Has anyone seriously parsed AC?

[pasi]

Originally Posted by heartbrand [You must be logged in to view images. Log in or Register.]

for tanking that's what I usually do. but I guess the crux of my post was more, does anyone know if AC has a cap here, what the hp > ac value approximation is etc. usually in other MMO's such as WoW mitigation is far greater than EHP [effective health]. do people find that to be the same case here? On EQ Live a number thrown around was 1 AC = 5 HP.

Hey pal.

The 1 AC = 5 HP figure was thrown around for an era that has some key differences from Project1999.

First off, in this era, Complete Healing is a full heal. This changes later on when CHeal goes down to 7500, and you can start to break 7500 HP. For practical purposes, CHeal is also the only strong heal in raiding this era.

In the long run, AC reduces damage taken and adds a small amount of survivability to NPC bursts via a left skewing DI curve. The issue is that your healing is coming through a CHeal chain. The goal of the CHeal chain is to keep the tank alive through damage spikes. While AC will slightly change the probability of damage spikes, your goal is to be spike-proof, not reduce overall damage intake. Slight reductions in damage taken matter a lot more when healing changes from CHeal chains to Light spamming.

With CHeal chains, the healing thrown at you and cleric mana expenditure are the same regardless of your damage mitigation. In other words, it doesn't matter if Tank A has 200 more AC than Tank B - you're still going to be running the same CHeal interval because you are playing around NPC max damage rounds.

So, you would be comparing AC versus HP for the sole purpose of surviving damage spikes. Furthermore, defensive discipline actually de-values the effectiveness of AC relative to HP as the DI is halved through defensive.

So, if your goal is raid tanking, HP is far, far superior in this era than AC.

With that said, it doesn't really matter. The high HP gear is almost all gonna have shitloads of AC (with hammered golden hoop being an obvious offender). It's not like there are augs or anything where you're distributing stat points. You can argue that Kael (AC) vs Skyshrine (stats) is a choice, but the reality is that you're just going to grab the first one that's available.

If I had to ballpark a conversion value, I would do 1 AC = 2 HP for the purposes of raid tanking.

[zanderklocke]

Maurice. Why the fuck are you not playing anymore? You're the hero this server needs.

[pasi]

I miss you gents, but I can't play EQ for more than 20 minutes without getting bored. Even Velious beta just feels like content that I've done a million times before. So, my free time is mostly going to WoW/HS.

I had a good time spending a day making those Velious videos though. Maybe I'll try to do something along those lines.

[Raev]

Originally Posted by pasi [You must be logged in to view images. Log in or Register.]

So, you would be comparing AC versus HP for the sole purpose of surviving damage spikes.

But AC is actually quite effective at this. I remember writing a simulator and expecting to find exactly the result you were suggesting, that AC sucks compared to HP. I coded up something really quick and dirty:

(defun simple-attack (min max min-rate interval-rate)
  (if (< (random 1.0) 0.35)
      0.0
      (let ((r (random 1.0)))
	(cond ((< r min-rate) min)
	      ((< r (+ min-rate interval-rate)) (+ (* (random 1.0) (- max min)) min))
	      (t max)))))

(defun test (min max min-rate interval-rate)
  (/ (loop for i from 1 to 100
	summing (loop for i from 1 to 100
		   maximizing (loop for i from 1 to 30
				 summing (simple-attack min max min-rate interval-rate))))
     100))

with the following results

CL-USER> (test 106 350 0.01 0.60)
7282.469
CL-USER> (test 106 350 0.01 0.75)
6831.041
CL-USER> (test 106 228 0.01 0.60)
5008.053
CL-USER> (test 106 228 0.01 0.75)
4770.086

Those are from my previous Spiroc Lord parse for Sakuragi. Sakuragi with a bit of resist gear is probably 175 item AC including the iksar bonus vs Argh/Bob probably rocking around 225-230. Let's call it 60, so my little 5-line sim suggests 60 AC is worth 230 HP (4:1) while defensive and 450 hp (8:1), although its obviously less b/c no one has 7300 HP to survive. Call it 6:1. However, the Spiroc Lord is an absolute pussy. Lets try something a bit harder . . like Vindi. I had to reduce Vindi to 12 rounds because he hits so much harder (about a 5-6s CH chain in practice)

CL-USER> (test 226 700 0.01 0.60)
6734.8774
CL-USER> (test 226 700 0.01 0.75)
6352.1973
CL-USER> (test 226 485 0.01 0.60)
4843.406
CL-USER> (test 226 485 0.01 0.75)
4594.7783

Which actually is almost the same. The point is that no one *really* provisions a CH chain for the absolute worst case, because that's very rare owing to Central Limit Theorem. Even over over 100 simulations, Vindi still didn't come close to his theoretical max of 8400.

Of course in Velious most warriors will probably hit Haynar's softcap, whatever it is, further halving or so the value of AC. Which brings us to 2:1 again.

P.S. HI MO

P.P.S. No I'm not going to explain these numbers

[pasi]

You could take your simulation a step further and compare AC's damage reduction and the 'additional healing' provided by having X extra HP with Y number of CHeals hitting. I think that would give you a more accurate number on what your simulator is looking to do.

However, I think you're missing what I'm saying here.

What I'm saying would look as follows
1) Establish NPC DB + DI
2) Establish Player HP
3) Establish Player AC
4) Set a number for NPC rounds (this would reflect the #rounds between CHeals)
5) Look at the number of times the tank dies.

Going off your numbers for DI and DB. Do 2 tanks, one with 5900 HP and 1200 AC and one with 5500 HP and 1400 AC. You can tinker with these, I just tried to maintain the ratio that I gave. Run the simulation for 2 rounds and for 3 rounds under both defensive and non-defensive. Lastly, record the absolute number of deaths in both groups.

With that said, 0.75 vs 0.60 seems a bit drastic. Thats a 25% difference [You must be logged in to view images. Log in or Register.]. Try something like 0.66 vs 0.6 for a 10% difference.

Although, I think this would make the most sense to do for content that isn't as trivial (as that is the goal of gearing). Most of these NPCs will be set at 1-2 rounds.

[Raev]

This is almost exactly what I did, except rather than measuring the proportion of tank deaths I measured the maximum damage over 100 parses of 30 attacks.

The 0.75 interval hits (25% max hits) vs 0.60 interval hits (40% max hits) come from parses of Sakuragi (in resist gear), Bobbarker, and Argh against the Spiroc Lord.

[maskedmelon]

So can we take away that for AC totals below soft cap without a MIT disc, 1AC is worth 6 to 8 HP? And any AC over soft cap is worth 4HP (again assuming no disc)?

[Haynar]

General rule of tanking in this era.

AC > all. That should simplify things.

H

[pasi]

Originally Posted by Raev [You must be logged in to view images. Log in or Register.]

This is almost exactly what I did, except rather than measuring the proportion of tank deaths I measured the maximum damage over 100 parses of 30 attacks.

The 0.75 interval hits (25% max hits) vs 0.60 interval hits (40% max hits) come from parses of Sakuragi (in resist gear), Bobbarker, and Argh against the Spiroc Lord.

Velious was originally launched with a hardcap in place on live. Meaning, AC was literally worthless for people for players with more than 289 raw AC from gear. Later on in Velious, they put in a softcap with a small return for melee only. It wasn't until the Luclin mitigation/avoidance patch where softcaps gave reasonable returns (along with a host of other changes that I won't get into).

Obviously, that doesn't necessarily apply over here (for good reasons, EQ has enough useless stats). So, I'm going to write the rest of this under the assumption that the server will have an AC softcap with decent returns.

It's a pretty simple idea, you need to be able to reliably survive in the short-run before you should be concerned with reducing damage intake in the long run. I typed up a long post explaining this in detail, but figured I could illustrate this concept in a much easier matter.

Let's take Derakor the Vindicator for example since you brought him up.

Vindicator has a DI of 25 and a DB of 200. Mathematically speaking, evasive discipline will be superior at reducing average damage compared to defensive discipline. However, any intelligent tank is going to be using defensive instead of evasive on Vindicator.

Why is this?

Because with defensive, you can guarantee that a tank can live 3 rounds while defensive. While it is unlikely that an evasive tank will die in 2 rounds, it is a possibility. Obviously, it is a larger possibility with 3 rounds (what I can guarantee a defensive tank living through).

The argument against AC is similar to the argument against evasive. With AC, you're looking at an effect that is seen over the duration of the entire fight. But, in reality, we're really only concerned about tiny sections of the fight (damage spikes) since CHeal is going to 'reset' combat every 1-3 rounds (pending on our rotation). As you decrease the number of rounds, the effects of AC on a smaller number of hits makes for far more variable rounds. With a mob like AoW, you're looking at 1-2 rounds of combat.

Yes, you'll have more of these tiny sections (damage spikes) of the fight with HP stacking, but less of these will matter (be lethal rounds). With AC, you'll have less of these tiny sections, but more of them will matter.

I was one of the biggest proponents of AC stacking on TSW, it's just a very, very dumb thing to do in the Velious era if raid tanking is your goal. With that said, I mentioned item availability in my prior post - it's not like you have much of a choice in stats until augs come along.

[Raev]

I understand your argument quite well: you are saying "well, if +X AC reduces your damage taken by 10% on average, that is inferior to +10% HP, because the worst case damage for the higher AC warrior is the same. Eventually in the CH chain you will hit that worst case". The problem is you are forgetting about two other effects. First, total damage is the sum of IID random variables, i.e. the individual attacks. This means that Central Limit theorem applies, and reduces the variance. Second, AC itself reduces variance by reducing the frequency of max hits. If AC simply reduced the value of max, interval, and min hits by X% it wouldn't be nearly as effective.