Gradual Reduction in MF Effectiveness

[LiquidDamage]

Probably because a low level character checked, and therefore spawned, her inventory, before you did.  Shopkeeper ilvl is equal to the clvl of the checking character + 5.  The inventory then doesnt change until some period of time goes by without checking it, or all players that are in that act go out into hostile areas.

Yes that makes perfect sense.

Let me see a level 90+ character on his own in his private game.

And she sells tiny potions on Hell?

What could that be?


Let me think, thinking, thinking..... no idea :D


In fact I have a pretty good idea what causes it. :P



Regards, Hunky

I have never seen minor potions for sale in a private hell game containing only high level characters. However, every time I check a potion vendor for the first time with a level 1 in hell, they sell minor healing potions.

You never said you were alone in a private game, so I really don't think the sarcasm was necessary.

Furthermore, I am not inclinded to believe you until I hear reports from other people with high level characters in private games getting minor healing potions, as there is a perfectly reasonable explanation: you weren't alone in a private game, and are remembering things incorrectly.

[LiquidDamage]

I do not deny, however, getting blue on more than one occassion which I attributed to a decreasing MF rate since my first few drops seem to always be great (may be failed uniques/sets?). I would like to mention that you say the first actual quest kill of the boss yields only rares or better, yet I seem to recall seeing blue plenty of times on my quest kills (perhaps a figment of my imagination from all this MF psycho-babble?  I wouldn't deny wanting it to be true so unconsciously convincing myself that is what I really saw).  Are you sure this is accurate? 

If you find magic items on a quest drop, they must be failed set items.

Also, if someone who has completed the quest kills your quest boss for you, then you won't get the quest drops, so normal magic items are possible.

If you doubt that quest drops are rares at a minimum, all you need to do to disprove it is have the boss drop a magic item with normal durability when it is killed by someone that hasn't done the quest (preferably in a game where the quest hasn't been completed, in case that bug doesn't apply to all act bosses).

[Crystalion]

When you use the -seed command there is no external time calls of any sort (for the seed part) since it uses the specified seed instead.

As I mentioned, holding -seed and repeatedly starting fresh games with restored (character) saves still results in differences in shrines and quest drops (norm HF tested). Since my understanding of your code reading theories doesn't really explain the observed results, we get my "debugging" desire to either audit the rnd usage or slyly hold the known variables to random seed generation constant, as a means of verifying or falsifying the assertions about the seeds usage, under -seed conditions.

It has always been my habit in debugging to verify "the obvious", when that is simple to do. As you say, when using -seed holding the external timer calls to (say) zero should not have any effect. This is actually precisely why one would do such a test.

It's a pity (for some other testing) that, at present, single player can not replicate exactly the same (initial) game (conditions) via -seed (and restored saves). But not terribly important. Thanks again for the info.

[JustAGuy]

You know, every run I have every completed has been on a map spawned by a character that HAS already killed Mephisto. It appears that my drops have been needlessly poor! The thing is, though, that they do get "worse". Over 100+ runs, it definitely gets "worse", in that perhaps more items fail to generate. Since I have not checked the durability of the yellows and blues, I can't say for sure.

Some more subjective MF run results; in 20 I did yesterday with a newly equipped level 87 hammerdin, doing runs in this order:
1. Sorceress or Hammerdin create game (both have killed Baal).
2. Both characters party up.
3. Sorceress finds Mephisto
4. Hammerdin jumps down and kills him (wearing 430% mf)
5. Hammerdin goes to act 5 and kills Pindleskin, Shenk and Eldritch, not in that order (varies each time to kill the boredom a bit)
6. Rinse and repeat

Just a quick note: the Hammerdin uses level 7 redemption (from items) to clear out the "dead" bodies around Pindleskin's entrance. Most times, if I run around the surrounding area with redemption on, I can clear about 85-90% of the bodies before they get up to distract me.

Now, with that order of running, each game lasts 5-10 minutes.

The same effect happens; "good" drops early, "bad" drops late. "Bad" drops include rares and blues. This is all subjective, since I haven't catalogued anything.

I agree that the burden of proof is on us "runners". So, I intend to do some testing. Just some raw numbers, for everyone's pleasure. Mmmm, numbers. I'm talking detail here -- every item along with durabilities dropped will be captured in a spreadsheet, and the results made available on this thread. I'll try to make the results available soon, but I'm definitely not going to be able to start this for a few days (my job has me driving back and fourth from Toronto all the time).

Here's the order I will follow:
1. Character that has not completed act 3 will create game.
2. Sorceress and Hammerdin will join game (the other character will vamoose).
3. Sorceress will find Mephisto
4. Hammerdin will kill him.
5. Write down all results (this is where two computers really helps, spreadsheet on one, diablo on the other)
6. Repeat

I'll do them in 25 run intervals, tracking item class (unique,rare,magic), item types, durabilities, time to complete run, and direction of the act 3 durance entrance. There was another thread that showed a picture of the likely direction of the durance level 3 entrance. For example, if the durance 2 way point spawns with the area leading up to it facing the north west, the area surrounding the durance 3 entrance will have a U-shaped pattern, and most likely be to the south west of your current position (although, sometimes the map creation logic puts the U-shape a bit to the north, but that's not a big deal).

Stats to follow in a few days. Maybe I can get that damned Herald of Zakarum finally...

Side note: My thread has hit the front page! *pats own back*... too bad I've been absent for the last few days.

[Ruvanal]

Here's the order I will follow:
1. Character that has not completed act 3 will create game.
2. Sorceress and Hammerdin will join game (the other character will vamoose).
3. Sorceress will find Mephisto
4. Hammerdin will kill him.
5. Write down all results (this is where two computers really helps, spreadsheet on one, diablo on the other)
6. Repeat

This will result in the drops using the non-quest drops of Mephisto. No different than if you had created the game with the Sorceress or Hammerdin. Is this what you are really trying to check?

Getting the quest drop is based on the quest state of the character that is killing the monster in question, not on the quest state of the character that made the game. To get a consistent run of quest drops from Mephisto, you would want to:

1. Character that has not completed act 3 will create game. i.e. probably the Sorceress
2. The character that has not killed Mephisto for quest credit will join the game.
3. Sorceress will find Mephisto
4. "The character that has not killed Mephisto for quest credit" will kill him.
5. Write down all results (this is where two computers really helps, spreadsheet on one, diablo on the other)
6. Repeat

The above will end yielding the quest quality drops. Since the game will not allow the credit to be given for killing Mephisto (due to the Sorceress making the game), the "new" character will be able to repeat the above time after time. Since the "new" character has its quest state for doing the killing of Mephisto in the uncompleted state the game will use this flag condition to determine that it should drop using the quest quality drop TC instead of the normal TC.


Based on these last responses of JustAGuy and MEAT, it looks like both were basing their claims on misunderstood information as to which TC is getting used under different conditions and just how the game handles determining the magical quality of the items (in particular that a failed set item is dropped as a x2 durability magic item).

Any others consider this 'Myth Busted'?
I do.

[HunkyDory]

You never said you were alone in a private game, so I really don't think the sarcasm was necessary.

Furthermore, I am not inclinded to believe you until I hear reports from other people with high level characters in private games getting minor healing potions, as there is a perfectly reasonable explanation: you weren't alone in a private game, and are remembering things incorrectly.

and the Earth is flat right?


Those are the facts, I was alone in an private game and no other character than me had visited the vendor.

Now when you are in a Hell game and you need Life Potions don't you think you will notice what kind of potions she sells?

And don't accuse me of lying please.
Obviously you do not know me or you would not say that.

I would never had thought of problems with the game inititialision if the above had not occured.

The fact that you never saw this does not proof that it does not exist.

Isolde only once answered to my questions about this and that was a very short reply which gave me the impression he did not want to talk about it as in if it was part of something they cooked up and did not want to become public knowledge.

In my opinion it is part of some anti-bot scheme on which drops are set to crap or the initialisation that went wrong.

Jarulf made some very interesting remarks about how the game is initialised.


Regards, Hunky

P.S.
Forgot, this happens in private games and public games.
But in public games you notice it and think about the level 1 explanation.
Only when you see it happening in private games you start thinking about it for serious.

I saw it happen a lot in 1.09, in 1.10 I saw it too but less frequent when doing Eldritch runs with my level 93 Amazon and the last month I have not seen it at all since I level my 89 Baba very slowly in long games.

This was/is on USEAST Ladder games, Ama is SC, Baba is HC.

[HunkyDory]

Jarulf,

I read your explanation about the seed thing.

Do not understand all of it yet but a couple of questions.


Is the number 666 in the Hiseed AND or ORed somewhere in the process.

Apart from being a fun number there is also the possibility to manipulate that particular byte to make the number positive or negative.

Just a tech question what bit signifies plus or minus on the Intel archictecture?

The most significant bit of the most signficant byte?


Last question :

How does the game know to make a Normal, Nightmare or Hell game when using the seed thing?

Or is that not part of this process?


Regards, Hunky

[JustAGuy]

Thanks for the clarification on the bug; I'll use the modified order to do my runs.

As for the myth being busted, not having the right information definitely contributes to possibly drawing wrong conclusions. The point of this thread is to find out whether there is a gradual reduction in MF or not. With more information on how drops are generated, it would appear as though "crap" drops aren't necessarily "crap".

It's not only two users that get poor drops, but many that do the same thing. I won't re-hash the supplied reasons/excuses for each side, but it can't be ruled out that there is some sort of problem with continued runs.

Just because the blues and yellows that are dropping COULD be failed uniques and sets, doesn't mean that they ARE.

[Obi2Kenobi]

Just because the blues and yellows that are dropping COULD be failed uniques and sets, doesn't mean that they ARE.

Unless we get some durability information. Then we can be certain (whichever way).

[Ruvanal]

Unless we get some durability information. Then we can be certain (whichever way).

Any runs like this that I have done in the past has always resulted in magic items that have durabilty, showing with the x2 daraability that indicates that they were failed set items. I have never experienced anything that would suggest there is some 'hidden' manipulation of the MF% that is being used. As I noted in an earlier post, such manipulation would have tendency to show itself in some, shall we say unusual drops that would be rather noticable. Since there have been no reports (or complaints) that match the type of pattern that this would cause, I feel that there is no such system in place.

The types of drops that JustAGuy and MEAT are trying to make a case of do not fit this type of pattern at all, but they do fit just the natural type of random distribution that should expect in this type of random item generation. If no one were to see this type of drop patten occuring at times, that would actually be indicative of a different sort of problem with how the items are being generated. A random system of this nature that does not exhibit certain types of 'streakiness' and psuedo cyclic patterns would not in fact be a real random system. This is an aspect of random systems that many have a geat problem in understanding and coming to grips with, they tend to expect that the result will look rather 'flat' and 'uniform' at any level of examination. With a diverse system like the item dros in this game, any smapling of less than around 3000 to 5000 items from a similar item source (i.e. drops that use the same TC and ilvl) will tend to display many sorts of 'patterns' that are really just the natural consequence of teh way a random system will look when examined in to close of a detail.

[Jarulf]

HunkyDory:
>Is the number 666 in the Hiseed AND or ORed somewhere in the process.

Ehhh, it is SET to that value on initialization. Then it is updated according to the algorithm for the prng I have posted above. From random number generation points of view, all numbers are treated as positive unsigned ones.

>Just a tech question what bit signifies plus or minus on the Intel archictecture?
>The most significant bit of the most signficant byte?

Most significant.
>How does the game know to make a Normal, Nightmare or Hell game when using the seed thing?
>Or is that not part of this process?

That has nothing to do with it. The seed thing is just about geting a random(ish) value(s) into a few places that will then be used for random number generation. There is no other inherant information in the seed values, they have no relation to difficulty or anything such.

For exmaple, selling minor potions instead of major potion has nothing to do with the seeds, since it is not really a random thing (it is part of the few items the vendors always sells and would be controlled by the data files for items (misc.txt I would say) that has a whole bunch of infor for each NPC selling items and what items though can sell and always should sell.

[LiquidDamage]

And don't accuse me of lying please.

I didn't accuse you of lying. I said that I think you are likely remembering things incorrectly.

You hadn't provided your "facts" about being alone in a private game until after I answered your question, when you used them as an excuse to be condescending.

Obviously you do not know me or you would not say that.

No, I don't know you, and now I don't want to, after you jumped into insulting/sarcastic mode when I attempted to provide an answer to a question you asked, and then continued the same behavior after I suggested it was unnecessary.

[Jarulf]

If no one were to see this type of drop patten occuring at times, that would actually be indicative of a different sort of problem with how the items are being generated.  A random system of this nature that does not exhibit certain types of 'streakiness' and psuedo cyclic patterns would not in fact be a real random system.  This is an aspect of random systems that many have a geat problem in understanding and coming to grips with, they tend to expect that the result will look rather 'flat' and 'uniform' at any level of examination.

I like to "quote" an article in a magazine I once read (Scientific American?? can't remmeber which one) about a professor in statistics whoe would in one of the first classes ask the students to go home and toss a coin 200 times and write down the result of each toss. Of course, many students simply skiped the tossing and wrote down their own randomly "appearing" 200 rolls. He usually had NO problem spotting those that cheated in this way. Why? Because they never had long streaks of the same result. Tossing a coind 200 times gives a VERY high chance of getting 5, 6 and even 7 identical results in a row. Most people would rule out such results as not "random enough". The stidents writing down results trying to make them look random, seldom had more than 3 of the same result in a row. Just shows how hard time people in general have to understand randomness and what is normal and what is not.

[Ruvanal]

Jarulf, thank for a good real life example. I have not had a problem of seeing this type of situation due to having done massive amounts of die rolling in games in the past. Some of the final results in the games could at times only be possible if there were some heavily skewed results in that happened a certain points.

An example was a wargame where the two opposing units each had a 1/3 chance of of killing the other unit (simultaneous combat). One of my units (exploring), I mananged to manuver through an enemy territory to only be encountering 1 or maybe 2 opposing units at a time. Result I killed 8 of their ships and explored most of their area before they finally destroyed my 'heavily armed explorer'. If things were very "even" in distribution I should not have made it past a third unit the way most people look at probability.

[whereagles]

Another thing people have a hard time understanding is that the normal result for 200 tosses is not 100 heads, 100 tails. Rather more like 90-110.. (either way, though)

[gekko]

That's the "beauty" of statistics and "randomness." In general, given such odds, you won't get past your third enemy. However, that doesn't mean you will not. You could in theory kill hundreds of the enemy before your scout gets killed. It's highly unlikely, but...

That's why I have a hard time believing people who swear they're 100% sure the drops worsen after 10, 50, 100 drops. Not that I don't believe they are seeing a reduction in quality; rather, I think they're just seeing a "bad run." That couples with the fact that there's more to having a "good" item drop than simply seeing gold or green letters.

Consider running hell baal twice with the following results:
1st run: baal drops 6 uniques -- face of horror, duskdeep, the chieftan, leadcrow, soul harvest and gleamscythe.

2nd run: baal drops 5 normal blue items and IK armor.

Given the rarity of the items (and the cost if you like to trade), the first run probably results in a grunt of disgust while the second results in a *whoop!* of victory. According to the game, the first run was far superior *in terms of the unique quality of each item.* According to just about anyone I know, if hell baal dropped six normal uniques, you had a terrrible run.

That example is obviously contrived and a bit extreme, but it shows how the perception of "quality" can greatly affect how good we think our luck is when hunting for items. This type of perception, coupled with a "good" or "bad" run that statistics tells us is quite normal, is what fosters the rumours surrounding MF.

gekko

[Jeger]

I also noticed (over two years ago) what seemed to be a treasure drop-off with increased runs. I did some quick tests then, that gave inconclusive results.

There are two possibilities if the effect is real. It could be intentional by Blizzard in an attempt to increase treasure drops for casual players while not increasing them as much for botters. I'm not with Jarulf on that possibility, as I think it would be a great idea. The second would be that it could be an unintential result of the random number generator or other part of the program. After reading this thread, I am inclined to write it off as psychological, but I am not as sure as some.

Check all your magic items that drop from him, and if they drop with double durability, then they are failed sets. These should then be considered "good" drops, even though they suck, in terms of testing out your MF.

I think counting failed items as "good" would be a mistake. If the drop-off is real, many possible explanations could result in more failed items and less real uniques.

A thread mentioned the Bishibosh bug, where the problem was the selection of the TC. This of course affects final item drops, and could be believed to be a “run counter effect”. Such bugs can exist of course. That is not the same as a feature with a counter that dimish item quality in runs, which is what this thread primary is about

Maybe a minor quibble here, but it could be the same if the bug occurred more frequently at higher run counts.

[adeyke]

I think counting failed items as "good" would be a mistake. If the drop-off is real, many possible explanations could result in more failed items and less real uniques.

Uniques fail iff that unique has already spawned in that game or if there is no unique version for that item with that ilvl.

Sets fail iff there is no set vesion for that item with that ilvl.

So if you get a double-durability set item, that has to be considered a set item for purposes of drop "goodness", since you would have gotten a set item if it was at all possible.

[HunkyDory]

For exmaple, selling minor potions instead of major potion has nothing to do with the seeds, since it is not really a random thing (it is part of the few items the vendors always sells and would be controlled by the data files for items (misc.txt I would say) that has a whole bunch of infor for each NPC selling items and what items though can sell and always should sell.

Would this indicate that if she sells those minor potions the is set to "Normal" despite being it a Hell game?

That is the only conclusion I can draw from your explanation as the data comes from files.


For that to be possible the game has to think it is in Normal instead of Hell.


Regards, Hunky

[HunkyDory]

Somebody should mod a game in that the Monsters drop one item on Normal, a different one on NM and a Third one on Hell.


And run some tests to see if the correct item for the difficulty gets dropped all the time after having done some runs.


This will rule out the possibility of the drop scheme getting screwed up.

Who is so kind to volunteer for this and do a couple of hundreds of runs?



Regards, Hunky