User talk:Dr ishmael/BWE Skill data collection

My only concern would be that the damage tooltip does not always match the actual damage done. Tooltip damage does not seem to take equipped weapon into account, whereas actual damage done does. Except for some skills, which don't seem to use your weapons at all. I'm guessing that at first we should get proper scaling factors from tooltip damage. Real-world damages are somewhat more annoying to get, since non-steady weapons have a range +-5-10% (and even with a sample size of 20 best estimates of medians are still +-1%), and steady weapons do so little damage that rounding frequently approaches 2-5%. Both of which are somewhat less precise than I would like. Capric 15:14, 24 May 2012 (UTC)

Yeah, tooltip numbers are what I'm primarily concerned with for now so that we can standardize the damage numbers across all skills here on the wiki. Real damage will always be affected by many things, including your weapon, the target's armor, whatever buffs/conditions/effects are on either of you, etc. Damage#Calculating damage could use a lot of updating based on our findings. —Dr Ishmael 15:38, 24 May 2012 (UTC)

I'm not quite clever enough to prove it, but I think that the divisor of 900 in the tooltip damage formula might be an artifact of the 300 difference in powers between the two data points. That and spreadsheet rounding; if you run the same slope and intercept calculations with higher precision you get a common denominator of 300 instead of 900. Annoyingly enough, the mesmer data I captured was also exactly 300 power different from the data you have up on the spreadsheet. It's almost as if a stat difference of 300 is very easy to set and reset in the PvP lobby. Anyway, I wouldn't say this is actionable or anything now, but it will probably be a good idea to get 3 or 5 data points next BWE so we can get a more proper linear regression going. ~ Capric 23:59, 4 June 2012 (UTC)

+300 Power is what you get from maxing out the trait line - which is what the person who made the thief video did to see the effects. Did you refund all your trait points before recording your data? My mesmer probably had Domination maxed in her default build.

You're right - all the numbers divide evenly by 3, and in doing so we reduce some of them to prime numbers, which is a good indication that we've found the "base" formula. It also makes the damage formula equivalent to the one for healing, which I feel is even more evidence that we're right. Thanks for noticing that! —Dr Ishmael 00:30, 5 June 2012 (UTC)

You didn't happen to record the condition damage for mesmer skills, did you? Thief doesn't have any skills that cause burning or confusion, and mesmer does, so it would've been nice if we could've worked those out before Friday. But, we only have to wait until Friday now, so I guess it's not that important. —Dr Ishmael 00:51, 5 June 2012 (UTC)

Sorry, I didn't record condition damage; I had actually thought condition damage was mostly solved to be honest.

Regarding the damage formula thing, my contention was that the 300/900 divisor was purely an artifact of the data collection. The slopes have been calculated as (damage1 - damage2) / (power1 - power2), and the intercept is calculated from that slope. However, in all the test points we have, (power1 - power2) is equal to exactly 300. And since (damage1 - damage2) will always be an integer, this will guarantee that all calculated slopes and intercepts can be expressed as an integer divisible by 300. Thus, I am not particularly convinced that the 300 is anything internal to the game calculations. A simple test case will be to have power differences of more than 300 and calculate if there are any common divisors across intercepts and slopes, or just get a lot of points and use linear regression. A better test will be to somehow increment power one by one and record precisely when the tooltip damage increments.

I don't actually think this is really a big deal at the moment, since for now it's most useful just to get a consistent set of numbers which can be used for qualitative comparisons. And also the data are still very limited; there should be a lot more numbers to base an argument around after this weekend. ~ Capric 04:21, 5 June 2012 (UTC)

I'll be playing and taking notes this coming BWE, and I'd be happy to share what I learn. Is simply recording the tooltip popups for skills sufficient to be useful, or do I need to experiment in combat? Also, I suspect I should repeat the gathering with each level up. Are there other times I should look? Dr Dread 05:21, 6 June 2012 (UTC)

For myself, I'm only concerned with the tooltip numbers, since that's what we can easily document on the wiki. Capric here is definitely interested in the "acutal" numbers, though, and I'm sure he wouldn't mind some help with that.

If you want to record numbers after every level up, that would be great! For at least one of those levels, also record with a different Power value (spend trait points in the line that increases Power or swap out an item with a Power mod). And finally, wiki-mail me so I can add your email to the list of editors (I'd rather not open the spreadsheet to anonymous editing). —Dr Ishmael 12:30, 6 June 2012 (UTC)

Thanks. I think it best if I leave editing the spreadsheet to others, but I will collect data. If I can play that is: my beta client crashes instantly on start up. :( Hoping it gets fixed before the event. Dr Dread 04:13, 7 June 2012 (UTC)

Sure, that's fine. I recommend you still use the spreadsheet as a guide to make sure you cover all the skills (and your character's stats/traits), then send me your screenshots or video and I can enter the data for you. —Dr Ishmael 04:22, 7 June 2012 (UTC)

I agree that tooltip numbers are the most important and what should be primarily documented on the wiki. By my thinking, "actual" damage numbers are useful in three cases:

• The tooltip damage is ambiguous - such as with multi-hit abilities, does the tooltip damage reflect the aggregate damage from all hits, or does it reflect the individual hit damage?
• The tooltip does not list a damage yet the skill still does some.
• The tooltip damage is wrong. Some abilities, like Illusionary Mage, have tooltip damages that are completely unrelated to anything, so far as I can tell.

Hopefully, all three (and especially the latter two) of these cases will become more and more rare as the beta process continues. ~ Capric 14:46, 6 June 2012 (UTC)

Whoa

Based on what I'm reading in the AMA, it sounds like there have been a LOT of changes to skills. Engineer and Mesmer have both been overhauled, and a lot of other specific changes were mentioned (ranger sword skill 1, guardian greatsword skills, etc.). Probably 3/4 of the spreadsheet is gonna be irrelevant. :( But I'm glad I at least got the framework set up beforehand. I think what I'll do is lock all of the tabs from editing, so it's obvious that people shouldn't start adding data yet, and then unlock them as I get them updated. I'll work on the updating as fast as I can after the beta starts. —Dr Ishmael 22:23, 6 June 2012 (UTC)

Bleh

I managed to totally mess up my fraps clips. I have like 6 necro tooltips readable, the rest is just o.O what did i film then? I sure hope someone else wasless blonde than i was. --Ee 18:17, 11 June 2012 (UTC)

Between Capric and myself, we got all professions covered with at least 4 data points in PvP (he did 5 on elementalist and mesmer, I only did 4 on the rest). Since you were probably in PvP to do that anyway, it's not a huge loss. But I am sad that your videos messed up and we won't have shiny skill animations on the wiki anytime soon. :( —Dr Ishmael 20:12, 11 June 2012 (UTC)