6-socket Orb of Fusing Community Log (done)
" " I'm no statistician but I calculate those to be approximately 1000000HEADASPLODE%%%% better chance than using fus/jwl on a non-quality item. I think we need more data. Have modem, will travel!
Silas' Gear & Gems: http://www.pathofexile.com/forum/view-thread/426367 |
|
" 100++ fusing, 400++ armourer scrap : 5L/6L = 0 ...
* Forum Avatar - $500USD ... -- Bex_GGG |
|
|
8fusing 20% - 1 5L
GMT+1 - You can speak German with me
|
|
|
IMPORTANT NEWS
Did a major update from my hiatus on April 24 until now, excluding quality results. I also took the time to separate the new data (April 24 onward) from the older. You can view both data sets, as well as the grand total, in the OP. These two batches of data are very divergent. Notice the 95% margin of error... according to the old sample, there's less than a 5% chance of the new sample existing, and vice versa. Either GGG was secretly changing rates behind the scenes, or there's some serious selection bias going on here. According to the old data, 5Ls are fairly hard to get, but 6Ls aren't so bad (only about five 5Ls per 6L); according to the new data, 5Ls are much easier, but 6Ls are a nightmare (about ten 5Ls per 6L). Since I didn't see anything in the patch notes about changes (until the obvious 0.11.0 change), and I don't think GGG would blindside us with stealth changes, my conclusion is that reporting here is very biased. Early on, the mass numbers of 6L QQ threads may have led to over-reporting of 6Ls; later, the decreased novelty of 6L QQ may have led to mostly bragging posts about easier accomplishments, and less serious reports of 6L attempts. I'm not sure of the exact cause, but I firmly believe that social forces are contaminating results here. Sorry, community, but your data is bad. (Feeling bad is optional.) Probably still a lot better than the average player's off-the-top-of-their-head estimate, but not good enough for further statistical inquiry. As a result, I'm not very interested in maintaining a 20% quality community log, or continuing data accumulation further in this thread. I do plan on conducting my own, independent research, which will take months (perhaps years) to gather a proper sample size, but will have data I know I can count on. If someone else wants to start a 20% quality community log, they are free to do so. I'd recommend creating a new thread so that the OP could edit in the updated results. When Stephen Colbert was killed by HYDRA's Project Insight in 2014, the comedy world lost a hero. Since his life model decoy isn't up to the task, please do not mistake my performance as political discussion. I'm just doing what Steve would have wanted. Last edited by ScrotieMcB#2697 on Jun 9, 2013, 1:05:29 AM
|
|
|
so i wanted to craft a karui maul for my leap slammer so i bought a ilvl 78 to 6L
took about 150 jewellers to 6S without using any whetstones stocked up on 154 fusings and 616 whetstones to 6L after 88 fusings and 352 whetstones: 5L: 0 6L: 1 now on to using alts+augmentations+regals for awesome dmg roll i think chances are definitely better with whetstones however, i did 6L a lioneyes before the patch with 14 fusings..that was pure luck though Last edited by hegotgame#7243 on Jun 9, 2013, 1:40:45 AM
|
|
|
30-40 jew to 6S before patch 587 fusings (196 fus after patch used) 260 chromatic didnt used any armourer scraps 5L - 5x 6L - 0x Now i will try with 20% quality with just few fusings (dont got much armourer scraps:P) @Eryslaf @Eryslav Last edited by ericcol2p#2566 on Jun 10, 2013, 2:11:07 PM
|
|
|
Fusing Used:47 5L:0 6L:0 |
|
|
I think data should be collected in this better way:
1) we have 5 possible link on a 6S, each link can be on or of. For each fus used, we record the Q% and the number of 'ON' links (5 links on means that the item is 6L). 2) I STRONGLY believe that, before patch, the probability was 1/4 for each link, as I discussed previously. In my opinion, a fast way to implement a prob gain with quality is to change the above ratio as a function of Q%. 3) If the theory is right, only few dozens (or one hundred) tests are needed for each Q%, since links on and off have higher occurency than 6L or 5L. Unfortunately i do not play much, and only have a few fus/arm to make tests. The low-link number data is far more important than 5L and 6L link data. Roma timezone (Italy)
|
|
"I strongly believe you're wrong, as GGG is on record saying that a single roll is used to determine the number of links in a link group (not five separate success rolls for a 6L). And even if that were somehow true, #3 is woefully ignorant of how statistics work. When Stephen Colbert was killed by HYDRA's Project Insight in 2014, the comedy world lost a hero. Since his life model decoy isn't up to the task, please do not mistake my performance as political discussion. I'm just doing what Steve would have wanted.
|
|
" The fact that each single link between sockets could have an INDEPENDENT probability is not related to the number of 'rolls' to obtain these links. We do not know how GGG rool dice technically, but of course we can suppose that a 'basic' roll is a bit which has 50-50% of 0 and 1. We can have more than one results with a single 'roll' if more bits are given with a single dice (in old computer a roll was made of 16 pseudorandom bits, maybe GGG use 64 bit, or whatever else). " I am not ignorant. I can ensure you that if the theory is right 50 test for each of the 21% (1050 tests) is enough to have a precise estimation of parameters. I also find demential to try to make statistics only on exceptional events when large amount of data is also available on more frequent events. Results for exceptional events can also be estrapolated from frequent events. I think that the probability of each link is 1/(A-B*Q), where A and B are parameters and q=0-20 is quality. To prove or disprove this statement, we NEED data for each fusing and on more frequent rolls and on how sockets are linked. Roma timezone (Italy)
|
|
