Hunting ammo test results

[nipon621]

I tested the corpse drop rates using savegame scumming - 10 events for noncursed vs blessed ammo of hunting against a corpse leaving target (intrinsic granting if I can find it).

Findings: 10 trials of killing a giant slug with uncursed hunting ammo using the same level 18 archer (who did not posses the food preservation skill) at the same starting turn had the following results

3 corpses
7 without a corpse

Same PC starting about 6 turns later (for blessing the ammo, lining up the shot) against the same slug with blessed ammo of hunting

6 corpses
4 without a corpse

I briefly toyed with calculating the number required to reach statistical significance, but decided it's probably not worth the time. The fact of the matter is that the number of trials I did cannot even begin to give something resembling a statistically significant result. Since I don't have the time to calculate the required sample size to achieve... oh, I'd be content with a 60 per-cent power to detect, I'm going to round to the nearest fifty per-cent.

So my conclusion, is that it is probably a 50 per-cent drop chance with a slight possibility that blessed hunting ammo gives more corpse drops

[anon123]

10 trials of killing a giant slug with uncursed hunting ammo using the same level 18 archer (who did not posses the food preservation skill) at the same starting turn had the following results

3 corpses

This should be enough evidence ammo of hunting is pretty powerful

Thanks for the research!

[JellySlayer]

This is pretty remarkable. IIRC, giant slug corpse drop rate is ~2%.

You might want to do a control test anyway though... just in case corpse drop rates are somehow affect by the fact that you're save scumming.

[nipon621]

This is pretty remarkable. IIRC, giant slug corpse drop rate is ~2%.

You might want to do a control test anyway though... just in case corpse drop rates are somehow affect by the fact that you're save scumming.

The last thing I want to be doing is re-rolling characters until I get similar ones, have them find slaying ammo, and then have them each find a giant slug at a similar number of turns into the game with all (really and truly all) other variables being equal.

Without some sort of supercomputer with custom coded state of the art character creation and botting software, making an actual, true control group would take at least a few months (if not years given the sheer number of variables), even for a sample size as small as 10 occurrences.

In short, I have no idea how to make a control group for comparison under "real world" (lol) circumstances without taking ridiculous pains.

Couldn't these results be at least partly confirmed by source diving? As much as I like to respect the wishes of TB, if this question is going to be definitively answered, it seems to be the most logical, authoritative and time effective option. On the other hand, I know next to nothing about coding, source code diving, or anything like it.

Sorry, but unless a better way of developing a control group is devised, I can't sign up to test one.

[JellySlayer]

The last thing I want to be doing is re-rolling characters until I get similar ones, have them find slaying ammo, and then have them each find a giant slug at a similar number of turns into the game with all (really and truly all) other variables being equal.

Without some sort of supercomputer with custom coded state of the art character creation and botting software, making an actual, true control group would take at least a few months (if not years given the sheer number of variables), even for a sample size as small as 10 occurrences.

In short, I have no idea how to make a control group for comparison under "real world" (lol) circumstances without taking ridiculous pains.

Couldn't these results be at least partly confirmed by source diving? As much as I like to respect the wishes of TB, if this question is going to be definitively answered, it seems to be the most logical, authoritative and time effective option. On the other hand, I know next to nothing about coding, source code diving, or anything like it.

Sorry, but unless a better way of developing a control group is devised, I can't sign up to test one.

Simple way: Run the same experiment as you did previously, but use regular arrows.

[grobblewobble]

I can confirm that this seems about right from experience. Soirana told me, he might know more about it. A fun way to start an archer is to get the heir gift and do the druid quest, getting =acid while at it.

[gut]

> just in case corpse drop rates are somehow affect by the fact that you're save scumming

I don't know of any drop rates being affected by save scumming... or anything for that matter.

[gut]

> I briefly toyed with calculating the number required to reach statistical significance, but decided it's probably not worth the time

for the kind of testing you are doing, you would prolly benefit very much from using adom-tas.
it is a program written by tuplanolla for tool assisted speedruns, meaning it is an uber-savescum utility among other things.
it works as effortlessly as a rewind button