In response to ztking: what I do is fraps the last couple of hits to the elite leader/last elite (the killing blow, when all other mobs are dispatched) then when all items have dropped I quickly mouse over with cntrl pressed (brings up the item tooltip), then pick up all rares and magic items if I am not certain about their affixes. Stop fraps, tp to town, log out, relog enter the same area, kill elite start fraps again. I keep doing this untill my inventory is either full with rares/blues or I have to go away/close the game.
When I have the time or when my inventory is full, I then open excel and open the movies I recorded using fraps and start writing down the data. It will go faster and faster I you do it a couple of times, and you will very fast see where the magic items dropped, and you can fill in 2 affixes right away. Also superior/white items only require a flash of time to write down. Ilvls are almost memorized for me as well now xD, most of them that is. The most time consuming are rares of which I do not know the affixes right away, I then check these items for the tooltip in the game. (I used to check all item tooltips in game, but tbh it's not needed for the white/superior/guaranteed magic items (always 2 affix) and the 1 affix magic items are easy to detect as well). When I was doing this I could get 25~35 entries / hour for warriors rest and I must admit I am always drinking/eating/doing something else in between as well.
- 8/17/2012 4:15:11 PM Posted in: Theorycrafting and Analysis
8/17/2012 1:04:33 PM
Very nice info on the bossess, this might also be in favor of the idea that the elite droprate is the same for all elites in all the acts for a given difficulty.Posted in: Theorycrafting and Analysis
It really is a shame that the bosses drop so few rares however, this makes it really pointless to do most bosses in most runs. Maybe they will change these droprates in the new patch, but who knows (I doubt they would change it tbh). 7 gear drops for one goblin is quite awesome too. It might be worth it to look into cyeron's goblin data and determine the average gear drops per goblin type from that data (currently not done I think?), It might be possible that goblins drop something like 3-4 items + NV + 2 slots for their own type (gem (seeker) /tomes (bandit)/extra gear/gold), no idea though. I might look into this myself sometime soon too.
About legendaries, yeah it seems like a huge boost to droprates if you play with high mf values, while the boost would be about the same (a little smaller because if the system rolls from top to bottom in terms of quality you get more legendaries, thus a little less rares) for rare items it just feels a lot more because you don't really see a lot of legendaries. Thus increasing your MF to higher values really feels worth it if you want to find legendaries. If the droprate is 1/2000 you will obtain a chance of 1/400 with 400 MF, which is in my opinion a huge difference. 1/400 means that 1 in 114 elite packs (without nv) will drop a legendary, with the NV rare it becomes 89 elite packs/legendary, instead of the about 570 packs (without NV) or 445 packs (with NV) at 0MF. You can easily get 80-100 elite packs in 2-3 runs I think, if you then also include the normal mobs that you kill on your way it wouldnt be to hard to get a legendary once per 1.5-2 hours with 400MF. If the base dropchance for legendaries is the same for all the slots of the bosses, these might be a nice source to get extra legendaries in the long run though. 11 gear drops for an end-act boss means that they might be worth it if you only want to farm legendaries (given that their dropchances are the same as for elites, or maybe even better than that of elites). But that would be about the only use of end game act bosses then...
An interesting question would then be if all types of legendaries and all types of ilvls can drop in all acts. This might be interesting with regards of the next patch. I think it is possible that the ilvls are actually determined the same as with all other items: dependent on the act. If you then can kill 40-50 elites/hour in act 2 with 400 MF this might be a nice farming spot for legendaries in the next patch (if you also take into account the 4x more droprates for normal mobs, although I doubt they would increase legendary droprate, prolly only magic/rare quality gets upped), if the remakes of the legendaries turns out te be so nice as they make us believe (it seems like everyone would need legendary weapons next patch).
A small note on gold drops, you mentioned you have seen 9 gold stacks drop at belial, have you ever seen this at normal/elite mobs, or is 8 the upper limit? And another question, does any1 think they see a difference in plan droprates at high MF :P?
8/16/2012 7:26:46 PM
Hello, It's been a while. I am not going to respond to every post though.Posted in: Theorycrafting and Analysis
Nice too see that blizzard is changing things (for the better on average). I do hope they won't change the dropchances for elites too much though.
About the 10/30/60 observed rare affixes distribution, yeah it may be something different, it indeed stuck with me because I observerd it myself and cyeron's data suggested it as well. For some reason I thought they were nice to follow and easy to work with, but combining nubtro's data, they might be something different. But this total rare distribution shouldn't be our main focus if we can obtain enough data for each gear slot, it would serve as something to check if you can approach it with the current dropchances (which we quite can already, ztking also showed that using the 2 main gear slots).
Anyway if I take the warrior's rest 0MF data on ilvl/affixes from Cyeron and then add the total results for the different gear slots by Nubtro and add my own data, it returns the following:
Affixes Amount Total Rares Chance 4 affix 207 322 0,64 5 affix 86 322 0,27 6 affix 29 322 0,09
About the slot system ZTking, I don't know if you know about the sheet posted a while back, but you can also edit every basechance for the different slots in that sheet (in the subsheets act 1-4) in case you would want it. (but if you have your own sheet now I guess it doesn't matter anymore). Link:
Nevertheless very nice work on the 2 slots that most of the time drop the rares. It's quite stupid tbh that the other 2 gear slots have such a low chance to drop rares that they become quite useless :(.
It's a shame I think the only thing we can do for now is wait on the new patch, what we (read I : p) need are more datapoints for the different gear slots for different kinds of mobs in different acts (bosses, more more elites, act 2/3). But I am afraid that Blizzard will actually change droprates for elites with the next patch :(. What would be the best thing to do next?
Edit: I deciced just to gather all data on legendary drops (except Shurafa's inbetween NV stacks data) because nobody here seemed to have done that and try to recalculate the legendary base observed droprate at 0MF. If the slot theory holds and we assume the game rolls quality from top to bottom, the legendary droprate would be 100% lineair with increasing MF. This means that we can recalculate the 0MF basedroprate for each entry by dividing the total amount of legendaries by (MF/100 +1). I currently also assume that each slot can drop a legendary and that the chances for each slot to drop a legendary are the same. This are the results:
Total Items Set/L MF NV MF Corrected Set/L Note 528 1 0 0 1 Murskautuminen Act I MF0 Sequence 1196 0 0 0 0 Nubtro Act I MF0 Sequence 907 0 0 0 0 Murskautuminen Act I MF0 Affixes 867 0 0 0 0 Cyeron Act I MF0 Affixes 1799 1* 0 0 1 Timza Act I & II MF0 1801 2 125 5 0.889 Timza Act I MF125 NV5 1392 2* 200 0 0.667 Timza Act I MF200 NV0 2245 3 200 5 1 Cyeron Act I MF200 NV5 479 1 238 5 0.296 HuiTzi Act II MF238 NV5 2425 3 267 5 0.817 syn83 Act II MF267 NV5 2700 6 300 5 1.5 Cyeron Act I MF300 NV5 446 1 313 5 0.242 Ghouul Act II MF313 NV5 Sequence 528 2 353 5 0.442 Cyeron Act I & II MF353 NV5 918 2 369 5 0.426 Nubtro Act II MF369 NV5 280 0 383 5 0 Shurafa Act I MF383 NV5
This gives me a total drop of 18511 items of which 24 items were legendaries/set. If we then sum the MF corrected legendaries/set we get 8.279 legendaries/sets in total. If we then divide this 8.279 by 18511, we obtain a observed chance of 0.000447 for MF0. Which is a bit close to 1/2000 (the chance I thought would be the legendary/set dropchance on average). This is just something I wanted to do, nobody seemed to have done it before. Something I found quite odd was that only 2 of these 24 legendaries were tagged as set items, and both set items were found in the data from Timza (see *, MF0 Act I & II, MF200 NV0 Act I). I also noticed you haven't used the HuiTzi data in the Elite analysis tab Cyeron, is there something wrong with this Data or is there no specific reason?
Is this method okay enough though?
8/9/2012 6:48:13 AM
Helllo there, this will be my last post here for the next few days.Posted in: Theorycrafting and Analysis
@nubtro, I agree with almost all of your points, do note that I made the sheet before I had most of my own data and your data (I made the most of the stuff this monday I think). About the chances, yeah they are nice round numbers that approach our total data, but they might be wrong. The reason why I used these round numbers is because I thought blizzard wont use numbers that are 17/397 and if i used these current numbers I would approach the main data points in the eltie farming section very close, like 0,3% off for most points out there, if I used the real chances we had got it still would be nice data but a bit more off. Problem really is those 3rd and 2nd drops that drop whites, we don't have nice chances for the magic items in these slots which makes it hard to get some nice data out of these slots with the calculated chances (in fact the magic chances we obtain are so high that you are likely to drop off the total white item count at high MF by 3-7% or something (compared to the data points in this topic) if i remember correctly.
About showing the entire dropsequence in the first tablem this is also possible but I don't think it is needed for the purpose of the sheet for the player. I entirely agree on points 2 and 3 tbh, we have a lot of tome and gem data, except plans/design data, which is actually the most interesting :(.
On the droprates for legendaries, yeah I dunno it has to be there with something, but it's likely a bit too high currently.
But you are free to work on it, I won't be here the next days and it is really in alpha stage or something, it is something I made quite fast during this busy week. Cya!
And I like the new update cyeron!
8/8/2012 8:15:20 PM
Hello, I am currently very busy with school and also planning a holiday this weekend. Next schoolyear will be the year I will do my Master's Thesis and I am currently setting up the basics to start out, getting everything okay including my plans for the coming month with my supervisor. I haven't had much time to reply here or gather data, but in the past few days I was working on the data collection for the different drop slots using fraps in warrior's rest and I was also working from time to time to create a very basic sheet to determine drops for each act related to time efficiency. Actually what you posted in your last post Cyeron.Posted in: Theorycrafting and Analysis
First I will link my datasheet again with the updated entries:
It's currently at 154 entries and I want to get it to 250 entries before Tuesday next week. I also included Nubtro's current results on affixes to determine the average chances for the different quality types for the different slots, this is what I got:
Affixes Drop I Drop II Drop III Drop IV Normal 0 0 264 102 0 Superior 0/1 0 81 32 0 Magic 1 0 36 21 0 2 327 12 12 332 Total Magic 327 48 33 332 Rares 4 48 3 0 46 5 17 0 1 12 6 5 0 0 7 Total Rares 70 3 1 65 Legendary/Set 0 1 0 0 Total 397 397 168 397
This is already a bit interesting and you can clearly see that the 2 magic guaranteed slots (in this case dropslots 1 and 4) drop more rares on average than the other 2 slots. You can also see that the guaranteed magic slots never drop less than 2 affix magics, again confirming the finding that there are at least 2 drops of > 1 affix magic quality. However the major problem I have with this data is that the sample size for the things that actually matter now for the slot system, which are the droprates for the rares in the 2 slots that do not drop magic items guaranteed. The amounts of rares in these slots is so low, it make it about impossible to estimate anything about chances for the specific affixes, also note the data on 1 affix magics is limited. It does however show that these slots have a very low chance to drop something that is worth anything in terms of quality for the player and these 2 slots contribute very little to the player's farming.
Now I am running out of time because I have to hurry for tomorrow's very short vacation, so I will quickly talk about the next thing I made, the thing Cyeron actually requested. The sheet is currently very basic and the layout is not top of the line, it's more about the principles in this sheet in its current state.
It can be found here:
The first page of the sheets is actually meant to be the only sheet that people will have to see or edit (hiding or fixing the other sheets is a possibility). It is currently not locked or hidden anywhere and you can everything in the sheet if I am correct.
The point of the sheet will be to determine, provided the player provides the sheet with his own MF, kills, time and deaths per act, to determine which act would be the best act to farm regarding rare drops, legendaries, tomes or plans. I am also planning to give some kind of abritary score to their input for each act which is relative to the time needed for a run in that specific act. This score is currently defined very easy and follows this: (#legendary items *10) + (#6 affix 63 rares * 9) + (#6 affix 62 rares * 8) + (#6 affix 61 rares * 7) + (#6 affix
The User would only need to change the variables in the first input sheet that are coloured in green, then the sheet automatically provides it's output in the same sheet below the data the player has provided the sheet. The player would then be able to see in a minor flash of time which act + parameters would be the best based on the final score. There are however some major problems with using this type of sheet in excel. The first one is that I need to have the magic find estimates and dropchances for every act laid out independetly in 4 different sheets, 1 sheet per act to determine droprates and data with it's own independent MF and NV values.
Another problem is the NV impact on magic find, to be absolutely accurate I currently would need to determine magic find and average drops for the first 5 elite packs that you kill with different NV values, because you currently go from 0-5NV stacks in those first 5 kills your magic find changes everytime you kill a elite. While these first 5 kills wil have very little impact on the total expected averages, it will be significant in smaller elite killing samples. What I currently do is determine the amount of elite kills and then determine for the first 5 kills the average MF over these 5 kills (which is +30MF on top of the normal MF you have at 0NV) and then use this average MF to determine the distribution of all rares/magics and other items for the first 5 elite kills.
But in theory I should have 6 different calculations to determine average item distribution:
1 for the first kill with 0NV + base MF
1 for the second kill with 1NV + base MF
1 for the thrid kill with 2NV + base MF
1 for the fourth kill with 3NV + base MF
1 for the fifth kill with 4NV + base MF
1 for all the other kills with 5NV + base MF.
So currently I use an averaged out MF for the first 5 of these calculations (Base MF + 0,5 * 4NV).
Other small things I need to note:
Each act currently has the same base chances for all the different slots, there is no difference in the chance to get a 6 affix rare between act 1 or act 4.
Ilvl distribution has been copy pasted from the latest blizzard hotfix and is used to determine average item lvl distribution.
the average drops for elites follows a slot system that rolls the item quality from top to bottom (Legendary -> Inferior (gray) item).
Average total items dropped is 3.5 (without NV) or 4.5 (with NV).
Most base drop chances for each slot are determined by analyzing data for 0MF 0NV data by Cyeron, Timza, Nubtro & Murskautuminen, most of the other data in this topic has been analyzed and can be used to check correctness of the system.
NO DROP IS "CALCULATED" BEFORE/INDEPENDENT OF SELECTING DROPS
NO IMPACT ON AVERAGE GEAR DROPPED BY: GOLD, TOMES, GEMS, POTIONS, PLANS
THERE ARE UP TO 5 SLOTS TO DROP GEAR (INCLUDING NV SLOT)
ONLY 4, 5, 6 PROPERTIES RARES DROP FROM INFERNO ELITES
MOST OF THE RARES ARE 4 OR 5 PROPERTIES
ELITES DO NOT DROP INFERIOR QUALITY GEAR
GUARANTEED DROP OF AT LEAST 2 MAGIC ITEMS WITH 2 AFFIXES
NEVER MORE THAN 3 ADDITIONAL ITEMS, ALWAYS 1 TOME, 1 GEM, 1 MYTHICAL HEALTH POTION MAXIMUM
THE 2 GUARANTEED MAGIC SLOTS HAVE A HIGHER CHANCE TO DROP HIGHER AFFIX RARES THAN THE OTHER SLOTS
Future plans (a lot of work to do everything):
determine gold, tome, gem, plans droprates (this last one will be near impossible).
check if all acts have the same base chances for item quality distribution (6/5/4 affix rares, 2/1 affix magics etc.).
determine drop tables for all NV affected bosses in each act.
Further determine base drop chances for the non-guaranteed magic drop slots.
Refine entire datasheet, including calculations and final score.
The current sheet seems to calculate average item quality distribution quite in line with the findings in this topic I would say. It has still to be perfectly optimized with the chances for the rares for the NV slot and the 2 non-magic guaranteed slots, but so far it looks like this:
MF NV %W %M %R %L/S 0 0 39.09 50.81 10.04 0.05 100 0 35.54 45.03 19.32 0.10 200 0 32.20 39.78 27.86 0.15 300 0 29.06 35.05 35.68 0.20 100 5 27.65 35.02 37.22 0.10 200 5 25.05 30.94 43.86 0.15 300 5 22.60 27.27 49.92 0.20 400 5 20.31 23.99 55.45 0.25
I have to go now and will be back on monday, I hope I didn't write this post too hasty and that you guys can use it. Otherwise my sincere apologies for wasting your time! Have a good weekend.
8/7/2012 2:55:38 PM
Posted in: Theorycrafting and AnalysisQuote from Cyeron
Just a quick note in general before I reply to your posts one by one: The slot system theory you have is an interesting system and combined with the drop sequence from Nubtro I believe it can turn out to become a single theory, possibly with a few tweaks. Or am I mistaken that the two findings can not be combined to one system?
Yes this is true, if you can determine the dropchances for the different item types (4 affix/5 affix etc) and each chance for every slot in the system with some accuracy we should be able to create a way to determine what elites drop on average. If we then apply this to specific data that people present themselves (like 30 elite kills in act 1 with 300MF in 60 minutes) we should be able to present them with the average amount of found rares and the amount of affixes. But if you want to be really precise we then should also determine the drop sequence chances and slot systems for every boss and for the elites from the different acts. If you have all that data you would be able to determine almost exactly what a person would find on his average elite farming run provided he gives the amount of kills and MF + NV. However we should then also determine the average dropslots for normal mobs that are not elites.
In theory it should be possible to generate a model that quite precisely determines the average amount of loot per run in a specific place.
Small update on my own data set: I am at 100 kills now nubtro. Going slowly today, think Ill do some more tomorrow.
8/7/2012 8:17:57 AM
It would be very appreciated if you can also sample at least affixes with it. I must admit I do find it very odd that nubtro finds not many 6 affix rares, in my own samples I am on my way to approach 10% 6 affix rares again. It's probably bad luck, but I my previous recordings I also seemed to quite steady approach a 10% 6 affixes amount. I do not know if nubtro checks every rare for the item code in the item's tooltip, but I guess he does. Which just means it's bad luck I guess, but yeah the most important data in these studies (in my opinion) would actually be the distribution of the amount of affixes per slot and not just rare/magic or not. It is possible to get a decent estimate for only the rare/magic distribution at 0MF 0NV (for act I mainly that is) already with the current data (especially if I also keep working on recording) I guess.Posted in: Theorycrafting and Analysis
But yes, you can try it out without recording # of affixes or ilvl. The new data from new acts is needed anyway, even without ilvl and # of affixes. Thx for helping out!
8/6/2012 3:41:42 PM
I always multishot everything and then pull the leader of the pack to the place where the ground is not cluttered with gold/globes/minions. So yeah I kill every minion and creep before killing the main leader. Your sheet looks good as well, I am not a fan of very small letters but it works hehe, the colouring is nice too.Posted in: Theorycrafting and Analysis
Well I will grab a few more kills and then see you tomorrow.
8/6/2012 1:15:14 PM
I have to go for dinner now, but I just whished to share my excel sheet with you nubtro. I just copied the basic layout of Cyeron's magic sheet and adjusted some things, I have 65 entries so far and in one of these entries the second slot (the guaranteed white/normal dropslot) dropped a legendary for me. Anyway let me know what you guys think of the sheet, I am always open for suggestions and feel free to edit whatever you like.Posted in: Theorycrafting and Analysis
And nubtro in response to the number of affixes, I think you will need to wait when you have more total rares. But must be some very bad luck so far xD.
Note: class means the item class that the tooltip trick gives us, so it basically is the number of affixes +2.
8/6/2012 8:31:37 AM
Well the ilvls are probably not different for each slot that is true. But we don't know that for certain and if you do spot a difference between slots, it's quite a finding! But yes it's extra work and probably won't show much of interesting data, combined with the fact that you need more than a couple of 100 items per drop slot to draw any good conclusions. We also noticed that the 3rd/4th drop are the ones that grant the 1 affix magics and that the 2 guaranteed drops are always 2 magic affixes or higher quality. Nice that you backed up again. I am very sorry it took me a while to finally start recording stuff with fraps as well, I had to go to several places this weekend so I did not have time to do anything on diablo 3. I am however going to start recording a couple of kill today, btw do you kill in warrior's rest act 1?Posted in: Theorycrafting and Analysis
Also: updated my excel sheet with the few corrections that were needed.
I just finished the data for affixes at 0MF 0NV for warrior's rest act 1 inferno. You can download the sheet here, please note that I marked 1 item entry in red, I forgot to write down the item level, but the Item class (3 for 1 affix magic) should be correct. The item in question was a ring with poison resistance for lvl 57 or up, so I decided to set the item level to 57:
EDIT: Updated datasheet with corrections.
- To post a comment, please login or register a new account.