More on rune drops

[sevenOfDiamonds]

More on rune drops

Following the recent discussion on rune drops I looked into the drop tables for A85 monsters. As the drop is tied to area lvl in Hell they are not suprisingly the same for all monsters (champions and bosses gives better drops of c).

This makes way for very easy calculations of total clear probabilities. The only input we need is the average number of (monsters+minions), champions and bosses in each area.

So fellow SPFers: what are your estimate of #monsters in Pit, AT or other A85s?

 

[Joosh]

Re: More on rune drops

More than 100 but less than 1000

No, but in all seriousness I'll try to keep an account of average amounts in future runs to update here if you want.

I don't think it'll be worth you're time though, I'm a little confused as to what you're trying to accomplish? The best area to run for a given rune VS Time/Runs Done?

Joosh :thumbup:

 

[sevenOfDiamonds]

Re: More on rune drops

That was a pretty accurate estimate

The aim for this project is to add more (statistical) color to the eternal question: "what will net me most loot per unit of time invested?"

The already available drop calculators works only for single targets. This is of course excellent if you are running pindle or meph, but if you want to compare say pindle vs AT you need something different.

Fabian has already done some excellent work on runefarming of cows, coucil etc.
I intend to extend his findings by including other areas, and when time permits include more items as well.

EDIT: The german drop calculator actually has a High Council "TreasureClass" which seems to assume 13 council members instead of 11 but accurate otherwise.

 

[Fabian]

Re: More on rune drops

I considered doing something like this back when I was messing with LK/Cows/Council, but never got around to it. I'd be interested in seeing the results, gl.

 

[Corrupted]

Re: More on rune drops

The problem is it varies so much. The selected monster spawn and the size of the location on your mapseed plays greatly into this number. I think your best statistical bet is to perform a big sample of runs on the given map and note how many runes you average per hour.

 

[sevenOfDiamonds]

Re: More on rune drops

I don't agree to this one.

1) The variance of rune drops in, say AT, is much higher than the variance of #monsters in AT, as the variance of rune drops depends on both drops and #monsters.

2) monster population in given areas are mechanically bounded. You will not see 25 bosses in a single AT run and probably not zero bosses as well. But you could see multiple HRs during a single AT run.

Thus my take is that for a limited number of runs it would be more benefical to observe #monsters than #runes.

 

[Corrupted]

Re: More on rune drops

But bosses are a very minor portion of rune drops. The amount of monsters doesn't vary that much in AT but it does in the Pits and particularly in the WSK areas. Furthermore, you have ghost-type monsters in WSK/RoF/CS which has quadruple odds of dropping runes.

All I'm saying is outside of AT and maybe CS, it should be more difficult to gather a correct average of monsters than a sample of average rune drops. I'm not talking about high runes, but any runes dropped as that'll point you to how statistically worthwhile it is for high runes.

 

[sevenOfDiamonds]

Re: More on rune drops

I finally had some time to work on this.
I did a small spreadsheet which can calculate the joint rune drop probabilities for multiple monster types, i.e. for area runs.

To test this I did a little experiment. I want to test what kind of drops I could expect for ten hours invested into three different projects.
1) P3 council with a 51% horker. Run time: 60 secs.
2) P1 AT with a sorc. She will on average kill 6 bosses, 4 champions and 20 normal monsters. Run time 90 secs.
3) P3 Cows with a sorc. Killing 400 cows in 300 seconds.

Here are my findings:
Project 1

Legend:
P(X>=1): the probability of getting at least a single drop
E[X]: The average (expected) # of drops.
MRDB: The average #runs between drops. I.e. "how many runs it will take on average to get a single drop"
MTBD: The average time (in hours) between drops.

Project 2

Project 3

To give a little idea of the distrubution of drops, I created the following chart based on the first project:

This show the probabilities of getting a given number of Vex runes (blue bars) and the probabilities of getting a given number of Vex+ runes (red bars).

I hope this can contribute to the eternal hunt for better drops.

Comments are very very welcome.


//D7

 

[squall78]

Re: More on rune drops

Great graphs. It's always interesting to see the stats behind rune hunting. So based on the info Council runs have the best rate of return if you can hork, although cows can come very close. Do player settings have diminishing returns after p3? I would think they do because the HP increase in the Council and cows.

 

[Quickdeath]

Re: More on rune drops

The average number of rune drops per clear of the Ancient Tunnels at /P8 is 0.6602 with a std deviation of 0.0279.

This is based on observing on each run the following:
Number of melee/cast/missile monsters (range was 82-109)
Number of champions (4-6)
Number of uniques (3-5)
Number of Type I and II Objects (7-14)
Number of Type I Locked Objects (0-2)
Number of Sparkly Chests (always 1)

The numbers in the above list are not independent, e.g. in a run when you encounter 6 champions, you tend to only encounter 3 uniques.

For other areas, you must factor in the number of Quills, ghosts, Flying Scimitars, Swarms, Cows, special chests, locked special chests, etc. because they all have different drop rates for runes than most monsters.

So based on the info Council runs have the best rate of return if you can hork, although cows can come very close. Do player settings have diminishing returns after p3?

Except cows do have a finite chance of dropping a Zod (about 3% in the example given), whereas the Council offers zero chance of a Zod drop.

 

[sevenOfDiamonds]

Re: More on rune drops

Great graphs. It's always interesting to see the stats behind rune hunting. So based on the info Council runs have the best rate of return if you can hork, although cows can come very close. Do player settings have diminishing returns after p3? I would think they do because the HP increase in the Council and cows.

For you'll get 10% more runes doing p7 Council vs p3 Council.

Must admit that it has turned out that my hork calc is flawed, but when fixed it should show an even larger advantage.

I was actually surprised to see that AT was way way inferior to council or cows.

 

[Quickdeath]

Re: More on rune drops

I was actually surprised to see that AT was way way inferior to council or cows.

Urflig counted the number of cows in cow level by modifing the textfiles in the source code so that each cow dropped a certain gem or rune with 100% probability, and champion cows, unqiue cows and chests in cow level dropped different gem/runes. He would then run the cow level and afterwards count the gems or runes of ech type. In this way he came up with the following results from three trials of cleariing the Cow level:

Cows: 464, 490, 489
Champion cows: 3, 8, 2
Unique cows: 5,0,4
Type I and II objects: 33,29,27
Type I locked objects: 6,5,5

Overall, Urflig calculated that these observations translated into this many rune drops per Cow Level: 3.221, 3.243,3.216 with 8 players

Doing 10 clears in Sewers Level 2 he got an average of 0.3892 rune drops with a std. dev. of 0.0378. (with 8 players)

Urflig also calculated the runes drops for The River of Flame: 1.4042 with a std dev of 0.2072 (with 8 players).

If your methodology ignores the chest drops you would calculate lower values for the above three areas.

Urflig only did one trial for Travincial, but he cleared the entire level. He observed: 98 normal monsters, 2 champions, 4 uniques, 11 council members, 9 Type I and II objects, 1 Locked chest, and a calculated number of runes 0f 0.848 (with 8 players). That is different than your scenario, doing only the council with a horking Barb. Does your hork calculation assume that all hork drops are "normal monster drops" or what?

if you need more info on these cases, let me know.

 

[sevenOfDiamonds]

Re: More on rune drops

@Quickdeath: As mentioned above my hork calc is flawed. I assumed that a hork was equal to get a percentage of chance of getting a "kill-drop". As discussed in another thread hork works in a different way (by affection the drop/nodrop prob directly).

And you are absolutely correct I have not implemented chest drops yet.

The study by Urflig is very interesting. Great concept by using drops as a counter.
The rune drops in Urflig's study I take as any rune, not just Mal+ as in my example.

Taking the easy way by using drop-probs directly instead of doing the drop/nodrop -> TC -> item way was a bad idea. I will look into redoing my calculator the right way this time in order to do horking correctly, chests and other stuff as well.