Goblinworks Executive Founder
|
The only thing you can conclude with a p value of 0,77 is that any correlation you have found is most likely due to chance. And thus, you have nothing that says the null hypothesis (same weapons) or the hypothesis you're testing (different weapons) is true. You can't reject either of them. Do more tests, find a correlation with a p below 0,05, then draw your conclusions.
Aaaaaah... so if we can't reject the null hypothesis, we have actually PROVEN it???!!!??? Wow. I never knew THAT particular trick of statistics...
OK
We are not trying to confirm the null hypothesis.
We want to confirm there are two separate formula at work and to do that we need to reject the null hypothesis.
We are not able to do that with two means that give a p=0.77 .
Intuitively, the reason for such a high p with so small a sample number is the standard deviation is over 3.0 for each sample set whereas the difference between the sample means is only 1.5 .
Bad statistics and bad science are bad.
First the null hypothesis (that things were working correctly as described) was rejected cleanly in the original data. The issue we have now is that we can't replicate the original results, so we can't write a proper bug report.
How was it "rejected cleanly"? There is probably a good reason you can't replicate it, I'd say.
Goblin Squad Member
|
For 3 keywords, the full damage in this case should be 98
Now also confirmed by Stephen (thanks!)
The issue here is that 79 =/= 98.
It's not not whether two sets of numbers come from different populations or how you should have defined the null hypothesis. Several of us are formally qualified to teach statistics, but that's not the topic here.
Fact: the OP observed results are due to some kind of error.
Fact: we know exactly where the number 79 comes from.
Fact: re-equipping gear (forcing keywords to be updated) solves the issue.
Fact: Stephen trying to reproduce the situation does not get the same error.
[/rant]
Since it's not repeated, we don't even know if it's an actual bug in the game code. Since it also seems very easy to fix, the devs really shouldn't put any effort into chasing it.
UNLESS someone can reproduce the bug and pin down the conditions where keywords don't update correctly. If that happens, let Stephen know.
EDIT: Capitolacracy, are you saying you saw the same bug (ie 79 max damage with a +2 staff) ?
|
Goblin Squad Member
|
I went to go and test this with shortbows last night, but unfortunately only got as far as taking the shortbows out of the bank and equipping one before I got called away.
If this can happen with a staff, it is likely it can happen with a bow. I will test swapping between +0 and +2 while out in the field.
|
Goblin Squad Member
|
I'm glad there isn't something repeatably wrong with keywords. Still, if the OP wasn't simple user error (and I'm not saying it was... :) ), the inventory and equipping mechanics seem to have some buggy edge cases.
I hope that the eventual combat logging will provide enough info to suss these things out! Like telling us what weapon, keywords, feat levels, etc were going into the damage function.
|
Goblinworks Executive Founder
|
| 1 person marked this as a favorite. |
How was it "rejected cleanly"? There is probably a good reason you can't replicate it, I'd say.
It's rejected because in the initial data sets, the series that should have a max and mode of 98 has one of 79.
The problem is that we cannot replicate how to get those results. In many circles, the presumption would be to consider it a failure to follow procedure and move on, instead of asking if there might be a bug related to updating keywords when weapons are changed out.
|
Goblin Squad Member
|
How was it "rejected cleanly"? There is probably a good reason you can't replicate it, I'd say.It's rejected because in the initial data sets, the series that should have a max and mode of 98 has one of 79.
The problem is that we cannot replicate how to get those results. In many circles, the presumption would be to consider it a failure to follow procedure and move on, instead of asking if there might be a bug related to updating keywords when weapons are changed out.
There is a difference between "not supported" and "rejected". Just because the null hypothesis was NOT disproved, does not mean it is true.
Stats can be challenging and I have not taught(tutored) it for 4 decades. Yet these arguments do not speak the language. Yes it is liars, damned liars, and statisticians. But even worse is poor statisticians who make claims with out full foundation. A little knowledge can be disastrous.
|
Goblin Squad Member
|
Negligible difference between one stats method and another. Is this the topic?
lol .. I kind of gave up, but I think the argument arose because I chose as the null hypothesis "the two staff are using the same attack equation" and then tried to claim that as true which technically is not correct :D
Eitherway the anomalie is interesting but was only reproduced by Capitalocracy suggesting whatever is going on (if anything) is more subtle than just "staffs are broken". If there is an issue its related to swapping out lower and higher plus weapons between combats.