Ridiculous droppings

[Slariton]

I decided to try a new run as an Infiltrator. (The tipping factor was getting annoyed at having to change my sphere of sight every time I attack an enemy. Now I remember why melee was always deprecated in Geneforge.)

 

Anyway, the first Thahd I killed dropped a Thahdskin Tunic. The second Thahd I killed dropped a Thahdskin Tunic. And the third Thahd I killed also dropped a Thahdskin Tunic. And I didn't pump Luck above 2.

 

Further Thahds have not, so the game isn't broken, but geez. What are the odds? If only I have this luck when collecting Glaahk Eyes.

[Schrodinger]

Hm. The odds are probably on the order of 1 in 1,000,000, given the 1% drop chance. If your luck adds a 1% factor, then between that and 1 in 30,000. Either way, pretty dang lucky. I've not received one yet.

[Student of Trinity]

I have long been suspicious of the Geneforge random number generator.

[Randomizer]

I got thahdskin tunics the first 2 times I played and haven't got one since until I reach West Illya Road. I must have tried over 12 times with no luck in Shardfield.

[Slariton]

Incidentally, has anyone ever gotten a random item drop, then reloaded and not gotten it? Or vice versa. I keep forgetting to try this and see if it works -- as I have never heard of anyone getting a random drop by repeated save and reload. I wonder if they are set when the creatures are spawned on zone loading?

[wz. As]

Quote:

Originally written by Everything Remains To Be Done:
Further Thahds have not, so the game isn't broken, but geez. What are the odds? If only I have this luck when collecting Glaahk Eyes.

Given that it already happened, the probability is 1. This is not just nitpicking — you didn't set out in advance to try and get three of these things in a row. If you started a game and thought, gee, I wonder if I'll get three thahdskin tunics off the next thahds I see, that would be a little more impressive when it actually happened.

As it is, though, it seems to be unlikely only because it was unexpected. The question you should really be asking is, "what is the probability that the random number generator should bring about an event that would bring my level of surprise past the threshold required for me to post something on Spiderweb about it?" I'd guess the probability for that would be greater than that for causing three thahdskin tunic drops.

[Kelandon]

Well, no. He is right to point out that this event was unusual, and it is not statistically immoral to calculate the odds of it happening again, which ought to be negligible if Spiderweb games actually used legitimate random number generation. And if something happens that ought to have an almost zero chance of happening again, it is reasonable to call that event unlikely.

 

Still, if you find that thahd droppings are ridiculous, you might want to check what they ate.

[Slariton]

Quote:

Originally written by wz. As:
Given that it already happened, the probability is 1. This is not just nitpicking — you didn't set out in advance to try and get three of these things in a row. If you started a game and thought, gee, I wonder if I'll get three thahdskin tunics off the next thahds I see, that would be a little more impressive when it actually happened.

I hope that what you meant was: "you didn't set up a controlled experiment and test for thahd drops." How is intention relevant to this kind of test? And if it is relevant, how they HECK is it desirable to have that potential bias?

You're right that I was only recounting an anecdote, but did I ever present it as anything else? Single random data points are not statistically significant, but they are not meaningless, and they are certainly not any less significant than single deliberately chosen data points with an equal lack of context.

[wz. As]

Kelandon: Agreed.

 

Slarty: Intention is relevant, because although the probability of getting three thahdskin tunics in a row is very small, there are many events which could be considered "unlikely," the occurrence of any of which might prompt one to post a new topic about it.

 

For instance, we could just as easily be reading a new topic about getting four consecutive vlish tentacle drops or doing 66 damage with an ice crystal three turns in a row.

 

I guess what I'm saying is that 1) although the probability of three thahdskin tunics is small, the amount of surprise derived therefrom may be unwarranted, and 2) the random number generator need not necessarily have any bias, since the odds of any surprising event happening are probably not too low.

[Student of Trinity]

The chance of this happening on any given triple Thahd kill is supposed to be one in a million. In a game you get maybe a few triple Thahd kills, people who could be expected to post a report of a triple-drop here have probably played ten to twenty games of one Geneforge or other, and there are probably 50 to a 100 such people. So the number of triple Thahd kills that these boards have effectively sampled is probably within a factor of 10 of 10,000. So the fact that we have now had such a report is indeed somewhat surprising, but at a probability on the order of 0.1 to 10%, it is not utterly incredible. Take it as a more grounds for suspicion of the RNG, but not quite as a smoking gun.

 

EDIT: Taking into account wz. As's good point that the RNG has lots more opportunities to generate surprising co-incidences, and we have only heard of these Thahd tunics, and not of similar events with Vlish tentacles and what-not ... the conclusion would seem to be that the RNG beats the rap, at least for now.

[Kelandon]

We do know from experience that the random number generator does not, in fact, generate random numbers, though, unless Spidweb has changed it since the last few games.

[Slariton]

What does it do? I mean, does it generate reasonably pseudorandom numbers, or not?

[Kelandon]

I'm not totally sure. I think part of the problem that I've had with it is that I've been looking at to-hit percentages, which are just not accurate. Monsters that have %5 chances to hit me hit me a heck of a lot more often than one time in twenty.

 

I don't know about drops, though. I should check that in BoA, where it's really easy to conduct a simple experiment: make a monster with a 5% chance to drop something, place about forty of them in a town, and then kill all of them and check how many actually dropped that item.

[Student of Trinity]

5% is a tough chance to call, though. In a hundred attacks you should get hit about 5 times, but the odds of those five hits coming nicely in 20-attempt intervals are very low. You're likely to have a few of them come quite a bit closer together, then have some longer sequences of straight failure. One tends not to notice that one has been missed 40 times in a row, when one is expecting to be hit 'hardly ever'. But we do notice 'Hey, that's three times in ten rounds!'

 

I still wonder, but it's a hard charge to make stick.

[Schrodinger]

Quote:

Originally written by Everything Remains To Be Done:
Incidentally, has anyone ever gotten a random item drop, then reloaded and not gotten it?

I received a fyoraskin cloak from a fyora in the forsaken lands, but reloaded since in the process I lost my creations. I did not get the cloak a second time, to my dismay.

[Kelandon]

Quote:

Originally written by Student of Trinity:
But we do notice 'Hey, that's three times in ten rounds!'

Three times in ten rounds would not have raised my eyebrows, but ten times in ten rounds started to. When I was consistently getting hit more times than not, I started to wonder.

I should point out that I'm not a fool. I am aware of the human element in observing probability. I took this into account when I realized that the displayed to-hit probabilities were not accurate.

[Micawber]

I think if your hit chance is 90% and you can still miss 5 times in a row, then there is something wrong. Either Geneforge sometimes gets "stuck" when generating a random number, or there is some part of the engine that does not properly feed through to the percentages shown.

 

Is it possible that the to-hit % is misleading for a similar reason to the armor %?

[Synergy]

Quote:

Originally written by Schrodinger:
I received a fyoraskin cloak from a fyora in the forsaken lands, but reloaded since in the process I lost my creations. I did not get the cloak a second time, to my dismay.

That'll learn ya to not reload after scoring a Fyoraskin Cloak.

-S-

[Kelandon]

Quote:

Originally written by Micawber.:
Is it possible that the to-hit % is misleading for a similar reason to the armor %?

Probably.

[Fort]

Quote:

Originally written by Micawber.:
I think if your hit chance is 90% and you can still miss 5 times in a row, then there is something wrong. Either Geneforge sometimes gets "stuck" when generating a random number, or there is some part of the engine that does not properly feed through to the percentages shown.

Is it possible that the to-hit % is misleading for a similar reason to the armor %?

But when you have a 90% hit chance you hit much more often than you miss, right? Rare incidents like this are not indicative of whether the RNG is evenly distributed, that is, fair. I think the RNG is slightly biased because I am pretty sure Jeff opted for a quick, standard number generating algorithm, but it is not a big deal.

If you are constantly being hit by 10% shots, then start being concerned.

[Slariton]

I think the issue here is less likely to be the RNG than the display of fatuous numbers. Adequate pseudorandom number generators are not hard to come by. Seeing as Jeff once studied statistics in grad school and is clearly competent with simple algorithms, I doubt he would find them difficult to handle.

 

We know from the Armor % display feature that not all displayed numbers are accurate representations of anything. I experimented with to-hit a lot in G3 when I tried to duplicate the A4 99% dodge build, and it failed because to-hit calculations do really weird things in Geneforge when they approach either extreme.

[Fort]

The armor percentages are of course rubbish because none of the actual mathematics behind any of his armor systems involve a sum total of percentages. Different pieces of army provide different individual percentages of protection for damage, and the asymptotic behavior of the caps make everything even more skewed.

[Synergy]

I'm increasingly curious about this. Can someone explain to me how a random number generator would not in fact generate truly random numbers?

 

Thanks,

 

-S-

[Emperor Tullegolar]

Random.org has pretty much everything you could ever want to know on the subject and more. Plus t-shirts!

[*i]

Quote:

Originally written by Synergy67:
I'm increasingly curious about this. Can someone explain to me how a random number generator would not in fact generate truly random numbers?

Random number generators use an algorithm and are given a starting "seed", usually an arbitrarily large number. This algorithm is recursive in that the input is the last entry and hence affects the output which is next entry.

What defines "random" is that the sequence of numbers are evenly distributed between 0 and 1 (the unit bounds; to get a different number, you may simply manipulate the result) and exhibits no skip, pattern, or repetition.

No sequence meets these requirements perfectly. Typically some seeds can lead to these subtle patterns. One often can get higher fidelity results by using lots of different seeds. As for lacking repetition, all of these algorithms repeat, but typically not until after 10^30 (or much, much more) numbers given an appropriate seed.

To get a more "random" number, typically the CPU pulls from the clock. In some applications, such as games, this is desirable. In scientific applications where replication is vital, a known starting seed is desirable.

So basically, a CPU is completely a deterministic thing when functioning properly. We humans need to coax it to give us a "random" number.

[Suspicious Vlish]

I remember boosting Luck in GF2, and getting two Roamertooth Bands (+2 to Fire Shaping) in a row, just outside the Magus Complex. Highly unusual, and quite a nice perk.

[Schrodinger]

I've had an unusual experience as well, with two clawbugs dropping clawbug carapace shields in a row. Could it be that the random number generator operates over a specific amount of time- that is, you have 3 seconds or so out of every 300 in which enemies drop their best goods?

[Emperor Tullegolar]

Is it possible that luck caps out or goes back to zero somewhere before 30? I cheated and played the game through with 30 luck, and other than maybe one fyora cape I didn't get anything special. No thahd tunics, no roamer rings, nothing.

[Synergy]

It's been apparent for some time that Luck does not affect droppings. The question then is, what does Luck do in GF4, and is it worth any bother at all? I vote no, by my experience. What happened with 30 Luck, ET? Did you survive zeroed hits more often or miss getting hit more often?

 

-S-

[Emperor Tullegolar]

Sorry, but I couldn't tell you. You see, I put everything to 30, which leads to the enemy missing you just about always, and makes the character destroy just about everything in one hit. Someone with more patience should make use of my editor and try only giving themselves 30 luck.

[Lilith]

I think that every point of Luck increases your hit and dodge rates by 1%, but that's hardly a reason to buy it in itself.

[Slariton]

By 2%.

 

Out of curiosity, did anyone ever figure out what Luck actually did in Exile other than save you from dying?

[Drakefyre]

Isn't that enough? I think it may also have an effect on poison % and damage.

[Schrodinger]

Each point of luck increases your stun, mental, acid, and poison resistance by 3%, which is pretty good. The clover boots are better resistance than several roughly equivalent items.

[Slariton]

Are you sure about all of those?

 

In previous games -- definitely in A4, and I think in older Geneforges -- having Luck would cause the status screen to display some real bonuses and also some spurious ones that were never factored in anywhere.

[Schrodinger]

That's the result from my testing with a shocktrooper. I recorded my resistance before and after investment of 3 points of luck, with those results (+9% to those resists).

 

I am probably using a different version than you, so maybe displaying those effects is one of the things Jeff changed from one version to the next?

[Lilith]

Quote:

Originally written by Schrodinger:
That's the result from my testing with a shocktrooper. I recorded my resistance before and after investment of 3 points of luck, with those results (+9% to those resists).

We're not arguing about whether it increases your displayed resistance; we're arguing about whether the displayed resistance reflects the actual resistance. In other words, have you, say, let a submission turret attack you 100 times with and without extra luck and seen whether it affects the average number of AP you lose?

This isn't just speculation; every point of luck in Avernum 4 supposedly increased your elemental resistances, but controlled testing showed that it didn't actually make a damn bit of difference -- you took just as much damage with 10 luck as with 0.

[Slariton]

The good news is this is DRASTICALLY easier to test for in G4, since the game tells you how much damage you take AND how much you resist. The proportion resisted is definitely not static, even given two completely identical attacks on the same PC, but it shouldn't be hard to figure out an average given a decent set of trials.

 

I want to do this, actually. Quantifying the reduction you get from Parry in particular would be useful.

[Lilith]

Testing Luck is hard, though, since most of the things it helps you resist don't do direct damage. I suppose the easiest way would be to go into an area that does continual poison damage and see if Luck helps reduce it.

[Randomizer]

So go to the Dumping Pit and kill the roamers in the entrance. You can sit next to the healing pool and just wait for the poison to hit you every few rounds.

[Schrodinger]

OK- I figured it out. It took some testing, and my statistics are based on a total of about 80 acid resistance events, and about 40 magic damage events for both with and without luck.

 

I made an artila and had it shoot me, recorded the damage and damage resisted with and without 4 points of luck (clover boots + charm = +4 luck).

 

Character's armor class with boots and charm: 54%, without 50%. Energy resistance with: 28%, without 26%. Acid resistance with 57%, without 45%.

 

The artila's to hit % with boots and charm: 76%, without 84%.

 

Total damage done by magic artila spit:

With boots and charm: 747 (270 resisted)

Without: 803 (241 resisted).

26.5% resisted with luck, 23.1% resisted without.

This is probably best attributed to the energy resistance difference due to the armor.

 

Total damage by acid from artila:

With boots and charm: 687 (288 resisted)

Without: 917 (384 resisted).

29.5% resisted with luck, 29.5% resisted without.

 

So at first glance, luck doesn't seem to enhance resistances at all. However, there's more to it, so consider:

 

The acid damage done per magic damage point with luck was 0.92.

 

The acid damage done per magic damage point without luck was 1.14

 

The ratio of these two (1.14/0.92 = 1.24) is roughly inversely related to the ratio of my relative acid resistances (57%/45% = 1.267). So the luck doesn't act on the "damage resisted" number, but instead seems to act on the total amount of acid layers given per attack. It does seem that luck affects these resistances approximately proportional to the values on the stat page!

 

EDIT: So this looks like the resistance percentage affected by luck is the chance to completely avoid a layer of damage, rather than resisting parts of it.

[Lilith]

Sounds right. I've long suspected that that's how acid and poison resistance actually work. (Which also means there's no benefit to putting on acid or poison resistance gear once you're already poisoned.)

 

By the way, is it just me or does there seem to be a cap on acid and poison damage per round? No matter how long I get beaten on by rotghroths, I never take more than about 50-70 damage from acid -- on the other hand, being hit more does seem to extend the duration of the acid effect.

[Randomizer]

Acid seem to work that way in Avernum 4. Multiple castings of acid spray increased the duration, but it had a maximum effect per round no matter how many were cast.

[Slariton]

Yeah. I think the max per round is higher, though. When I was carting around seven Artila, they seemed to max out around 90 acid damage per round.

[Upon Mars.]

I killed a drakon, a eye beast and a rothrogoth and they didn't say thanks they didn't give me any money this is outrageous!!!

Off with there heads!!!

Droopings are irregulliar and it hapens like that you can have a cake killing a servile and have a tash killing a crystal luk make droppings to be more suchessfull.

[John S]

Quote:

Originally written by upon mars:
I killed a drakon, a eye beast and a rothrogoth and they didn't say thanks they didn't give me any money this is outrageous!!!
Off with there heads!!!

aaarggghhhhhhh! In addition to being annoying, this sort of post really doesn't help your case for remaining a member. If you actually read this, please make sure you are aware of the future consequences.

[Drakefyre]

Quote:

Originally written by upon mars:
I killed a drakon, a eye beast and a rothrogoth and they didn't say thanks they didn't give me any money this is outrageous!!!
Off with there heads!!!

Did you read the PM I sent you?

[Upon Mars.]

I leafed this post on a hurry sorry i didn't completed it .

[Lennyhipp3]

isn't the %hit chance factored not only on your character's ability to hit (or be hit) but the weapons themselves sometimes add or subtract to the hit chance, the enemies might have armor/ spells that add or subtract to your hit chance, and some cast spells that add or subtract to your being hit chance.

 

i don't know that it's so much random, but it would explain if you built up your armor to 50% and you still get him more than half the time.

 

LH

[Slariton]

Here's how hit chance works.

 

Each form of attack has a default chance to-hit. Almost all of those are between 50% and 80%.

 

Each level of attack strength increases hit chance by 5%. For creations and NPCs, attack strength = Strength + skill with the attack ability. Skill is a hidden stat on creations that does not change with level. Strength is used even for missile attacks and yes, even for spells!

 

For the PC, attack strength depends on the type of attack.

Melee: Strength + Melee Weapons + bonus from equipped weapon

Missile: Dexterity + Missile Weapons + bonus from missile used

Magic: Spellcraft + appropriate magic category + skill in individual spell

 

If the attacker has Luck (PCs + Gazers, basically) add 2% chance per level of Luck.

 

Remove 2% chance per level of defender's Luck.

 

Remove 5% chance per level of Dexterity the defender has.

 

(Also remove 5% chance per level of Nimbleness, a hidden stat that Thahd Shades and Plated Artila (and no one else) had in the first three games. I can't remember if Plated Artila still have it in G4.)

 

In summary:

Default chance to hit <-- attack type

+5% / lvl of attack strength

+2% / lvl of Luck

-2% / lvl of Luck

-5% / lvl of Dexterity

 

Note that if the final number is capped on top by 99% and on bottom by 1%. However, the actual results stop following the expected distribution when you get close to either of those numbers, so in practice the limits are closer to 95% and 10% or so.