Infernal Pit SpecialChest runs: high probability of Zod and other HRs

[vmiguli]

There are many known methods for rune hunting in Diablo 2. Most commonly, gamers pursue Lower Kurast chest-opening, Cow-level slaughtering, or Chaos Sanctuary, Pindleskin, and Baal runs. In my opinion, the most effective are the LK runs, however, both the monotony of the repeating procedure and the rune-level restriction (up to Ber only) are the major drawbacks of this approach. On the other side, the fighting strategy requires dealing with enormous amount of monsters, since the probability of a rune drop from a killed monster is by an order of magnitude lower than a similar drop from a container at the same player settings.

Is there any possible way to perform effective chest runs in order to get HR’s (including Zod) that would time-wisely compare with the known algorithms? Looking for an answer to this question I went through different Act 5 (Hell) areas searching for high chest density locations. Earlier in this forum my game colleagues showed an ineffectiveness of the SuperChest hunting in Act 5 due to its irregularity in appearance, as compared to Act 3 (LK) and Act 4 (RoF). And Hell Act 5 is the only place in game where Zod might drop from any container.

In this post I would like to describe my short research of a chest rune hunting in Act 5 based on repetitive runs to the mini-hell levels: Abaddon and Infernal Pit. Actually, the third such a mini-level in the game, namely Pit of Acheron, has appeared way too far from the Arreat Plateau WP in my working map, so I have abandoned an idea of getting there after 100 time-consuming runs.

All these mini-hell levels offer one SpecialChest (NOT a SuperChest like in LK) at the end of the pathway. This chest has a higher rune drop probability than any other normal container in this area; at the same time, any repeating combination of the dropped goods (that is observed for SuperChests) was never recorded for this type of chests, to my knowledge, thus not confirming an assumption of a reasonably limited amount of possible drop combinations (like the “65,536 hypothesis”). This means that only a substantial amount of runs could have shown the potential of this place for rune hunting.

My strategy included running/teleporting from a WP to the corresponding portal and then to the final SpecialChest. While moving, I tried to hit as many containers (chests, corpses, stashes, etc.) as I could without losing too much time. Avoiding fighting was crucial for my task since it would dramatically increase the time needed for a run, and thus, impeding the proper estimation of the chest-opening approach. However, the monster-killing was sometimes unavoidable when the amount of the dropped units from the SpecialChest exceeded the screen-seeing capability (and that happens every other time at Player Settings = 7), and I had to clear up most of the junk to be sure not to miss a Zod.

I also kept track of Uniques being dropped from the SpecialChest, however it might depend on the MF% of my gear, so I don’t know if these numbers are that useful.


Abaddon – 1,240 runs

10,700 containers (total)

1,240 SpecialChests (total)

368 runes (total) = 212 (cont) + 130 (SC) + 26 (monst)

16 Unique drops (SC only) included 25 units (total)

HRs (Pul+): 2 x Mal, Gul, Ohm (cont); Lo (monst)

% cont = 2.0% – an average probability to get a rune from a container

% (Pul+) cont = 0.037% – an average probability to get a HR from a container

% SC = 10.5% – an average probability to get a rune from a SpecialChest

% Unique = 1.3% (drops) or 2.0% (unit) – an average probability to get a Unique drop/unit from a SpecialChest


Infernal Pit – 1,960 runs

32,200 containers (total)

1,960 SpecialChests (total)

829 runes (total) = 536 (cont) + 200 (SC) + 93 (monst)

37 Unique drops (SC only) included 49 units (total)

HRs (Pul+): Pul, Um, 2 x Mal, Ist, 2 x Gul, Vex, Ohm, 2 x Ber, Jah (cont); 2 x Ist, Gul, Ohm, Lo, Cham, Zod (SC); Pul, Um, Vex (monst)

% cont = 1.7% – an average probability to get a rune from a container

% (Pul+) cont = 0.037% – an average probability to get a HR from a container

% SC = 10.2% – an average probability to get a rune from a SpecialChest

% (Pul+) SC = 0.36% – an average probability to get a HR from a SpecialChest

% Unique = 1.9% (drops) or 2.5% (unit) – an average probability to get a Unique drop/unit from a SpecialChest


The found data is fairly consistent for both Abaddon and Infernal Pit. Indeed, the “productivity” of regular containers appeared to be very similar in both areas with almost equal numbers of rune dropping probabilities. However, the SpecialChest behavior was drastically different. Despite the probability of seeing a rune being dropped from a SC is the same for both mini-levels (10.2-10.5%), the amount of HRs in Infernal Pit SC equaled 7 (out of 1960 runs) while in Abaddon NO SC HRs were dropped in 1240 runs.

These findings allowed me to suggest that the chances of EACH HR being dropped from the SC in Infernal Pit are significantly HIGHER than those of Abaddon (and, most likely, of Pit of Acheron also), thus making Infernal Pit the best choice for Zod hunting. In other words, IP SC should provide runes of higher quality than Abaddon or PoA SCs within the same total amount of dropped runes. Analogously, even if mathematically a total of 10 regular containers in Act 5 (Hell) provide the same probability of dropping a HR (10 x 0.037%) as compared to one SC in Infernal Pit (0.36%), the probability of encountering a “better” HR is higher in Infernal Pit SpecialChest, as I can judge from my personal experience.

See for yourself, in less than 2000 runs IP SC has provided me with 2 x Ist, Gul, Ohm, Lo, Cham, and finally Zod (run # 1960). What monster killing or other chest-opening could be comparable to this??? My guess is that this SpecialChest has initially been programmed to have the highest chances of top rune generation.

This suggestion might be easily checked by posting here the obtained results for series of runs to IP SC by different players. So, for those who are still interested in Diablo research (like myself) or those who are specifically looking for HRs to make a runeword – testing this hypothesis might be a way to go. If it proves to be true then we discover the most sufficient way to get a HR in D2 with some exciting findings appearing below in this thread.

As for me, the last thing in this game I had been looking for was a Zod rune. Unfortunately, the capability of cubing it up does not even closely compensate all that excitement of seeing it being dropped. So, my purpose was to discover the most effective way of finding it in order not to spend my lifetime doing so. Thus, my Infernal Pit research has paid me back in both ways, and just after 1960 runs…:

Now the search is over. Thanks for playing!
_______________________________________________________________________________

Several tips on the Infernal Pit runs:

1. If you decide to go for the SC specifically (let’s call it a “short” way) then find (roll up) the most convenient map! You need a portal to IP in a close proximity to the WP in Frozen Tundra.

2. Teleport is a must-have! If you are not a master of this ability use the speed increasing run/walk equipment. The best time-saving option in this case will be running and inevitably fighting on Players Settings 1, switching EVERY time to PS = 7 right before you open the SpecialChest. And then back to PS = 1 again and again. Until you get a Ber and a Jah for your Enigma.

3. If you decide NOT to clear junk from the SC and depend only on your monitoring capabilities of detecting that orange flash, my estimation is that you will miss about 2-3 runes per 1000 runs. However, this strategy will save up your time greatly. Together with 1. and 2. you probably could squeeze 1 run in 20 sec, which will add up to 1000 runs in about 6 hours. Not bad at all!

4. If you prefer a “long” approach (in case, for example, if the monotonic repetitions are too depressing for your mind, or you don’t want to lose your map, or by any other reason) then find an optimal pathway to the IP portal in FT in order to effortlessly open as many containers as possible. My calculations show that 10 regular containers (chests, corpses, stashes, etc.) in total have equal chances of providing a HR as 1 IP SC. Same applies for the path inside IP. In this case the time-consuming clearance of the SC junk is justified (the price for missing a single SC rune during “long runs” is much higher).

N.B. All my runs were, obviously, “long” ones. I cannot guarantee that the preprogrammed probability of a HR drop from the IP SC is a constant. If it’s not, then a “short” way might influence it in the same way as, for instance, the route to an armor stand in LK had led to different (but repetitive) armor being dropped from it. Or, in my runs I could control an appearance of a mana shrine in a certain place in FT on a regular basis by choosing a specific route to it. Is it possible to increase the chances of a HR drop by a “certain behavior”? Unfortunately, there are not enough data in my hands to make any conclusion about that at this point. But it will be cosmically interesting to get more info on this subject!

5. Delirium headgear should be accompanied by the Beast second weapon (which provides lycanthropy). Otherwise, you will get stuck with no TP for a minute from time to time.

6. I suggest you make a screenshot of any Pul+ SC finding. You never know what other “coincidences” might be discovered for D2 SpecialChests.


Good luck to you all!



FT – Frozen Tundra

HR – High Rune

IP – Infernal Pit

LK – Lower Kurast

MF – Magic Find

PoA – Pit of Acheron

PS – Player Settings

RoF – River of Flame

SC – SpecialChest

TP – teleport

WP – Waypoint
_______________________________________________________________________________

Screenshots of 7 HRs (2 x Ist, Gul, Ohm, Lo, Cham, and Zod) found in the Infernal Pit SpecialChest:

View attachment 9091

View attachment 9092

View attachment 9093

View attachment 9094

View attachment 9095

View attachment 9096

View attachment 9097

 

[pharphis]

Too few runs to draw any real statistical comparison between the two areas. Also, since the chests very likely aren't truncated to 65k drops (instead 4 billion) they may not be any more efficient than popping objects in general (though they probably are slightly more efficient). With 4 billion patterns, have (random number choice) 10 zods +/- 2 and choosing the chest with slightly more doesn't give a real advantage (especially since we don't know how many zod patterns there are).

Still cool to see a zod, though

I'm surprised you didn't try the chest in nihl's halls

 

[T72on1]

IIrc these chests have been discussed before, and the conclusion was more or less as @pharphis says.

Nevertheless @vmiguli, all those juicy HRs definitely tickled my interest. Certainly because we are talking about 7 HRs from the SC alone, in 2000 runs. I mean, RNG and drops are what they are. They can sometimes do very silly things, as I have been able to judge myself in the first round of the RFL. In that respect the Zod from the IP SC is likely to be once in a lifetime luck. But even then it would be extremely messed up RNG to get less than let's say 3 HRs in the next 2000 runs.

So I'm definitely very tempted to give this a try myself. Only problem is when, with RFL round 2 starting soon and a Sept character to finish.

 

[vmiguli]

My point is that the IP SC was preprogrammed to have better chances of HR generation. So, there are not "10 zods +/- 2" (out of 4 billion scenarios), but many more, if my theory is correct. In other words, the final amount of HR patterns is in direct dependancy with the programmed probability of the appearance of each rune. I cannot state that the amount of Cham patterns exceeds the amount Zod ones, after all it is RNG to decide, but I presume that the total number of HR patterns for the SC are by orders of magnitude higher as compared to the regular containers. That is what I see with the IP runs.
2000 runs is not enough to postulate, I agree; that is why I welcome everybody to share their experience on the subject.

 

[pharphis]

My point is that the IP SC was preprogrammed to have better chances of HR generation. So, there are not "10 zods +/- 2" (out of 4 billion scenarios), but many more, if my theory is correct. In other words, the final amount of HR patterns is in direct dependancy with the programmed probability of the appearance of each rune. I cannot state that the amount of Cham patterns exceeds the amount Zod ones, after all it is RNG to decide, but I presume that the total number of HR patterns for the SC are by orders of magnitude higher as compared to the regular containers. That is what I see with the IP runs.
2000 runs is not enough to postulate, I agree; that is why I welcome everybody to share their experience on the subject.

Pretty sure they're all the same object type (in this case the kind of gold chest that drops sets and uniques) and so there is no pre-programming of one over the other. Alvl probably affects what items could be dropped but since it's Av they should all drop the same runes.

@GalaXyHaXz checked about a year ago for me if they were the same object type, iirc. There's not much reason to believe that these are that much more efficient than other objects if they all have 4 billion drop patterns. The reason I say this is you could pop 20 objects in the time it takes to pop just one of those. Of course each sparkly chest is better on average.. but for run-times...?

The reason LK is efficient is
1) 6 chests per run and
2) truncated patterns (66k patterns) = more variation on expected HR/chest, so there's a high likelihood in each version (ie 1.13 vs 1.12) to have different optimal p settings as well as a reasonable chance for significantly abnormal distribution rates. With 4 billion patterns, the effort it takes to find if any given setting is beyond the usual density is enormous (impractical even if we all had bots running for months) and the likelihood of it being significantly different is smaller (much bigger pool of patterns = closer to expect drop ratios)

I may not be writing clearly so I hope that makes sense. My pretend number of 10 +/- 2 is of course not reasonable. It would be more like 1000 +/- 100 or probably more (too lazy to do the math). Compare that to LK having for ex/ an "expected" value of say 5 surs. 11 surs is twice as much as we would expect but that quite of variance isn't too insane for small-ish number of events.

 

[maxicek]

From Ulrik’s Rune Finder Guide:

“Although it offers the best chance in the game, an Act 5 (H) Sparkly Chest (popped at P8) would require 625,000 runs on average before you see a Zod. Kind of depressing ain't it?
I wish you the best of luck. ”

Even allowing for better chance of high runes since 1.13, I don’t think it is worth it.

Although I am slightly interested in how quickly they can be run - The runs on the video are way too slow and I don’t think the build choice / setup are optimal.

 

[vmiguli]

Although I am slightly interested in how quickly they can be run - The runs on the video are way too slow and I don’t think the build choice / setup are optimal.

I rerolled my map yesterday just to prove my point in the post for the "short" runs. From the SECOND reroll I got a map with all three mini-level portals being in close proximity to the corresponding WP (10 TPs / Ab; 4 TPs / PoA; 4 TPs / IP). On my old PC with a short practice I have managed to get numbers as low as 30 sec / Ab; 20-25 sec / PoA; 25-30 sec / IP for the full circle. Thus, I am pretty sure in my estimation of 20 sec per run for a skilled gamer with a more efficient character. Even 15-20 sec per an average run might be reported.

As for probability numbers, after 1.13 the whole picture has changed dramatically. I don't think it is worth relying on previous generalizations in evaluating probabilities in the current patch. But even if the numbers stay unchanged, it does not matter - it is not what I am trying to emphasize in my posting. I try to keep everybody's attention to the IP SC because it "behaves" differently! It is NOT like any others you see around in D2. I am pretty sure this is the case, although I don't know an algorithm for the rune generation in detail. I just came by something unusual with a huge potential in it and described it. And when it will be checked and, hopefully, confirmed - then we will see numbers of justified explanations for such a behavior.

 

[Grisu]

Thanks for sharing your findings with us! Certainly some interesting ideas and some especially good finds!

Yes, 1.13 changed drop statistics for Diablo 2, but not how objects behave in general. Also, object type didn't change either as far as I am aware. Nowadays, I don't have enough time at hand to dig in the game code though...

also, statistics can play tricks one humans' minds.

Back in 2007/2008 pre 1.13 and again in 2009/2010 (1.13. beta), jjscud did lots of research (including simulating millions of runs for research purposes). With that and the game code it is as good as certain, that regular Sparkly chests do not have seeds, but 4 billion possibilities. Since Infernal Pit always spawns in the Frozen Tundra and the superchest there is "guaranteed", it is certainly a sparkly chest, not a special chest with seeded drops . Special chests with seeded drops are for some reason scarce beyond Act III - i.e., there are no regular appearing special chests, only irregular apperaing ones in the RoF (Act IV) and in the Frigid Highlands (the one near the camp fire wit crumbling walls around it IIRC) (Act V). As pharphis mentionend, the chests in Infernal Pits/Abaddon etc. have the same object type - jjscud checked this thouroughly, thus almost missing that there wer irregular appearing special chests in Act IV and V as well, initially assuming there were none beyond Act III.

Also I am quite certain that the superchest in Infernal Pit did drop Sets and Uniques for me, very much confirming it is indeed a sparkly chest and not a special chest.

As much as I like to believe that the drop patterns are seeded, I very much doubt it

But I will be gladly proven wrong!

EDIT: Here is the link to the thread where pharphis asked about Act V superchests.
https://www.purediablo.com/forums/threads/act-5-superchest-questions.837347/
Some ideas there maybe
Oh, and this thread: https://www.purediablo.com/forums/threads/lk-1-13beta-special-chest-drops.751457/page-6
esp. posts 113 and 114. There is much information about how superchests (both sparkly and special) behave.

EDIT2: On a second thought though: jjscud confirmed that special chests in Act V can indeed be static and do look sparkly and that there is no way to tell if they drop like "regular" sparkly chests or "LK-like" special chests...still I'm quite certain, that the chest in the Infernal Pit does drop S/U items, eliminating the possibility of being a special chest like the LK ones.

They are actually static, at least as static as any others. But yes, the problem persists that there are no grouped special chests in act 5. The two I have used the most are the one in Frigid Highlands and the one in the Halls of Pain.

The tricky thing with act 5 special chests is there the same graphic is used for sparkly chests and special chests. The only way to tell the difference is to pop them a hundred times or so and see only blues and rares.

Did you have any S/U drop from the superchest in the Infernal Pit, @vmiguli?

 

[vmiguli]

Thank you for your comment! It is even better now

So, putting aside an assumption that I got misled by this lucky High-Runic shower (which will be eventually proven by the rune hunters), let us consider this situation logically.

1. The IP SC is, in fact, a Sparkling Chest with no seeds and just 4 billion combinations.

2. The probability of any rune drop from the IP (Abaddon, PoA) SC I evaluate to be around 10%, 5-6 times higher than of a regular container. The probability of a HR drop from the IP SC I evaluate 10 times higher than of a regular container. (Based on just 2000 runs, but that is what I have, sorry, folks). This is the first observation for my conclusion: different chests give different probabilities of HRs in the same area.

3. IIRC, when the “Rune” object is chosen it goes through a cycle from Zod to El to generate the dropping rune from the list. The probabilities of a rune to be chosen are obviously "skyrocketly" increasing throughout this sequence. No Zod -> Eld rune has fit – you see an El.

4. Now, my hypothesis is that for each Sparkling Chest there might be an individual preprogrammed probability for that sequence. Like in LK, the chances for Zod, Cham and Jah equal zero. It has been programmed to start each cycle with a Ber. I don’t see why the cycle probabilities might not be INITIALLY different from Chest to Chest.

5. So, my suggestion is that the probabilities of HRs from the IP SC (sparkly chest) are HIGHER than from the other similar (sparkling) chests, like in Abaddon and PoA. That is how I see it right now.

EDIT 1. Just saw your links, Grisu. Yes, they drop Set and Unique items, and mixes of Magic and Rare. They are not SuperChests (like in LK) for sure. That is why I called them SpecialChests. No sparkling shining, though, like in Pit 2.
But it does not matter for my hypothesis .

 

[Brak]

Congrats on your finds and thanks a lot for sharing the detailed info!
but...

2. The probability of any rune drop from the IP (Abaddon, PoA) SC I evaluate to be around 10%, 5-6 times higher than of a regular container. The probability of a HR drop from the IP SC I evaluate 10 times higher than of a regular container. (Based on just 2000 runs, but that is what I have, sorry, folks). This is the first observation for my conclusion: different chests give different probabilities of HRs in the same area.

its not 5-6 times more likely, its just dumping out 5-6 times more stuff

and I dont think you understand how undersampled 2000 instances is against 4 billion possibilities. You just don't have a representative enough sample to conclude something that goes against a substantial amount of evidence to the contrary. I'm fairly sure that sparkly chests operate as poppables do and chests were supposed to. They should drop from similar tables to creeps in the same alvl area.

 

[Grisu]

3. IIRC, when the “Rune” object is chosen it goes through a cycle from Zod to El to generate the dropping rune from the list. The probabilities of a rune to be chosen are obviously "skyrocketly" increasing throughout this sequence. No Zod -> Eld rune has fit – you see an El.

4. Now, my hypothesis is that for each Sparkling Chest there might be an individual preprogrammed probability for that sequence. Like in LK, the chances for Zod, Cham and Jah equal zero. It has been programmed to start each cycle with a Ber. I don’t see why the cycle probabilities might not be INITIALLY different from Chest to Chest.

That's not how item generation works for chests though - Just a few things regarding item generation/chest drops:

No Rune above Ber in Act 3 is due to the restriction of which runes can drop in which Act: It's up to Ber in Act 3 Hell:
Hell difficulty:
Act 1: Vex
Act 2: Lo
Act 3: Ber
Act 4: Cham
Act 5: Zod
So that's right so far.

However, there are no different TCs used. The higher chance for high runes is based on the truncated seeds with 16 bits instead of 32 bits, not because it uses higher base runes to start to drop from. Once the "rune sequence" (Runes 17 TC) is chosen when an item is generated as a chest drop, it cycles through exactly the same sequences:

TC      Picks  Item1 Prob1 Item2 Prob2   Item3   Prob3
Runes 1     1  r01    3    r02      2
Runes 2     1  r03    3    r04      2    Runes 1   2
Runes 3     1  r05    3    r06      2    Runes 2   5
Runes 4     1  r07    3    r08      2    Runes 3   7
Runes 5     1  r09    3    r10      2    Runes 4   12
Runes 6     1  r11    3    r12      2    Runes 5   22
Runes 7     1  r13    3    r14      2    Runes 6   45
Runes 8     1  r15    3    r16      2    Runes 7   90
Runes 9     1  r17    3    r18      2    Runes 8   180
Runes 10    1  r19    3    r20      2    Runes 9   360
Runes 11    1  r21    3    r22      2    Runes 10  720
Runes 12    1  r23    3    r24      2    Runes 11  1066
Runes 13    1  r25    3    r26      2    Runes 12  1519
Runes 14    1  r27    3    r28      2    Runes 13  2170
Runes 15    1  r29    3    r30      2    Runes 14  2941
Runes 16    1  r31    3    r32      1    Runes 15  3957
Runes 17    1  r33    1    Runes 16 5170

Infernal Pit has an area ID of 127, thus it drops from the highest chest TC possible, same as the chest in Abaddon: Act 5 (H) Chest C. Then, if Act 5 (H) Good TC is chosen (2/142), it checks if runes 17 is chosen (14/130). Then the sequence above starts.

So it doesn't per se start with a different cycle, it just fails to create a Zod, Cham or Jah in LK and checks the next "event" (runes 16 or r32 or runes 15)./edit: this is wrong, see below

What you suggest is, that for certain chests the game doesn't take it's data from treasureclassex.txt is when it checks which items drop, which is very unlikely. Yes, in theory a chest could have a different starting TV somehow (hardcoded?) but it would still jump to one of the tunes TC I showed above. Else the chest would have her very own treasureclass table and not use treasureclassex at all...

Here's by the way the explanation, why LK differs from Act V:

Now for some technical info I dug up last night.

I wanted to cover my bases before I went nuts on the Act 5 Special Chests so I dug up some facts from Jarulf about the special chests. All Act 1 - Act 4 Special chests are object type 580 or 579? (mostly 580). While most objects have descriptions in the objects.txt file, objects 573+ (including 580) are hardcoded somewhere.

No other objects appear to have this limited seed so it appears that its this hardcoded function that has the 16 bit seed for drops. I finally found an old post by Jarulf where he confirmed that all chests have a 16 seed BUT he later states that the drops appear to be governed by the game seed. This doesn't appear to be the case for the hardcoded object 580 drops and so we get the 2^16 (65536) limited drops.

So last night, I dug up a d2 map editor and a mpq extractor and started examining some maps. First I checked LK to make sure I was looking at the right thing, found the campfire area and sure enough there were the type 580 objects.

Then i dug up some of the outdoor act 5 map areas and found one of the chests there that sparkle. It turns out its type 455. A quick check in the objects.txt file and sure enough, type 455 is a special chests. So that's confirmed, at least some of the sparkling chests in act 5 are actually special chests.

Here's the catch, type 455 is well defined in the objects.txt file and uses standard initialization and operation functions. All evidence points to these functions drawing from a proper 32 bit seed and so, no limited drops and no unusual drop odds on HRs. Its not a guarantee, but that's how it appears.

Edit; sorry, realised my mistake: in LK, it doesn't fail to drop Zod and then checks again then next (lower) runes TC, but can't choose from runes 17 TC at all. That's because of the area ID which determines the initial TC to choose from,which is of course not Act 5 (H) chest C. I Have to look up the corresponding TC tomorrow though, I am writing from my phone... Still, the rest holds true - the cycles and probabilities are the same once you are in the rune TCs.

 

[vmiguli]

Infernal Pit has an area ID of 127, thus it drops from the highest chest TC possible, same as the chest in Abaddon: Act 5 (H) Chest C. Then, if Act 5 (H) Good TC is chosen (2/142), it checks if runes 17 is chosen (14/130). Then the sequence above starts.

What you suggest is, that for certain chests the game doesn't take it's data from treasureclassex.txt is when it checks which items drop, which is very unlikely. Yes, in theory a chest could have a different starting TV somehow (hardcoded?) but it would still jump to one of the tunes TC I showed above. Else the chest would have her very own treasureclass table and not use treasureclassex at all...

jjscud:
Here's the catch, type 455 is well defined in the objects.txt file and uses standard initialization and operation functions. All evidence points to these functions drawing from a proper 32 bit seed and so, no limited drops and no unusual drop odds on HRs. Its not a guarantee, but that's how it appears.

Thank you, Grisu! Now we are getting closer
I do not suggest treasureclassex.txt is different for the IP SC. I think it goes by the same algorithm as for any other Sparkling chest.
What I suggest is "when it checks", as you say it, there is another hardcoded parameter that is taken into consideration (which I call a HIGHER probability of a HR drop for the IP SC). In other words, not only the numbers from the table (like 5170, 3957, 2941, etc.) define the probability of a certain TC, but there also might be another preset parameter (that is different for each chest) that is being involved into the mathematical procedure of "checking it".
We need to know how the game makes this yes/no check for each runic TC (from Runes 17 to Runes 1) and see if the values for this procedure are identical for EACH sparkling chest. If that is the case, it will prove me wrong, since I suggest that these other preprogrammed values (A) exist; (B) are different for each chest. And according to my assumption, in case of IP SC they are HIGHER than for the other chests.

 

[pharphis]

I think this information would have been realized a decade ago if it were true

 

[Grisu]

I really appreciate your enthusiasm about this and your hope of finding something awesome and new in D2 After all I was one of the people to suggest the RoF Special chest might be worth running for Cham and Jah some years ago, despite all skepticism.

Still your explanation seems too farfetched for me. I will probably still do some runs just for good measure and in the hope to get lucky, but your explanation would mean an entirely different feature/bug than ever experienced before: AFAIK the object type for the sparkling chests in the Hell areas (Infernal Pit and the like) should be the same, so there should be no reason for them to behave differently. To be entirely sure, one would have to use a map editor and MPQ extractor to see if object types truly are the same though...

Assuming they are, than indeed there would be the need of some hardcoded parameter, which is very unlikely, because Act V was added much later and patches usually didn't change anything via hardcoding. For example the rune drop probability change: It was changed in the treasureclassex, not hardcoded. Here are the rune TC probabilities from 1.11 and 1.07 so that you can compare them to the 1.13 ones above.

1.11
Item1  Prob1  Item2  Prob2 Item3 Prob3
----------------------------------------------------
r01    3    r02    2     
r03    3    r04    2    Runes 1     9
r05    3    r06    2    Runes 2     20
r07    3    r08    2    Runes 3     30
r09    3    r10    2    Runes 4     50
r11    3    r12    2    Runes 5     90
r13    3    r14    2    Runes 6     180
r15    3    r16    2    Runes 7     360
r17    3    r18    2    Runes 8     720
r19    3    r20    2    Runes 9     1440
r21    3    r22    2    Runes 10    2880
r23    3    r24    2    Runes 11    5040
r25    3    r26    2    Runes 12    8820
r27    3    r28    2    Runes 13    15435
r29    3    r30    2    Runes 14    27011
r31    3    r32    2    Runes 15    47269
r33    1    Runes 16    82721     

1.07
Item1 Prob1 Item2 Prob2 Item3 Prob3
----------------------------------------------------
r01    3    r02    2     
r03    3    r04    2    Runes 1     9
r05    3    r06    2    Runes 2     20
r07    3    r08    2    Runes 3     30
r09    3    r10    2    Runes 4     50
r11    3    r12    2    Runes 5     90
r13    3    r14    2    Runes 6     170
r15    3    r16    2    Runes 7     330
r17    3    r18    2    Runes 8     650
r19    3    r20    2    Runes 9     1290
r21    3    r22    2    Runes 10    2570
r23    3    r24    2    Runes 11    5130
r25    3    r26    2    Runes 12    10250
r27    3    r28    2    Runes 13    20340
r29    3    r30    2    Runes 14    40970
r31    3    r32    2    Runes 15    81880
r33    1    Runes 16    32770

It doesn't make any sense that blizzard should make changes to drop odds in the txt files AND hardcode additional changes. That's so much more workload that I don't think that happened.

So the only possibility would be, that - when creating the expansion - Blizzard deliberately hardcoded some additional modifiers for drops without mentioning that in the change logs/expansion logs and then never changed those again. Those modifiers however could not have been linked to the object type (assuming it is the same), but would have been linked to something else, like area ID. That's a lot of effort to make for Blizzard when they could have so much easier changed the treasureclassex file...

I'm not a programmer though (FAAAR from it ), so take this with a grain of salt

EDIT: Funny to see that Zod is actually more likely to drop in 1.07 than in 1.11, but other HRs are less likely to drop...

 

[vmiguli]

Well, there are only two logical explanations, as I see it right now.

1. IP SC is fully identical in it's behavior to other sparkling chests in Act 5 (H). I got a lucky streak of 2000 runs out of 4 billion combinations and have made conclusions that are not valid. "Forget about it" (c) DB

2. My observations are justified and require an explanation. The only explanation that is not rebuffed by programming, mathematics and common sense is a hypothesis that each chest has an affined value that is being considered (together with treasureclassex probabilities) while the game chooses a TC for an object (a rune, in this case). Such a value can easily go with each chest, in the patch or in the original version, and is being included in a mathematical procedure (which is hardcoded) of a yes/no answer to each runic TC when chosen.

There are two ways, that come to mind, to find a solution of what option is true.
1. Make lots of runs by multiple players to the IP SC and compare the obtained data with any other chosen SC from Act 5 (H). If my theory is correct the total amount of runes should be the same (+-) for both chests, however, the amount of HRs will prevail in IP.
2. Get to the bone of the mathematical operation of yes/no checking for each runic TC. Are there any other values (except for those from the table) being used? What are these values? Is it different for each chest? What is a crucial procedure of getting a yes/no answer for each TC? I don't have answers for those, I am not a programmer either.

Just thinking logically...

 

[GalaXyHaXz]

For farming runes, it's either p7 LK, p5 Cows, or p3 Travincal. End of story. @Gripphon did the math last year and each area has a slightly different chance for each rune, Cows being the best for Zod but LK being the best for Sur/Ber. Travincal for Jah/Cham.

The gold chests in A5 do not have truncated drop patterns because the "always magic" flag is not set. They do however have their own special drop system, which can make them potentially worth it in rare cases.

The chest has two different types of drops. The first type is the "regular" drop, where the chest is hardcoded to drop a few potions and gold pieces (ever wondered why you see minor mana potions and 1 gold piece on hell??). Along with the hardcoded junk, the "Chest C" class is chosen for the regular drop. This part takes on the assigned level to the chest and not the area level. Act 5 Chest C is level 85.

The second drop is the "super" drop. I haven't checked the code for the exact %, but something to the extent of: 5% rares, 3% sets, 1% uniques. The chest will not drop the hardcoded junk and all items in the chest will be of that quality.

What I would recommend doing, as suggested awhile back, is running Snapchip instead and popping the gold chest behind him. Snapchip can drop almost everything and has better Unique odds than Pindleskin, coupled with a quick pop of his chest for two picks in the Chest C group or all uniques. The only difficulty is making a dual element build and rolling a map where he's right by the WP.

That's the only gold chest worth running IMO. 1.10 made the MF system a lot more viable throughout the game, so you can actually play the various level 85 areas instead of milking the same target. In older patches where things were broken however, such as 1.08, running gold chests could be a viable way to get Windforce, as described in the 1.08 thread.

Edit: and in pre-1.09 the gold chests had very high levels, allowing them to drop level 93+ charms for rolling

 

[pharphis]

I believe it's 2% for unique ie 1/50. But depending on success/fail for item types it can drop 0-many of them. I think the upper limit was like 12 uniques or something insane

 

[Shuffleblast]

198/9, if only it was eth as well (ss cleaned of gold and potions).

View attachment 9127

The drops do seem comparable to a LK chest and are better than some other sparkly chests in a5, like the one in glacial trail and frigid highlands (are there different kinds of sparkly chests?). I can see it dropping a zod, even if just because of how much stuff it drops on /p7.

The difference between abbadon and infernal pit chests doesn't seem statistically relevant though. It's 0 hrs/130 runes and 7 hrs/200 runes (or ~4,55 hrs/130 runes). The chances of this happening even if both chests can drop the same are too high. But anyway, I hope some people do some more runs, I've had enough of racks and chests for a while.

 

[vmiguli]

198/9, if only it was eth as well (ss cleaned of gold and potions).

The drops do seem comparable to a LK chest and are better than some other sparkly chests in a5, like the one in glacial trail and frigid highlands (are there different kinds of sparkly chests?). I can see it dropping a zod, even if just because of how much stuff it drops on /p7.

The difference between abbadon and infernal pit chests doesn't seem statistically relevant though. It's 0 hrs/130 runes and 7 hrs/200 runes (or ~4,55 hrs/130 runes). The chances of this happening even if both chests can drop the same are too high. But anyway, I hope some people do some more runs, I've had enough of racks and chests for a while.

Shaffleblast, what is 198/9, I can not interpret for sure: runs, runes, uniques?

Your statement that the IP SC is "better than some other sparkly chests in a5" contradicts the stereotypical belief that was expressed in this thread (that all sparkly chests are identical in Act 5) and indirectly supports my theory of extraordinary behavior of this particular chest.
Thanks for sharing!

 

[Grisu]

Shaffleblast, what is 198/9, I can not interpret for sure: runs, runes, uniques?

It's the variable stats value for Titan's Revenge: 198% Enhanced Damage, 9% Life leech

Your statement that the IP SC is "better than some other sparkly chests in a5" contradicts the stereotypical belief that was expressed in this thread (that all sparkly chests are identical in Act 5) and indirectly supports my theory of extraordinary behavior of this particular chest.

Well, I (and others) simply tried to point out the problems of your theory. I certainly don't want to impress my opinion on you. But with over 4 billion possible chest drops, one should be aware of the fallacy of small numbers Shuffleblast's statement isn't backed up by any statistics, but simply by feeling, which is the common cause of most fallacies. You ignored Shuffleblast's second statement...so I am sorry to ask this, but might it be that you simply want people to agree with you? (Well, don't we all want that in life )

No offense meant, but if you were more interested in people supporting your theory than in actually discussing the likelihood of your theory based on statistics and knowledge about the game, then I'll save me the trouble of thinking about the lieklihood of your theory based on what I know about the game mechanics and spend my time on other things. To emphasize this: no offense meant, I am just not sure about this

Some other thoughts that - from a logical standpoint - that make me hesitant to accept your theory: If there was such a hardcoded parameter, why would Blizzard have implemented it just for this chest? Or just for sparkling chests? Or just for objects? Wouldn't it be safe to assume, that they implemented such parameters globally? One could argue, that some super unique monster in Area A drops better than another super unique monster in Area A because they have different parameters hardcoded which affects their drop probabilities. But then - why would Blizzard have bothered with MPQ files? They could have simply hardcoded all drop probabilities.

I think that running those chests can still be worth it (after all, they can drop rellay cool stuff!), but I doubt it's more effecitve than LK, Travincal, Cows (or HF rushing for that matter...)

The second drop is the "super" drop. I haven't checked the code for the exact %, but something to the extent of: 5% rares, 3% sets, 1% uniques. The chest will not drop the hardcoded junk and all items in the chest will be of that quality.

It's 2%, 4%, 6%,20% for uniques, set, rare, magic. Mind you, not for a unique item, bur for the chosen item to be uniuqe. So if for example "Arrow" or "Emerlad" is chosen, it won't drop a unique item even if it rolled option 1 (unique item). See Warrior of Lights Item Generation tutorial for more information.