Diablo® III

[Discussion] Crafting Strategies & Odds

170 crafted DEX amulets, no trifecta, not even one with low stats...however I did well with my dual amulet so I'm OK with it and stopped after 170 crafts. I may pick it up again at some point.

Another thing...I have good high DEX trifecta none crafted gloves...now they could be better but it's going to take a heck of a lucky RNG roll, so I've only crafted about 30 pairs of gloves...didn't even get a dual attack modifier roll on any of them and quickly abandoned that pursuit.
Reply Quote
The truth is, when somebody comes to me to build a set and wants to craft away, I'll almost always say that's not a good idea. I'll just about always propose that he let me buy on ah first, so that I can guarantee the EHP stats that I want.

100% agree. I truly believe that EHP stats are the hardest to fulfill because of how specific the requirements are. It is hard enough to roll for 4 or 5 specific affixes to improve or keep up your EHP but to also demand great rolls? Far better to get it through the AH. So yes, I also cringe when someone is suggesting crafting for an EHP slot. It'd be the last thing I'd be spending any resources on.

03/30/2013 09:29 AMPosted by Silver32
Are we absolutely sure that every affix has an equal chance of rolling?

I've only assumed that each affix has an equal chance of rolling, but honestly, I have no idea. But if there are coefficients for each affix to determine likelihood of each affix rolling... that will be hidden to us until someone collects enough data (with a meaningful sample size) to estimate what that is. And if someone ever figures that out, it wouldn't take much for Blizz to change that up on us without us know. So my answer? Who the hell knows? But I'll just stick to equal probability just for simplicity's sake.
Reply Quote
I just have one question and I may be totally wrong here.. but does MF increase the chance of the RNG being in your favor for found gear? If so, does the MF affect your crafting? And if we come to a conclusion of yes the higher the MF the better the gear COULD be, then would it not be smart to say, craft only in MP10 and with full MF gear (of not P100 of course)
Reply Quote
@XxNitewolfxX
I don't think MF has any impact on how RNG works. As far as I know, MF only affects monster drops and I don't think it impacts crafting in any way shape or form. I am not an expert in this, so I'd welcome others to respond with more authoritative word.
Reply Quote

I've only assumed that each affix has an equal chance of rolling, but honestly, I have no idea. But if there are coefficients for each affix to determine likelihood of each affix rolling... that will be hidden to us until someone collects enough data (with a meaningful sample size) to estimate what that is. And if someone ever figures that out, it wouldn't take much for Blizz to change that up on us without us know. So my answer? Who the hell knows? But I'll just stick to equal probability just for simplicity's sake.


I can absolutely guarantee this is not the case and that some affixes are much more likely than others. Just looking at the data I put in the thread earlier about vit-bracer affixes (roughly 100 crafts collected which is 500 non-vit affixes) we can see that the variability of chances is too high for an equal weighting of each affix to occur. (Post #20)
Edited by Xoran#1546 on 3/31/2013 10:33 AM PDT
Reply Quote
03/29/2013 10:01 PMPosted by Nameless
Is this reasonable? Or is this completely non-sensical?

I think of it as a two-by-two matrix:

  • Gear: You're either well-geared, or you're not.
  • Gold: You either have a lot, or you don't/

This results in four possibilities:

  • You have good gear and a lot of gold. Upgrades on the AH will be unbelievably expensive, so you should craft for upgrades.
  • You have good gear, but you're now out of gold. You should craft, but you can't afford to. Start going on runs for crafting mats, and then craft for your upgrades.
  • You have poor gear, but a lot of gold. If your gear is rather poor, you can probably upgrade easily by crafting. Given how cheap decent gear is, however, you should look at the AH first.
  • You have poor gear and no gold. You should save up a decent amount and hit the AH so you can buy guaranteed upgrades.

I fell into the third category (poor gear, lots of gold), and crafting worked out fine for me. I don't think that it works best for well-geared characters, primarily because well-geared characters will have a very hard time crafting across-the-board upgrades. I think it works both ways for poorly geared characters, though. If you have the money, roll the dice on crafting; if you don't, hit the AH first. If you're in a weird in-between financial state, scour the AH for deals, and craft only if you can't find anything good.

Now, I'm no AH expert, so this might be off, but that's how I think of it.
Reply Quote
I've only assumed that each affix has an equal chance of rolling, but honestly, I have no idea. But if there are coefficients for each affix to determine likelihood of each affix rolling... that will be hidden to us until someone collects enough data (with a meaningful sample size) to estimate what that is. And if someone ever figures that out, it wouldn't take much for Blizz to change that up on us without us know. So my answer? Who the hell knows? But I'll just stick to equal probability just for simplicity's sake.


I listed out the coefficients in the google doc...
Reply Quote
For whatever its worth most scientific calculators have a combination function to do these sorts of problems. Usually they're of the form cNr but how you enter the fields from there can vary as the 'correct' way to do so doesn't lend itself well to a single line for entries.

Outside of that though I agree with the premise of the thread that both Nameless and Vrkhyz are offering, if your poor and starting out the AH is way more budget friendly. I've spent about ten million to get moderate stuff out of crafting for my monk when there are several slots I could easily upgrade for that amount.

Shame I enjoy crafting more then playing the ah refresh game!
Reply Quote
That PDF is craaaaaazy informative.

Too bad it's all a little over my head :D

Also, if this is a double post, I blame teh intrawebs.
Reply Quote
03/31/2013 10:31 AMPosted by Xoran
I can absolutely guarantee this is not the case and that some affixes are much more likely than others. Just looking at the data I put in the thread earlier about vit-bracer affixes (roughly 100 crafts collected which is 500 non-vit affixes) we can see that the variability of chances is too high for an equal weighting of each affix to occur. (Post #20)

I will need to take another look. The thing that we do have to keep in mind is that double-stat rolls can be a little sneaky. I think there are times that it could give you a double-stat roll where one of the stats is the main one you are crafting. I think that's the case in the gloves I have right now. Neverthelss, I would like to take a look at this.

03/31/2013 04:07 PMPosted by Chazhang
I listed out the coefficients in the google doc...

Man, if there are coefficients, I wouldn't know how to integrate that into what I'm trying to figure out. Any ideas? I think the simplest (but not necessarily the most accurate) way of doing this is to apply the coefficients after the fact? *shrug* Again, not my strong suit.

My whole thing in here is that the odds of you getting the rolls you want work exponentially against you with each affix you want to try to fix, and this doesn't even take into account the likelihood of getting decent rolls in any of them. Balance that with the crafting costs (hidden or otherwise), and I think we will come to know what we should already intuitively sense... the odds are hugely stacked against you with each affix you demand in your craft. And saving gold you'd otherwise use to craft in huge batches may be better used towards AH purchases in the longer run if you are a mid or lower tiered monk, and AH prices for the level of gear you are looking to buy are still reasonable.
Reply Quote
You can estimate it quite accurately (maybe within 10%?) using equation 4. The more affixes you want that have a very disproportionate roll rate (ie IAS/pickup/thorns) the less accurate it will be.

With that you should be able to calculate the odds of rolling say:
9 IAS (affix level 62 or 63 roll; affix weight of 3)
8.5+ crit chance (affix level 63 roll; affix weight of 6)
35+ crit dmg (affix level 63 roll; affix weight of 6)
on gloves.

If you're really adamant about getting things accurate to the dot, you'd have to create a very complex nested combinatorial sequence that would have a total number of terms equal to N*(N-1) +1 in the numerator and denominator similar to how the exclusive affix property formula was written in equation 1, where N is the number of possible affixes the item can roll.

The middle ground and easiest solution is to do what Mandlebarb did. Program in all of this into a simulation program in Matlab. Run a simulation of 100 M or so items, determine the probabilities based on the roll rate within that 100 M item sample.
Edited by Chazhang#1745 on 4/1/2013 1:34 PM PDT
Reply Quote
Affix weighting is the easy part of this equation IMO. All you need is the available range of each affix as well as the desired range, and for each affix multiply by the probability of the desired range being in the available range. This is assuming a linear affix spread which is likely due to the simplicity of coding a random number and then multiplying it to fit the range.

There is the problem that ilvl63 amulets may be able to roll ilvl62 and lower affixes of the same stat, which actually changes the probability of getting a certain range. You would have to figure out the chance of rolling each affix then multiply by the chance of that affix being in the desired range. But I dont know the chances of getting ilvl62 CD roll compared to ilvl63 CD roll, so can't comment on that (they may only roll ilvl63 affixes faik)

The hard part is the exclusivity of some affixes which I haven't yet got my head around fully, because it depends on which order the affixes are rolled in, rolling an exclusive affix early increases your chances of getting other affixes later on, however if you roll it last then it has no effect at all.
Reply Quote
Bumping this thread for its informative contents.
Reply Quote
04/15/2013 12:37 PMPosted by Tobes
The hard part is the exclusivity of some affixes which I haven't yet got my head around fully, because it depends on which order the affixes are rolled in, rolling an exclusive affix early increases your chances of getting other affixes later on, however if you roll it last then it has no effect at all.

I would treat those as "umbrellas" in that you treat all exclusive affixes as one, and then from within that, determine your odds of getting the exact affix you want from within that umbrella. So in the case of elemental resists, I would imagine that the coding would look something like

(1) did it roll an elemental resist? if yes, then
(2) choose one of 6 (?)

Likewise, do the same for double stat rolls... (dex/vit, str/vit, int/vit, dex/str, dex/int. str/int) - 6 possible I think. In any scenario that rolls any singular resist or double stat roll, multiply by 6 and 36 if it has both.

Again, that's just my guess on how that works to get the total number of possible combinations. I have yet to work out the math, but I've done enough math to know that to go from 4 desired affixes to 5 desired affixes become exponentially more difficult, and even more so if you are looking for decent rolls in those. Keep the requirements lower, and it'll be "easier" to get.

If you check my profile, my current ammy is kind of a placeholder and I plan to roll all kinds of amulets at this stage of the game. Once I get something good, I'll find a way to rebalance my gear to accomodate it (likely through my inna's pants if I don't get a life% or vit roll in my ammy). I'll probably start crafting those soon.
Reply Quote
@Nameless True that is a very good way of looking at it, but the problem is you actually have extra chance for that "umbrella" stat ie you are six times more likely to get a single resist stat than another specific stat
Reply Quote
Affix weighting is the easy part of this equation IMO. All you need is the available range of each affix as well as the desired range, and for each affix multiply by the probability of the desired range being in the available range. This is assuming a linear affix spread which is likely due to the simplicity of coding a random number and then multiplying it to fit the range.


Since you can only roll an affix once it doesn't work out this way. Its more like cards where P(drawing a Q and K in a hand of 5) is not equal to P(drawing a Q in a hand of 5) * P(drawing a K in a hand of 5)

There is the problem that ilvl63 amulets may be able to roll ilvl62 and lower affixes of the same stat, which actually changes the probability of getting a certain range. You would have to figure out the chance of rolling each affix then multiply by the chance of that affix being in the desired range. But I dont know the chances of getting ilvl62 CD roll compared to ilvl63 CD roll, so can't comment on that (they may only roll ilvl63 affixes faik)


The approximate probabilities were shown earlier.

The hard part is the exclusivity of some affixes which I haven't yet got my head around fully, because it depends on which order the affixes are rolled in, rolling an exclusive affix early increases your chances of getting other affixes later on, however if you roll it last then it has no effect at all.


This part isn't that hard, see equation 1. Both exclusive affixes (specific resistance and combo stat) are shown in that combinatorial sequence and how to incorporate it.

I would treat those as "umbrellas" in that you treat all exclusive affixes as one, and then from within that, determine your odds of getting the exact affix you want from within that umbrella. So in the case of elemental resists, I would imagine that the coding would look something like


This is the correct way to go about it, the problem is that you have to nest everything. Given the number of affixes the nested combinatorial sequence would be ridiculously long.

I was able to dig up Mandlebarbs simulation from 10M amulets he programmed into matlab:
https://docs.google.com/spreadsheet/pub?key=0ArooAQ_ASbpjdHRpWnowSTNNcm41ZFVsS09ZUkpwc2c&output=html

Ignore the "top half of rolls" thing since he assumed linear affix level probabilities which was shown to be false, but the rolls to craft is quite accurate.

Like I mentioned previously, all the formulas are there, just need to plug in the specific stat combinations that are considered "desirable" to you and figure out the probabilities like Mandlebarb did.
Reply Quote
04/15/2013 05:00 PMPosted by Chazhang
This is the correct way to go about it, the problem is that you have to nest everything. Given the number of affixes the nested combinatorial sequence would be ridiculously long.

Perhaps. But I don't think it'll be too huge of a deal. I'll give it a whirl with amulets to calculate the total number of combinations (taking into account exclusive affixes) and see how closely calculations matches up with Mandlebarb's resulsts (assuming equal probability for all affixes... although I am aware that there may be different affix weightings... I think it makes a difference, but not so much that it will make a significant impact on the final numbers). I just want something of a rough estimate so that we know what kinds of odds we are facing when it comes to crafting and targeting specific affixes.

Once I work something out, I'll post some results... Thanks for bumping this and keeping it alive again. :)
Reply Quote
Nameless, after perhaps 150-200 useless Dex ammy crafts, I've started to keep track in an excel file. If you are planning to do the same we should 1) combine and 2) perhaps combine with someone in General Forums who did same, but for Vit ammies.

P.S. Tracking in Excel has given me worse RNG--don't think any crafted ammy since then came closer than -20k DPS, and my ammy has no EHP to give up either. My best earlier-crafted dex ammy is on my DH now
Reply Quote
@crushkyle -- how have you been tracking it? I wasn't planning to track it -- it might take too much time. Could you post it in a google spreadsheet, share it, and then provide me with a link? Maybe I can start putting in some of my stats in your spreadsheet to track the results.

Truth be told, I haven't devoted too much gold to crafting yet (outside of the initial couple of days/weeks to test it out) because I have been saving my gold for actual gear / weapon upgrades. I might set aside 5M for crafting over the next while and focus on ammys and gloves and see where it goes.
Reply Quote

Please report any Code of Conduct violations, including:

Threats of violence. We take these seriously and will alert the proper authorities.

Posts containing personal information about other players. This includes physical addresses, e-mail addresses, phone numbers, and inappropriate photos and/or videos.

Harassing or discriminatory language. This will not be tolerated.

Forums Code of Conduct

Report Post # written by

Reason
Explain (256 characters max)

Reported!

[Close]