Refining for cheap

WarrenWolfy - Sanctuary

Here's the spreadsheet I promised:

Refine Costs (Google Docs)

Google Docs will let you view it without signing in, but in order to actually use it you'll need to sign in and save a copy to your computer using "File"/"Download a copy". You can then use Excel, OpenOffice, Google Docs, or whatever else you'd like to actually use it. If I could, I'd make it usable as-is, but unfortunately Google Docs doesn't support any sort of "Permission to Edit, but not Save" option yet.

- EDIT -
Another version is available at Zoho Sheet, which can be used directly on-line:
Refine Costs (Zoho Sheet)

To use it, enter the current market prices in the blue boxes. If you want to refine weapons, remember to double the price of Mirages.

The most important results are in the green boxes, which tells you what the best refining choice is along with the average price you can expect to pay. Please remember that you're gambling! While the averages are (I believe) correct, it's quite possible to end up paying double or more of the average and still end up with nothing!

Other results include how many times you're going to have to click "Refine", on average, before you finally succeed, as well as a price comparison against using only Dragon Orbs instead.

The charts at the bottom of the spreadsheet show all the work-in-progress calculations for each type of stone. The costs are cumulative, and always assume that the previous best stones were used to reach the current level.

As a final note, please keep in mind that I may have made some errors in my calculations. While the results appear to be correct, I would appreciate some mathematical proof-reading, especially with the Tisha stone calculations, since they are the most complex ones.

WarrenWolfy - Sanctuary

Asterelle - Sanctuary wrote: »

Someone made an incomplete list of the principles of refining here:

4. Restart the game before you refine your gear. Me? I would reset my computer instead.
5. Close refining window every time you finish refining
7. Go on if you were successful in the second refining of the day, otherwise you'd better wait for tomorrow.
8. If you fail a refining attempt, wait a minute before you try again. Test with anything lower then [sic] level 5 until you are successful, then refine your important gear.
9. Each time before you do refining, lift your head up and shout "Praise be to the Heavens! Praise to the Earth! Praise to technology! Praise to the Internet! Praise to the game industry!", then kneel down with your head to the ground. This will get you in the mood for refining.

Principle #9 is by far my favourite. b:chuckle

People unfairly single out #9 and treat it dismissively, yet I've checked #4-#8 and I can confirm that #9 appears to be equally effective.

Deceptistar - Sanctuary

WarrenWolfy - Sanctuary wrote: »

Here's the spreadsheet I promised:

Refine Costs (Google Docs)

Google Docs will let you view it without signing in, but in order to actually use it you'll need to sign in and save a copy to your computer using "File"/"Download a copy". You can then use Excel, OpenOffice, Google Docs, or whatever else you'd like to actually use it. If I could, I'd make it usable as-is, but unfortunately Google Docs doesn't support any sort of "Permission to Edit, but not Save" option yet.

To use it, enter the current market prices in the blue boxes. If you want to refine weapons, remember to double the price of Mirages.

The most important results are in the green boxes, which tells you what the best refining choice is along with the average price you can expect to pay. Please remember that you're gambling! While the averages are (I believe) correct, it's quite possible to end up paying double or more of the average and still end up with nothing!

Other results include how many times you're going to have to click "Refine", on average, before you finally succeed, as well as a price comparison against using only Dragon Orbs instead.

The charts at the bottom of the spreadsheet show all the work-in-progress calculations for each type of stone. The costs are cumulative, and always assume that the previous best stones were used to reach the current level.

As a final note, please keep in mind that I may have made some errors in my calculations. While the results appear to be correct, I would appreciate some mathematical proof-reading, especially with the Tisha stone calculations, since they are the most complex ones.

looking at the sheet for my try the cost seems pretty accurate. The cost for my 12 gears +5 each cost roughly a little over 4M

Mumintroll - Heavens Tear

WarrenWolfy - Sanctuary wrote: »

Here's the spreadsheet I promised:

Refine Costs (Google Docs)

Google Docs will let you view it without signing in, but in order to actually use it you'll need to sign in and save a copy to your computer using "File"/"Download a copy". You can then use Excel, OpenOffice, Google Docs, or whatever else you'd like to actually use it. If I could, I'd make it usable as-is, but unfortunately Google Docs doesn't support any sort of "Permission to Edit, but not Save" option yet.

To use it, enter the current market prices in the blue boxes. If you want to refine weapons, remember to double the price of Mirages.

The most important results are in the green boxes, which tells you what the best refining choice is along with the average price you can expect to pay. Please remember that you're gambling! While the averages are (I believe) correct, it's quite possible to end up paying double or more of the average and still end up with nothing!

Other results include how many times you're going to have to click "Refine", on average, before you finally succeed, as well as a price comparison against using only Dragon Orbs instead.

The charts at the bottom of the spreadsheet show all the work-in-progress calculations for each type of stone. The costs are cumulative, and always assume that the previous best stones were used to reach the current level.

As a final note, please keep in mind that I may have made some errors in my calculations. While the results appear to be correct, I would appreciate some mathematical proof-reading, especially with the Tisha stone calculations, since they are the most complex ones.

Absolutely excellent. THANKS

Brael - Dreamweaver

WarrenWolfy - Sanctuary wrote: »

Here's the spreadsheet I promised:

Great job, I'll take a look at Tisha but I have to agree, they're a pain to do.
Is your boutique price for gold taking the 2% fee into account somewhere? Google won't let me see cell formulas.

WarrenWolfy - Sanctuary

Brael - Dreamweaver wrote: »

Great job, I'll take a look at Tisha but I have to agree, they're a pain to do.
Is your boutique price for gold taking the 2% fee into account somewhere? Google won't let me see cell formulas.

No, it's not included. I could certainly put it in, though.

Chillum - Dreamweaver

WarrenWolfy,

Do you still want refine data or are you done now that the spreadsheet has been put together?

Regards.

Asterelle - Sanctuary_1381265973

Very nice spreadsheet. I checked the formulas and it all looks accurate. Very nice.

WarrenWolfy - Sanctuary

Here is a slightly newer version hosted by Zoho Sheet. This version can be used directly on-line, although you can still download it if you prefer.

This version has slightly clearer formatting and also automatically adds the 2% Auction House fee to the calculations.

Perfect World - Refine Costs (Zoho Sheet)

Chillum - Dreamweaver wrote: »

Do you still want refine data or are you done now that the spreadsheet has been put together?

Regards.

I am curious to see how well people's experiences match up with the calculations, so yes, I'd be interested to see your data.

However, I'm now virtually certain that these are the correct numbers, so people need not exhaust themselves keeping 100% accurate records of their refines anymore. Out of respect for the forum rules regarding not discussing other versions of PW, I haven't given the exact details of how I confirmed these numbers, but unless PWI has chosen to break from other versions of PW, then these values are correct.

shootingstarvn

we got the area of Avalanche Canyon which duke black claims to have better chance to refine to +3 i can test now to refine a bit there

Vitenka - Dreamweaver

Yeah, I'd like to see some refuting of the rumours. (Well, I'd like to see substantiation more, but refutation will do)

WarrenWolfy - Sanctuary

Vitenka - Dreamweaver wrote: »

Yeah, I'd like to see some refuting of the rumours. (Well, I'd like to see substantiation more, but refutation will do)

I tested the rumour that you get better +1 refining if you refine at The Harshlands Officer by refining with 250 Mirage stones at the Officer, then 250 Mirage stones at the Elder. There was no statistical difference, with both coming in at ~49% success for +1.

I also tested these rumours:

4. Restart the game before you refine your gear.
5. Close refining window every time you finish refining
7. Go on if you were successful in the second refining of the day, otherwise you'd better wait for tomorrow.
8. If you fail a refining attempt, wait a minute before you try again. Test with anything lower then [sic] level 5 until you are successful, then refine your important gear.

They are all false.

The reason I've been (annoyingly, I admit) vague about my testing methods is because they involve both a different version of Perfect World and the use of a special debugging mode that allows me to see the exact refining odds at the moment of refining. I can't/won't do this on PWI, and I also can't/won't give any details of how to do it, because it would be a violation of the ToS.

Vitenka - Dreamweaver

Wait - refining odds are known on the client?

That's insane!

Strongholder - Heavens Tear

First off, let me say good job on putting together a very good spreadsheet.

I had a couple of questions regarding some of the formulas that i hope the creator or anyone else that may know, can answer.

The data through A26:I47 seems to all inlcude the cost of one mirage (makes sense bc one is needed per refining chance) + the cost of the refining item (ie tisha, tienkang, etc; which also makes sense bc you use one of each as well) and now heres where I get lost. All of that gets added to the cost of the previous level's refining item.

For Instance, the Tisha stones shown starting at H35. At levels 2 through 8 the success rate is shown to be 33+.035 %. If this was the case, why does it cost MORE to refine from level 2 to 3 than it does to level 5 to 6. The ONLY way it can cost more is if the succes rate has been decreased (or you have really crappy luck).

I am only trying to understand why this was done, not point out the flaws in the spreadsheet. I think a lot of work was done into this and i think its wonderfull. I hope someone can help me understand the why though.

WarrenWolfy - Sanctuary

Strongholder - Heavens Tear wrote: »

The data through A26:I47 seems to all inlcude the cost of one mirage (makes sense bc one is needed per refining chance) + the cost of the refining item (ie tisha, tienkang, etc; which also makes sense bc you use one of each as well) and now heres where I get lost. All of that gets added to the cost of the previous level's refining item.

The reasoning is that the cost of refining includes the potential for loss.

So, for example, when you use a Mirage the formula includes 70% of the cost of everything you've invested so far in the cost of attempting another refine, because you've got a 70% chance of losing everything.

For Tisha stones it's a more complicated calculation, because you don't lose everything if you fail, but rather you only lose 1 level of refining.

As for the second part of your question regarding it costing more to go from 2 to 3 than from 5 to 6, I'm not sure what you mean. Can you give me some more details?

Strongholder - Heavens Tear

alright the mirages make sense now, but what about the dragon orbs?

For example dragon orb 10 currently is 75 gold in AH, @ 440k per gold your sheet says it costs 33.66M for this 10 star orb (G15).

Why is it then that you add the cost of dragon orb nine in C45. Dragon orbs are 100% refine, so there shouldnt be a need to include the cost of the one before it, right?


Edit: Additionally, why does the price of the Tisha stones change (I45:I47) when you change the price of the 10 star orb? What does the price of the orbs have to do with the Tisha stone price calculation?

WarrenWolfy - Sanctuary

Strongholder - Heavens Tear wrote: »

Why is it then that you add the cost of dragon orb nine in C45. Dragon orbs are 100% refine, so there shouldnt be a need to include the cost of the one before it, right?

The section of data you're referring to contains running totals of all costs so far, assuming you are starting from +0.

For the per-level costs for Dragon Orbs, they are calculated in a different section from G6 to G16.

There's no equivalent per-level costs for any other type of stone because I felt those numbers would be misleading because they rely far too much on just 1 lucky/unlucky refine attempt. For example, the average per-level cost of going from +9 to +10 using a Tienkang Stone is ~110million, but in reality that's a misleading number. In reality, the cost has a 35% chance of being a mere 208k and taking you just 1 click, and a 65% of being about ~200million and taking you all-day.

Asterelle - Sanctuary_1381265973

Strongholder - Heavens Tear wrote: »

alright the mirages make sense now, but what about the dragon orbs?

For example dragon orb 10 currently is 75 gold in AH, @ 440k per gold your sheet says it costs 33.66M for this 10 star orb (G15).

Why is it then that you add the cost of dragon orb nine in C45. Dragon orbs are 100% refine, so there shouldnt be a need to include the cost of the one before it, right?


Edit: Additionally, why does the price of the Tisha stones change (I45:I47) when you change the price of the 10 star orb? What does the price of the orbs have to do with the Tisha stone price calculation?

Rows 36:47 represents the total cost to refine to that level given optimal choices before it. For instance row 44 is the total cost for refining to +9. C44 is refining from +8 to +9 with a dragon orb, E44 with a mirage, G44 with a Tienkang, and I44 with a tisha. There is a 25% chance for mirage to succeed going to +9 so on average +9 requires 4 refines to +8. The value of refining to +8 comes from C29. If there was a sale with nearly free +8 orbs than the cost of getting +9 with a mirage decreases since it is easy to get the needed 4 +8 refines.

WarrenWolfy - Sanctuary

I think the one last thing that this spreadsheet needs is some sort of "Chance of success" odds like Vitenka suggested. Unfortunately, this is where my math skills fail me and I must admit I'm at a loss.

I've been trying to figure out how to calculate what the chances are for each possible result when you specify how many starting resources you have vs. your target refining level.

For example, if you've got 4 Mirages and your goal is +1, the odds for the number of successes are:

0: 6.25%
1: 25%
2: 37.5%
3: 25%
4: 6.25%

I can do this example by hand because it's quite simple, but I can't figure out how to do it for more complicated cases, especially ones that involve Tisha stones.

Anyone able to help?

NaiaLeFey - Dreamweaver

refining chance is 1/3

Nelae - Heavens Tear

NaiaLeFey - Dreamweaver wrote: »

refining chance is 1/3

May i ask how did you come to this conclusion?
And I assume you are talking about all refines from +1 to +12

Zhoubotong - Heavens Tear

@ Warren

I can't remember the math either so that's why I ran the simulations. For example, those results then say how bad can things get. In the following output, on average it takes 126 tries to mirage up to +4. That is, 50% of the time it'll take <=126 tries. However, at the 90 percentile of bad luck, it will take 290 tries, so 90% of the time it'll take <=290 tries.

Level Base 75%ile 90%ile
1 2 2 4
2 10 13 22
3 37 54 90
4 126 177 290

Is this what you were asking?

WarrenWolfy - Sanctuary

Zhoubotong - Heavens Tear wrote: »

I can't remember the math either so that's why I ran the simulations. For example, those results then say how bad can things get.

That seems like a really good idea! So I made some Excel macros that will run simulations. It's not as elegant as having the actual calculations, but I have to admit it works pretty well.

Unfortunately, Zoho Sheet is painfully slow and it doesn't render the macro buttons properly. You have to manually select "Macro -> Run Macro -> ResetTrials" because the button isn't showing up. I highly recommend exporting this to Excel, since it will run hundreds of times faster (literally) and the buttons will show up correctly.

Refine Costs with Simulation Macros (Currently buggy on Zoho Sheet)

I'll post a better version if I can figure out how to work around Zoho Sheet's bugs, but for now you can just export it and run it off-line.

Here's a sample of the results I've looked at so far:

Starting with 8.2million (roughly the cost of what you'd pay for Dragon Orbs up to +5), trying to get as many +5 items as possible before running out of cash. 4500 trials runs.

0 items refined to +5: 8.8%
1 item refined to +5: 23.5%
2 items refined to +5: 29.1%
3 items refined to +5: 20.5%
4 items refined to +5: 11.6%
5 items refined to +5: 4.7%
6 items refined to +5: 1.3%
7 items refined to +5: 0.3%
8 items refined to +5: 0.02%
9 items refined to +5: 0.02%
10 items refined to +5: 0.02%

So basically, this means that 8.8% of the time you'll lose everything, 23.5% of the time you'll break even with what you could have gotten with just Dragon Orbs, and the remaining 67.7% of the time you'll end up with 2 or more +5 items.

To me, those are extremely good odds! With only an 8.8% chance of failure, and the potential gains so high, I'll gladly take the risk.

Devoted - Lost City

was never intending to post data here ever. this kinda forced me to:


0->1 : 62/235
1->2 : 3/62
2->3 : 0/3

counted 15 fails in a row once in a sample of 300.... I should buy a lottery ticket, the odds of it accidentally being printed on solid gold is higher. 3/62 @ a 30% is just as absurd.

turned off my computer and got:

0->1 : 1/1
1->2 : 1/1
2->3 : 1/1

and stopped... I've never had such troubles refining before in my life, my hh90 gear is all +5 without orbs and I think I used less than 300 mirages for my entire set. The above case should just be forgotten, it's just gross.

Devoted - Lost City

another 400 and i got +6, didn't record stats i just rage-refined

Mistimancer - Heavens Tear

Well I must be quite unlucky then ROFL

Here my experience with most of my refining ( don't remember the level of gears but am over 80)

-Ring +1, wanted to bring to +2.... after 30 mirages, I was still at +1 haha
-Weapon, TT 80,.. spent over 50 mirages to bring it to +2.. another day I tried to bring it to +3.. spent 60 mirages and it was still +2 so I gave up and got an orb hahaha
I tried many times and never been able to refine over +2 without orbs.. well I guess I could have if I would spent millions of mirages per items.. if lucky

hahaha

WarrenWolfy - Sanctuary

Cholla - Harshlands wrote: »

For example, I think rumor 5 is about avoiding the problem where you replace your stack of +1 dragon orbs with a +2 dragon orb but you do not drag it in properly and you wind up wasting a +1 orb. If you close and re-open the window between each refine you will not run into this problem. And I do not think you proved this rumor "false".

I suppose you could interpret rumour #5 this way, but if the question is whether or not closing the window between attempts increases your odds, then the answer is no, it does not.

Cholla - Harshlands wrote: »

For example, I think rumor 7 means that under some conditions your odds will be significantly lower than under other conditions. Statistically speaking, the odds of getting "two in a row" will be significantly worse when the overall odds are lower. For example, if your chance of refining were 10% your chance of two in a row would be 1% where if your chance of refining were 1% your chance of two in a row would be 0.01%.

Sure, but I'm not sure what that has to do with whether or not you should give up or keep trying based upon whether or not your second refining attempt of the day succeeds. Whether or not you succeed at your second refining attempt has no effect upon your odds of further successes.

Cholla - Harshlands wrote: »

And, from my own personal experience, sometimes odds are much worse than other times. So, again, I do not think you proved this rumor to be "false". (For example, even if you had disassembled the code that was being used, other code might be used under different circumstances.)

If you can not give us the full details of your testing methodology we can not reproduce your tests. And if you can not even describe exactly what you thought you were testing, your tests can not mean anything for us.

You're right. I have not proven the rumours to be false, due to the fact that I haven't given full details of the tests that I used.

As I've stated before, my reasons for not giving specifics is because the tests were done on a non-PWI version of Perfect World, and out of respect for the forum rules I am trying not to get into a discussion about non-PWI versions of the game. Furthermore, doing the tests requires mucking about with the game files in ways that are ban-worthy if done with PWI.

The only reason I've even posted as much info as I have is because I see the info as being similar to the situation of the oft-cited pwidatabase. In that case, discussion of the data that's available from pwidatabase is accepted (or perhaps I should say tolerated) here on the forums, but a discussion giving technical specifics about how exactly to acquire the data would be ruled as a breach of forum rules and locked.

As I've pointed out, everything I've said comes with the disclaimer that my tests are only valid under the assumption that PWI uses the same refining rules that the non-PWI version I tested uses. Since Attempting to confirm the same tests on PWI would be, A) impossible, and a violation of the ToS, I have no desire to attempt to do so.

For me personally, the fact that the empirical data we were all gathering turned out to exactly match the results I ended up with from the non-PWI version is enough evidence to convince me that the rules for refining in PWI are now known with a high degree of certainty.

As for Duke's rumours, Duke Blacke gives out all the exact same rumours on the version of PW that I tested. I know with certainty they are false, yet a search of the forums of the version I tested reveals many players convinced that they are true. My conclusion is that they are false in PWI too, yet I don't begrudge people their superstitions.

If a rumour, false or not, gives people the confidence to attempt refining in a way that I know statistically gives them favourable odds, then to me that's a positive superstition. I view it in the same light as Dumbo's magic feather, because even though it had no real magic, nevertheless it was essential in giving him the confidence he needed when he needed it.

I really do sympathize with the fact that, "because WarrenWolfy said so," is a truly lousy answer. If I were you, I wouldn't trust my word either. But the forum rules are the forum rules. I genuinely apologize.

_DarkSeph_ - Sanctuary

I'm beginning to really believe this thing about the random generator "sticking" ^^

Dawdlers belt
+1 (2/3)
+2 (1/2)
...then

Cuisses of Seacaptain
+1 (2/2)
+2 (1/2)
+3 (1/1)

The belt was an easy +2 and the cuisses went to +3 even easier. I usually don't go above +2 but I was getting such a lot of "success strings" so I went for it on the spur of the moment. Oh and...

Armlets Of Antiquity

+1 (2/3)
+2 (1/2)
...the other day. All in archosaur.
Not enough to help much with statistics but just sharing my general experiences.

I have a few hundred mirages saved and 15 tienkangs and tiesha's from the tide-quest, so I'll be having a shot at a no-orb, high as possible refine on my GX pretty soon.

VashtaNerada - Dreamweaver

I really don't expect to see this kind of luck again - knocking on wood, just incase - when it comes to refining, but:

Level 72 protection belt: + 2 (2/2) & TT70 Jaden Spear +2 (2/2)

Before those though, I had pretty rough luck with refining. My cape absolutely did not want to reach +2 w/out d-orbs (still went w/out) and I burned more mirages than I care to admit out of tenacity.

WarrenWolfy - Sanctuary

The follow up to my spectacular failure to refine my Cape of Tauren Chieftan to +5:

Started at +3, refining at Archo Elder.

+1: 7/9 Mirage
+2: 4/7 Mirage
+3: 3/4 Mirage
+4: 3/4 Tienkang
+5: 1/3 Tienkang

Total cost: 27 Mirages @ 14k each + 7 Tienkangs @ 90882 each (ie. 330k Gold prices) = 1,014,174

Obviously a good run of luck, which roughly balances out the horrible run of luck earlier.

According to the Excel spreadsheet, after 50,000 trials, the odds of getting to +5 once using 1,014,174 at the given prices is 21.0%, twice is 1.43%, and 3 times (!) is 0.018%.