Personal temper findings

[SpankWustler]

Your information had inaccuracies so I was looking for people to actually back up what you said with solid information. Currently there has been no solid answer derived via reverse equations of known % values to give a working equation to work with. Additionally, without a working parser for test servers let alone the fact that we can't even do the +2 relic trials at the moment, it would be more beneficial for numbers to be ran and verified via community.

In conclusion, your numbers are not substantiated at this time and I'm looking for people to focus on that rather then troll you. Don't jump my case for trying to correct the course of this topic. If you want to argue semantics of what a rdm can/can't do then go make a new topic for it and argue to your hearts content.

Well then, you might like to hear about something Byrthnoth posted some time yesterday.

Temper is weird.
404 Skill = 10% DA rate, recasting every 500 seconds
410 Skill = 12% DA rate, recasting every 500 seconds
435 Skill = 13% DA rate, recasting every 600 seconds
471 Skill = 17% DA rate, recasting every 600 seconds

So it is either non-linear (seems very unlikely) or there is something else going on. All precisions are within 1%.

As I understand it, Byrth's sample sizes were over 5000. THE PLOT THICKENS.

[hideka]

could you link his thread? the fluxiations between our numbers indicates 1 of 3 things
1). my sample size was too small and didnt allow certian tiers to ramp up enough.
2). he has outside DA equipped and Temper provides +% instead of +ammount
3). he just invented numbers to fill in his gaps.

[SpankWustler]

could you link his thread? the fluxiations between our numbers indicates 1 of 3 things
1). my sample size was too small and didnt allow certian tiers to ramp up enough.
2). he has outside DA equipped and Temper provides +% instead of +ammount
3). he just invented numbers to fill in his gaps.

Here's his original post. At least, I assume it's the original post. I only follow 2-ish threads on BG and exactly 0 threads on other community forums, and I think he posts on almost all of them.

Not a lot of information is included, but his sample size is mentioned further down that page. It's safe to assume that he eliminated painfully obvious confounds such as Double Attack equipment given the experience that Byrthnoth has with testing spells and abilities in FFXI, though.

If he just invented numbers, he would probably have invented numbers that made it scale in a linear fashion rather than...however it scales. Also, he would have invented cooler numbers. Like "schlexen", the number with a speech impediment that lurks between six and seven.

[saevel]

Of all our enhancing magic, none of them scale in a purely linear fashion, other then stoneskin. Most are tiered in some fashion. Enspells / Phalanx / Gain-Stat spells all change calculations once you pass a certain skill point, then also each require you to hit the next tier before additional enhancing magic adds to potency. For example, 460 has the same potency enspell as 479 enhancing magic.

[Byrth]

Sorry, those were parses of Hpmedes with a D0 weapon, which lets me afk for a few hours while I collect data. Useful, considering I don't have a controller atm and am severely lacking in keyboard coordination.

What you do is:
* Get a D1 weapon (like Trainee Knife)
* Make sure you have 65 STR or less (at least)
* Go find a high level Hpmede (I use Teleport-Holla, run towards Temenos, and look for a Hpmede with >300 evasion).
* Pull it, and make sure you're swinging quickly enough that you keep its interest through casts, but slowly enough that the parser doesn't blur your attack rounds (~>180 delay)
* Having low STR will give you an fSTR of -1, which lets you turn your D1 weapon into a D0 weapon. This means you can never hit for damage. Put yourself somewhere without agro and afk with the parser running.
* I'm not sure exactly where this method came from, but it was introduced to me in the BG Critical Hit Rate thread where they dissected the dDEX relationship.
* You can figure out how to cast Temper every 500-600 seconds and use Composure once in a while for an indefinite period of time if you really think about it.

The original post for Temper was actually the bgwiki discussion page. I put it in the random fact thread because I wanted some input. I ended up skipping out on the followup with DA gear in favor of confirming whether or not THF gets Triple Attack II (they don't), but I can do it tonight instead of figuring out whether or not WAR gets Double Attack II if people care.

I think I saw your data on wikia, and N=500 explains it. In general (for your higher sample stuff at least), you can reduce what you're measuring to a binomial distribution if you set up your experiment correctly (make sure there are only two states). The linked page has the statistics for determining your 95% confidence interval.

As for why my data is weird . . . I don't know. The parses were almost continuous, I never changed gear, Temper never wore off, etc. Why it jumps 2% in 6 skill and then 1% in 25 skill is a mystery to me. I wanted the story to work out to a nice 1% every 10 skill as it originally appeared it would. If you want work this out faster and with less static about people caring/not caring about Temper and the RDM Meleeing that it represents, I'm Byrthnoth on BG and you can just PM me.

PS. Saeval, those things you're describing are called "piecewise linear." Nonlinear things look like this:
http://wiki.bluegartr.com/bg/Talk:Afflatus_Solace

[saevel]

My point was that none of our spells are a straight 1 skill = ~something something~, their all staggered depending on which developer wrote the code. So while you can make a graph with the tiered points and draw a line through it, the line itself is a lie as it would suggest a difference between 360 and 379 enhancing magic on Enspell potency which in actuality there is none. It's a stair stepped appearance, no reason to think Temper would be any difference. Eventually people will figure out the tiers.

[Byrth]

Ah! I have just the page for this situation!

See, the second equation you're referring to actually is linear. It just involves a floor function. I've never verified that the first equation (nonlinear) is legit. It seems unlikely to me that it is, but it also isn't very relevant content anymore.

[hideka]

well, my guess, is that a base stat could possibly affect how your DA effects Ramp up, thats the only explination for the large diffrence were seeing between our numbers

[hideka]

one thing that bothers me, is that my projected outcome for Max skill DA rates are identical to his max skill outcome rates, but none of our lower Skill tiers match up... this indicates that the .125 DA per point of skill over 400 might be correct and for some reason our %'s arent matching up for some odd reason...

.125* 43 = (5.375 + 11.63) = (17.005%) double attack.

( the 11.63 was the ammount of DA i was pulling down at 429 Skill, the 43 Represented the skill i was missing to cap)

[Byrth]

well, my guess, is that a base stat could possibly affect how your DA effects Ramp up that's the only explanation for the large difference were seeing between our numbers

It could be, but Stoneskin is the only enhancing magic known to be related to a base stat thus far.
* I lost 10 MND and 5 INT going from 471 skill (17%) to 435 skill (13%).
* Then I lost a further 13 MND going from 435 skill (13%) to 410 skill (12%).
So if it's MND related, the effect is pretty weak and/or not there.

Also, your values and mine don't necessarily disagree. The closest to disagreeing we are is at your 402 sample, and my going from 404 to 410 and seeing +2% DA rate shows that area is kind of weird. Also, it's good to keep in mind that the confidence intervals with your sample sizes are a little too large to draw strong conclusions from.