Item rarity research

[JellySlayer]

It's not really 1/100000. In fact, one would add all the artifacts up and then divide by number of trials, since each artifact can only be had once, to get chance of generating a random artifact.

For surge of power, I think if you wanted to do the statistics properly, you'd also have to weight against the number of random artifacts already generated.

[cromat]

Is it hard to add macros for automatic item generation for DLs 1-50? I'm also curious about how the whole thing was organised (since it's p9). Is it possible to access the framework to do some other tests (obviously with references on derived work)?

Well, it's a bit tricky, but should be also doable.
I consider framework far more useful then just pure results, I'll write the discription and post it soon. In short: "tmux send-keys" for sending input and running adom through tee to collect output, later lots of regular expresions to extract data from huge tee's log and aggregation in GNU R.

Regarding artifact. Here are the turns of "type-identifiable" ones:
2064:A silver key
2747:A silvery two-handed sword
5456:A scorched spear
5561:A rune-covered sling
6016:An ebony spear
9023:A rune-covered long sword
11354:A silvery arrow
12801:A silk scarf
16564:A black long sword
28845:A dirty cloak
36555:A sapphire ring
Here there are only 11/21 artifacts generated, silk scarf is 8th of them, I'll estimate the chance to be 1/1000.

[biomateria]

Haha, thanks so much for putting it up on Google Docs. I'm a lazy mofo and probably wouldn't have ever seen the results. Really interesting, thanks again!

[SirTheta]

I am not sure, JS. Sure, technically, spawn of artifact reduces both total artifact chance and also increases spawn chance of less common items (since an item has been taken out of the pool and that's kind of how ADOM works, to my understanding). I don't see it being that significant, though, and a very reasonable approximation can be made by just taking #artifacts/scrolls read.

I have no idea what my earlier word salad of a post actually conveys, so here is what I basically meant: simply taking one artifact's chance is misleading. If you take #artifacts/scrolls read, you get a pretty good idea of artifact generating chance.

[cromat]

Framework described and uploaded(linux only), 1.2p5(before "more meaningful loot") in progress, I'll upload in in the evening

Small research regarding artifacts:
turns; artifacts
2082;four arts
2083;two arts
2085;two arts
2085;two arts
2538;five arts
4169;two arts
10018;eight arts
10018;10 arts
10018;10 arts
10018;nine arts
10018;nine arts
10018;nine arts
10018;eight arts
10018;10 arts
10018;11 arts

Definitely looks like 1/1000

[cromat]

Looks like spliting into smaller parts makes huge speedup. 11*10000K by Grey Elf Wiz@DL50@1.2p5 done

Top increases in probability(p5->p9, "more meaningful loot"):
helms of leadership 8.1
loaf of spider bread 5.0
fletchery set 4.5
torch 4.1
anvil 4.0
bandage 3.6
amulet of hunger 3.5
eternium quarrel 3.4
ring of gain attribute 3.4
ring of the clear mind 2.9
helms of mental stability 2.9
scroll of literacy check 2.3
amulet of chaos 2.2
helms of teleportation 2.0

[anon123]

http://www.adom.de/forums/project.php?issueid=1277

I guess I am to blame for the extra anvils, bandages and amulets of hunger.

[cromat]

REALLY exhaustive research done: over 30 000 000 generations altogether(true DL can be calculated through average gold drop). I'll analize them later

[cromat]

Ok, results partially analised(probabilities by DL yet to be calculated), all the DLs for "Create Item"-available items (it was one character, so not all artifacts) determined.
http://bit.ly/adomitems

[SirTheta]

This is some truly stunning work, cromat.