Spellbook distribution research thread.

[Blank4u47]

It was pointed out by soirana that we should post a thread instead of derailing his. I quite agree.

So here goes.

The goal here is to create a character and compare the results after accumulating 100 spellbooks and counting them. This only counts for monster drops, not guaranteed sources such as the sage etc.

The purpose of this is to see what percentage of spellbooks are frost bolt/lightning for necro/druid respectively.

Anyone who can lend a helping hand in this endeavor will be helping to possibly improve the game's spellbook distribution.

[JellySlayer]

Relevant RFEs 3798 and 3696

[Soirana]

Relevant RFEs 3798 and 3696

Well, reports are prone to deliver extremes of possible RNG... since people are more likely report abnormalities

More on topic my first 25 books (troll necro r60 wand born I2-5) starting books not included

Calm monster
Cure serious wounds
2xDarkness
Farsight
5xFrost Bolt
2xInvisibility
Know alignment
2xLight
Magic Missile
2xRemove Curse
Scare monster
2xSlow monster
Slow poison
Strength of atlas
Teleportation
Web

p.s. also dropped Confucius book

Gonna keep SoA and drop rest.

I think the book problem actually varies in between versions, but I rarely play casters.

[Cactus]

I'd be very happy to generate some data.

Thanks for the links, JellySlayer. I wasn't aware of the more recent one. It appears TB already tried to reproduce the problem:

I just used debug mode to test this further with a dwarves druid (currently on D: 16). I generated lots of random spell books and received the following distribution (which looks totally ok to me)

So how exactly were these spellbooks generated? What is the difference between debug mode and a normal game? Maybe that's where to look for the problem?

In the meantime, generating more resilient data from actual games surely cannot hurt.

[Soirana]

So how exactly were these spellbooks generated? What is the difference between debug mode and a normal game? Maybe that's where to look for the problem?

Normally debug mode is not something retail version is shipped with.

Go ask Thomas if you need details.

[Soirana]

Got another 25

Acid Bolt
Calm monster
Cure light wounds
darkness
Fire  bolt
7-Frost bolt
Ice ball
Improved fireball
Knock
Light
Magic Missile
neutralize poison
2xSlow monster
Slow poison
SoA
Summon monsters
2xWeb

Overall 12/50, So my crude estimate (lol) would be 24%...

[yhal003]

RNG in adom is very strange sometimes. 100 books as a sample size should be plenty if there is a simple percentage, but there is no way you get 66 books or more with 0.25 chance - the probabilities are minuscule. Try playing with this for example http://www.vassarstats.net/textbook/ch5apx.html.

I don't have much experience with spellbook drops playing non-spellcasters most of the time. But some other item types often drop in a very bizzare fashion. For example my last game had no eternium boot drops and only single non-brass bracers pair. It is like there is a seed rolled on start that determines drop probability for the rest of the game.

I don't think this question can be resolved without some code diving or preloading custom random functions.

[Blank4u47]

RNG in adom is very strange sometimes. 100 books as a sample size should be plenty if there is a simple percentage, but there is no way you get 66 books or more with 0.25 chance - the probabilities are minuscule. Try playing with this for example http://www.vassarstats.net/textbook/ch5apx.html.

I don't have much experience with spellbook drops playing non-spellcasters most of the time. But some other item types often drop in a very bizzare fashion. For example my last game had no eternium boot drops and only single non-brass bracers pair. It is like there is a seed rolled on start that determines drop probability for the rest of the game.

I don't think this question can be resolved without some code diving or preloading custom random functions.

Yep, rng can be wierd sometimes, which is why I'm asking for the distributions from several people. 100 as a sample size is too low even for real life statistics.

[yhal003]

100 as a sample size is too low even for real life statistics.

Depends on percentage but you are probably right in this case. With 25% and 100 books, 3 deviations is between 12 and 38 drops which is quite a wide range (still 66 is way outside that).

[Soirana]

Depends on percentage but you are probably right in this case. With 25% and 100 books, 3 deviations is between 12 and 38 drops which is quite a wide range (still 66 is way outside that).

I hate to repeat, but is where actual list of 100 dropped books in which 66 are same?

Because all I've read were samples of <20, citations of spell knowledge and reported impressions (2/3 look differently when you read each book upon finding and actually calculate). Plus the only reports with numbers were r57.

Cause I grinned few mist elf necros in small cave and that seriously is not that bad.

p.s. slight decourse into Maths
66/100 and 12/50 is 0.52 if combined
with triple deviation on 100 set it would be -- 52+/-15
with triple deviation on 50 set it would be -- 26+/-11
So theoretically both observations still can be generated if chance is somewhere around 0.5. (They would be extreme lucky vs extreme unlucky, at 0.51 100 is in range of 36-66, while 50 is in 14-36 (with some stretches to rounding)).

p.p.s Sorear thought spells have rather straight drop chances. Well, again it just observation but in the last versions I don't really believe Light+Dark being ~20% in r60ish
http://www.it-is-law.com/adom/index.php?title=Spells