between your version of tasks.data and its own version. So given all that, the only way they could pull one over on us is if tasks.data were modified on the fly, server-side, after you login to reflect the "true" refining rates. Not only would this be kludgy and the devs would have to go out of their way to do it... but I'm pretty sure the server would be working on just one copy of tasks.data, meaning they couldn't modify it on the fly or else the next user to login would be getting parity errors.

I mean, you guys can play the whole "OMG COVERUP!!1" angle if you want - it's your credibility - but this one can just be disproven by basic logic. :-/

TheDan - Sanctuary

I actually have a pre-saved video of me +8ing some gear a month ago in 3x speed (13 minutes) that I was going to make a commentary on but then got lazy so it just sits on my hard drive now, but if you wanted it slowed down enough to be able to actually count everything you'd have a 40 minute video and I doubt anyone would be interested in watching that tbh.

There's times it takes me 1000 mirages to +5 something, other times it takes me only 50 mirage to +9 something, the true rates might not even be reliable seeing how people like to exaggerate or use their outliers as fact. For each horrible refining experience I've had, I've also had about as many in epic refining experiences, which seems normal to me. While I don't think the currently accepted rates are 100% precise, I think they're accurate enough to be as guidance of the true rates on average.

bitewarrior

I honestly don't know anything about wheter it is true or not, but one fact remain true for me. When orb are on sale my refining chances is way worst , coincidence...maybe. But at every sale i get this coincidence...doubtly.

In the end this is a major revenue for them now since most all players on server as swicthed to new gear such as 3rd cast nirvana and R9 3rd cast that again coincidently cost 5 mirage to refine now........let's just say it is a big influence on player to be looking at orb for refining.

on a side note. Reminds me of SFC they really took there time to fix while they made multiple R9 sale and hyper sale during that period. End though is, anything financially driven and the dynamic behind it is the only true fact.

So, make your own conclusion based on fact and random is as random as they want it to be.

Regards

SylenThunder - Twilight Temple

Okay, I did a LOT of digging in tasks.data, element.data, and poked around a bit in general....

It would appear that the rate to refine items is hard coded into the server files. And when I say hard coded, I mean this isn't something that you can just arbitrarily poke at and change. Also, given my experience with the developers, and their current skillset, I sincerely doubt that they would be able to go in and make any changes to this.

Also made an interesting discovery. As early as server version 1.3.6, there were already a large majority of the quests and items built in the game for the Tideborn race, skills and other things. Didn't find much at that point for EarthGuards, but there are a lot of untranslated quest slots.

On a side note, It would appear that Tienkangs increase the refine chance by 0.15, Tisha by 0.035.
Also an interesting table on refining with Chienkun stones..
Refining Level 0: 100%
Refining Level 1: 25%
Refining Level 2: 10%
Refining Level 3: 4%
Refining Level 4: 1.67%
Refining Level 5: 0.77%
Refining Level 6: 0.47%
Refining Level 7: 0.25%
Refining Level 8: 0.13%
Refining Level 9: 0.07%
Refining Level 10: 0.04%
Refining Level 11: 0.02%

TheDan - Sanctuary

SylenThunder - Sanctuary wrote: »

Okay, I did a LOT of digging in tasks.data, element.data, and poked around a bit in general....

It would appear that the rate to refine items is hard coded into the server files. And when I say hard coded, I mean this isn't something that you can just arbitrarily poke at and change. Also, given my experience with the developers, and their current skillset, I sincerely doubt that they would be able to go in and make any changes to this.

Also made an interesting discovery. As early as server version 1.3.6, there were already a large majority of the quests and items built in the game for the Tideborn race, skills and other things. Didn't find much at that point for EarthGuards, but there are a lot of untranslated quest slots.

On a side note, It would appear that Tienkangs increase the refine chance by 0.15, Tisha by 0.035.
Also an interesting table on refining with Chienkun stones..
Refining Level 0: 100%
Refining Level 1: 25%
Refining Level 2: 10%
Refining Level 3: 4%
Refining Level 4: 1.67%
Refining Level 5: 0.77%
Refining Level 6: 0.47%
Refining Level 7: 0.25%
Refining Level 8: 0.13%
Refining Level 9: 0.07%
Refining Level 10: 0.04%
Refining Level 11: 0.02%

The chienkun refine rate table is visible when you hover your mouse over the item lol, in case you weren't aware. But the Tienkang Tisha rates are interesting.