I might, what sample size you want, 1 cast or 2?

>>>>>
I might, what sample size you want, 1 cast or 2?
<<<<<

+/- 3%, 95% confidence interval is what I'd like.

The original tests didn't reach that point, mind you. More like +/- 10% with 95% confidence interval.

On the other hand, the original tests were remarkably consistent - we didn't see results just scattered over that +/- 10%. What we saw were very precise results (variations of less than 1% for the most part), then a large jump, then more very precise results (variations of less than 1% for the most part), repeated till the next jump.

Which suggests strongly that while his results lacked statistical rigor, they shouldn't be discarded out of hand without at least an attempt at verification/falsification.

Especially given that Mythic does, from time to time, change things without telling us. And sometimes changes things without even realizing they are changing things.

EDIT:
Note, by the by, that in at least one case, the discontinuity was greater than 10%. Which suggests even more strongly that while the precise values may be incorrect, the presence of those discontinuities is probably real.

 

-----signature-----

Thanks Angharad. Shouldn't even have to do the math to see that there is an obvious discontinuity. And this isn't a one time thing that we all have to accept, every single person should be able to repeat the test and get essentially identical answers. Oh well, he's just trolling anyways. I'd love to try and crack the code to these numbers though, being a university math student. But this is probably the busiest week i've ever had so far in university, i've done nonstop math for 3 days, and I have nothing to do for the next 2 days except sleep, and more math.

 

-----signature-----

You don't know what you're talking about. LOL @ trying to use statistics on something that has no variance. That's like you going to the store and buying a candy bar and telling the cashier you don't believe that the price is correct because the sample size is too small.

Then, even if there is some variance (say from random lag or some unknown wacky function) you have no idea what it is so your attempt to sound smart and say what the CI of the original tests were is 100% full of fail.

 

-----signature-----

I don't know if anyone else has been following this, but I finally bumped my Wiz (2.8 dd and 2.0 bolts) up to 392. I didn't do any 4000 cast tests or anything, but it is a noticeable difference. Thanks for this thread.

 

-----signature-----

Well if you lok on the ywain version of this you cans ee that the tests are not so good as ppl thought, as they get diff speeds on diff tests with the same spells.

Well yes, and we confirmed that the menta NS for example is bugged and doesn't cast as fast as other 2.0s spells. The breaking points are the same (at least the one at 392), but other 2.0s spells we've tested casts a lot faster.

Feel free to aks any questions you have and i'll do my best to answer them.


Btw, i love how you say my tests are wrong when the other guy used an even smaller sample and it still refutes your 386 dex theory and confirms my tests on pretty much all points.

 

-----signature-----

You are silly and old.
They are months that we tested these datas and that they are true.

If you dont test & continue to ask around "the truth is behind a 6 years old paradigma", plz go play tetris.

No. And you continue to be more pathetic every time you post on this subject.

What doesn't make sense about the game-tick theory is that it would be consistent across all the various casting speeds and associated DEX values. However, your breaking points jump all over the place from one spell to another, when all that changes is the base casting speed of that spell. The tick increase/decrease would seem to result in a consistent change to the old 386 DEX Cap rule if it were true, and if the old rule was true naturally. (A tick change is a linear event, so it would just accordingly raise/lower the DEX cap by a set amount too if such a cap did exist.)

I think that this looks a lot more like the usual "rounding" errors, as in not rounding at all but truncating as computers usually do in integer math. (Rounding requires analysis, which requires extra processing time that gets very expensive across a large amount of calculations per second.)

Everyone knows this phenomenon from the way in which %Duration gear affects the duration of a 9s Stun. In order to make that stun last the maximum time you need 23% Duration, because that's what takes a base 9s to 11s, as any fractional component is simply truncated. If you have only +22% Duration you end up with a 10.98s duration, but the .98 gets simply truncated and you get the same exact effect as if you only had +12% Duration in your temp.

I'm guessing that somewhere in Mythic's casting formulas there is also integer math being used that just ignores fractional components and thus you end up with apparent "breaking points."

I agree that the results are so blatant that it appears the sample size is enough to prove that something odd is going on. I suspect that people who do not see this on the basis of looking at the data alone have trouble visualizing numbers on paper, and should probably go try to test it for themselves to see how blatant a difference that has to be in order to change the values that much across 100 casts. (I do agree that the 4x25 thing is a little too bogus. You can't draw a valid conclusion from that in general, but that is not to say that you can't wing it to find a "breaking point" that you're pretty sure is in a certain range of DEX.) Lag would not generally have such a significant and especially not such a consistent impact across various tests.

What would be a good test to do is simply disprove the 386 casting cap without all this granularity. Just choose the best case from your sample above and retest it with ONLY 386 DEX vs. 400+ DEX and you can easily prove that the 386 Cap is wrong. Once that is accomplished plainly and without plausible doubt, even the most thick-headed people will be more open to new ideas/input.

 

-----signature-----