(Undocumented) Game mechanics discussion

So, not all game mechanics are properly documented in many games. Is weapon speed normalized for attacks? How about for chance on hit procs? How do AP/SP and crit relate to actual dps for a given build? How much of each should one have, and in what ratio? When does AP become more important than Crit, and vice versa.

To answer such questions, some people go out and test mechanics, gather data, and formulate an answer. Sometimes this is easy and general (weapon speed IS normalized for attacks) and other times it is hard and build specific (AP/SP vs Crit questions).

So...
Do chance on hit effects, such as poisons, proc more often with faster weapons, or are they normalized to a procs per minute system (like WoW)?
I did 2 ~= 30 minute parses on a normal dummy, just auto attacks and Virulent poison. Used "Tin Sword"s and "Tin Dagger"s.

The data, PPM/proc chance rounded to 3 significant digits:
Swords
1932 swings
342 procs
2089 seconds
9.82 PPM
17.7% chance to proc

Daggers
1828 swings
326 procs
1702 seconds
11.5 PPM
17.8% chance to proc

Thus, weapon speed is not normalized for chance-on-hit mechanics.

I'll try to test other such mechanics as I have the time.

Does the parry cap being 20% from Parry Rating count?

Dodge is 20% as well. Found out that the only way to queue as support as a Cleric is to have a role with at least 22 Justicar in it (seem to remember someone asking that the other day).

"1932 swings 342 procs 2089 seconds 9.82 PPM 17.7% chance to proc Daggers 1828 swings 326 procs 1702 seconds 11.5 PPM 17.8% chance to proc Thus, weapon speed is not normalized for chance-on-hit mechanics." I'm a tad confused. To me, an engineer, I feel like 17.7% and 17.8% IS the same number. With anything there's some errors in data collection and so most data is reported as, for instance, 17.7% +/- 0.2%. The error I just imagined is only 1%, pretty darn low. Even with that low of an error spread, 17.7~17.8 and you can safely assume they are the same number. The 0.2% can be decreased if you do a lot more samples. So if 17.7~17.8%, doesn't that mean "weapon speed IS normalized for chance-on-hit mechanics"? Or at the very least, "it appears weapon speed is normalized..." Part of this question is about statistical analysis and the other half is me just trying to understand your conclusion.

» Edited on: 2011-06-08 13:22:01

in other words, if all other things are equal except the proc it's better to use faster weapons since the proc will occur more times in the same period of time.

Ah I think I understand now. By normalized, you mean it should cause the same damage over the same time period regardless of how fast the weapon attacks in the same time period. IE most players don't want to see a dagger proc'ing more times in one minute than a really slow 2h hammer just because the dagger hits more often. (Granted the guy with dagger wouldn't mind.) In this case, I'd would definitely agree, the proc isn't normalized unless they make all weapons attack at the same speed.

Yep. "Normalization" would drop the proc chance for the faster weapon so the poison would do the same DPS. Since the proc chance stays the same it isn't normalized. Just did tests on the values of AP vs crit for my 44 Sab/20 Assassin/2 Ranger build at range, using the "5x spike, detonate, 5x shrapnel, detonate" rotation. Over a 5 minute parse with no buffs I had the following damage breakdown according to ACT: 33% spike 21% HE 21% shrap 13% Det 6% auto 4% charge Booster 1% vir poison 1% gibbous shadesource 0% Lethal poison I tested 5x of each charge, and recorded damage on crit vs the boss dummy. I then used a legendary scroll of strength (+40 str ie +20 AP) I repeated the crit damage tests. (BuffedPowerCrit - BaseCrit)/CritMultiplier/PowerBuffAmount = PowerCoefficient Power Coefficient = how much damage is added by adding 1 power. 410 AP -> 430 AP 920 spike crit -> 926 Spike coefficient = 0.167 1007 Detonate crit -> 1025 Det coefficient = 0.5 149 auto attack -> 149 Auto coefficient = 0 617 High Explosives -> 617 High Explosives coefficient = 0 406 Shrapnel -> 410 Shrapnel coefficient = 0.11111 0.143 normalized damage per AP by damage %s from 5 min parse. 1 crit rating adds 0.38% chance to crit, flat. So my average damage for all the above abilities after multiplying by the total % damage they represent was 366. 366*1.8*0.00038 = 0.25 dmg/crit point So to calculate the value of a piece of gear: Damage = 0.143 × AP + 0.143 × (Str + Dex) / 2 + 0.25 × Dex + 0.25 × crit