User talk:RolandOfGilead13141

How do you know all that skill data for level 18? are you testing it at the proper shrines? Foo 08:34, 9 September 2006 (CDT)
 * Oh ok, now that I see the attributes for the level 18 and 20, it seems you do. gj. Foo 08:36, 9 September 2006 (CDT)
 * No I don't gather hard data for the highest 2 levels. However, skill effects seem to follow a strictly linear curve depending on its linked attribute's level, taking GW's rounding method into account. For levels 0-16 I have tested this theory extensively and have not yet found an exception to it, so I have full confidence that the theory is correct and can be applied to levels 17/18 and 19/20, respectively. RolandOfGilead 08:49, 9 September 2006 (CDT)
 * ok, that sounds better then just guessing numbers. can you say more about it? Foo 09:06, 9 September 2006 (CDT)
 * Sure. The rounding method applied in GW is: x.5 and greater is rounded up, otherwise down. For example, 5/2 = 2.5 = 3 in GW. There are exceptions, but AFAIK they are alwas explicitely stated.


 * The linear curve I mentioned is the basic equation describing a line in a Cartesian coordinate system:


 * y = mx + c, or, for our purposes,


 * effect = effectIncrease*attributeLevel + offset


 * where effectIncrease is the increase per level and offset simply the value at level 0 (but before rounding, so offset can be a fraction, even though a fraction is never displayed to the player).


 * For example, if the skill says "target foe is poisoned for 5-17 seconds", then effectIncrease = 1.0 and offset = 5.0. This is trivial if the increase is a whole number. But what about fractions?


 * Example: Life Siphon effect duration, from level 0 trough 16, and the increase per level below it:


 * 12 13 14 14 15 16 17 18 18 19 20 21 22 22 23 24 25


 * N/A +1 +1 +0 +1 +1 +1 +1 +0 +1 +1 +1 +1 +0 +1 +1 +1


 * Do you see the pattern in the lower line? The increase is always 1/1/0/1/1. You could also say it is 0/1/1/1/1 or 1/0/1/1/1, as long as the pattern returns without fault. Either way, now we know that the effect will increase by 4 every 5 levels. Thus, effectIncrease = 4/5 = 0.8. The only thing left to do is to find out the exact value of offset, which can be determined by what my Maths teacher would've called "a sharp look". In this case, offset is 12.0. Now we can calculate the duration of Life Siphon for an arbitrary level. Hope that helped. RolandOfGilead 10:01, 9 September 2006 (CDT)


 * Hehe, maybe I should have said I'm a mathematician first. it is still a guess, an assumption that the line has nice and round values. how short should a pattern be so that you will consider it reliable? there was a discution about it already, and I think it said that we will only put in values confirmed in-game. (energy cost, after experties and other effects is always rounded down). Foo 11:40, 9 September 2006 (CDT)
 * As long as we don't get to see the source code of the game, every formula and every rule with predictive power the community deduces about game behaviour is an "assumption", unless confirmed by ANet, so "assumption" is not a very useful term to begin with here. But, staying with the term, all assumptions are not alike; there are good ones and bad ones. The good ones are founded on large amounts of data, can explain all data and are not contradicted by even one piece of data.


 * The theory of strictly linear skill progression is a good assumption in the above sense, because it fulfills all these criteria. Feel free to disagree, but to the best of my kowledge (and I have put a lot of time into this), every single one of the hundreds of skills in GW follows the linear curve in levels 0 to 16, which is 80% of the 0-20 range which currently matters. I do not exaggerate when I say that I have checked this behaviour for every (!) unique skill range in Prophecies skills, i.e. 5-17, 20-50, 1-10, etc. etc. ad infinitum. If you want proof, I can upload a csv-file which is readable by the spreadsheet program (excel, OOo calc, etc.) of your choice. As long as there is no shred of evidence contradicting the theory, it would be plain unreasonable to dismiss it in favour of dull data collecting without generalization, just because "it could be different just 1 level further". Yeah, but so what?. All evidence points towards linear progression without any deviation and none speaks against it, so why not go with it? If it should turn out wrong later, ok, let's dump the theory with or without replacement, that's how it goes, no problem at all. But until then, this theory provides an equation which not only correctly describes all known skill values, but also has potential predictive power for non-observed level ranges. Who is afraid of a good assumption which has a non-zero, but imho small, chance of being wrong? RolandOfGilead 13:03, 9 September 2006 (CDT)


 * Progressions are linear through the values for 0 and 15, as in a line through (0, f(0)) and (0, f(15)). Round non-integer results.  Using 0 and 16 doesn't work (Aura of Restoration is one example). See Talk:Skill_details/Archive.  --Fyren 15:46, 9 September 2006 (CDT)
 * Ok, I stand corrected, thanks for discussion. You could've saved me all that by just linking to the formula :) Anyway, at least I was right about the linearity of the function, because the one in Skill details is linear as well, only the coefficients are a ittle different. For example, for Aura of Restoration the equation is y = 152 + x*16.5333333, versus my y = 152 + x*16.5. Obviously the difference is very small but big enough to be noticed, if only in a very few cases. Also, if (v15-v0) is a multiple of 15, then my approach gives the same results, so my method is really a special case and an approximation. RolandOfGilead 17:42, 9 September 2006 (CDT)


 * I'm not Foo (if you're mistaking me for her, heh). I just happened to see your discussion with her so chimed in.  I also have some wiki templating written up that auto-generate progressions, but I don't know if we're going to use it.  See, for example, Sandbox/Symbiosis and the source for it.  The values for 0 and 15 are used to generate the progression table automatically, so hopefully we can ditch all this manual progression data crap.  (We didn't do this from the start since ParserFunctions are needed to do the math, and PF weren't installed till a few weeks ago.)  --Fyren 18:12, 9 September 2006 (CDT)


 * "her"? how did you get to that conclusion? Foo 18:20, 9 September 2006 (CDT)
 * If you insist, I'll use "it" in the future. --Fyren 18:22, 9 September 2006 (CDT)
 * :O Foo 18:25, 9 September 2006 (CDT)