Please upgrade your browser for the best possible experience.

Chrome Firefox Internet Explorer
×

Dear Developers, Accuracy Considered Harmful

STAR WARS: The Old Republic > English > Classes
Dear Developers, Accuracy Considered Harmful

KeyboardNinja's Avatar


KeyboardNinja
09.14.2012 , 11:02 PM | #1
Let's look at the breakdown in a recent downing of HM Firebrand:

Spoiler


First column is the number of times an ability was activated (or ticked). Second column is the ability, third column is average damage, and the fourth column is the miss percentage. And yes, I do use Double Strike a lot. I'll post a new thread soon detailing why, don't bother hijacking this one. :-)

Here's the important thing to realize: discounting Saber Strike, almost nothing missed to a significant extent. This is with only 91% accuracy (the 1% is from maxing my companion). We can determine the DPS we lost according to the following formula:

dmg / (1 - miss%)

Thus:
  • Saber Strike: 238.11 ==> 0.647 DPS (1.294 TPS)
  • Double Strike: 526.21 ==> 1.430 DPS (2.860 TPS)

Of course, we're forgetting that Double Strike procs Particle Acceleration. Thus, we need to fudge up the DPS loss by 13.77% and the TPS loss by 15.51%. (drawing on the numbers I'm assembling for my longer post on Double Strike) This gives us a net DPS loss (from Double Strike misses) of 1.627 and a net TPS loss of 3.304.

Neither of which are game-breaking, or indeed even noticeable in the face of all of the other statistical variance we see in a raid boss that is more dynamic than a straight tank-n-spank. Even looking at things from a "non-variability" of threat and damage, we're still not losing anything noticeable. A miss on my Double Strike means a very small, momentary loss of threat on a short rolling average. Even multiple misses in a row isn't a serious issue. 4.55% of 88 hits is exactly 4 misses, which is two GCDs. Assuming the worst possible RNG, that's 3 seconds of zero threat. Chances are, I still won't notice it. In the worst case, I have to pop a taunt.

Looking at this another way, lets look at the abilities I used on Firebrand and their exact accuracy (note that I have absolutely no points in Accuracy on my gear):
  • Force Breach: 101%
  • Combat Technique: 101%
  • Spinning Strike: 96%
  • Saber Strike: 91%
  • Slow Time: 101%
  • Double Strike: 96%
  • Crushed: 101%
  • Project: 101%

Note: Double Strike and Spinning Strike are both "special attacks". Among other things, this means that they get a bonus 5% accuracy boost. This is why they have 96% accuracy overall.

As you can see, even without any accuracy whatsoever, nearly all of my attacks are guaranteed hits, and the ones that aren't are still at a very high probability. Adding accuracy would improve this, certainly, but not in a way that's actually worthwhile.

At my current gear levels, 50 more points in shield rating is worth 0.3693% m/r mitigation, which is about 7.936 DtPS on HM Kephess (much higher on high damage bosses like Foreman Crusher or Fabricator). 50 points in accuracy is worth 1.7933% accuracy, which is less than a third of what I would need to make my Double Strike hit 100% of the time. Thus, playing a bit of rough napkin math, we can say 50 points of accuracy is worth 0.858 + 0.1941 = 1.0521 TPS.

Now, 8 DtPS is almost nothing, but I think we would all agree that trading 1 TPS for 8 DtPS is significant improvement.

Every tank is different, no question, and Guardians/Juggernauts are certainly going to be more accuracy-dependent than Shadows/Assassins and Vanguards/PowerTechs (due to a lower percentage of special attacks). However, I think these numbers show pretty conclusively that even using a Double Strike-heavy rotation (which is more accuracy-dependent than the popular low-Thrash variant), accuracy is an extremely low-value stat for tanks. It's not worthless, but very close to it.
Computer Programmer. Theory Crafter. Streaming Dilettante on The Ebon Hawk.
Tam (shadow tank) Tov-ren (commando healer) Aveo (retired sentinel) Nimri (ruffian scoundrel)
Averith (marksman sniper) Alish (lightning sorcerer) Aresham (vengeance jugg) Effek (pyro pt)

SamuelAU's Avatar


SamuelAU
09.16.2012 , 08:44 AM | #2
Nice article. I've never really understood the need for accuracy on my tanking gear.

On a side not, I'm interested to see why you use double strike so often. I also use to it activate particle acceleration but not with the frequency that you do. I'm also wondering whether you use harnessed shadows and telekentic throw at all, but that's not really relevant to the thread.

Thanks for sharing!

KeyboardNinja's Avatar


KeyboardNinja
09.16.2012 , 09:00 AM | #3
Quote: Originally Posted by SamuelAU View Post
On a side not, I'm interested to see why you use double strike so often. I also use to it activate particle acceleration but not with the frequency that you do. I'm also wondering whether you use harnessed shadows and telekentic throw at all, but that's not really relevant to the thread.
I'll definitely go into that in another post. Short answer: I absolutely use Telekinetic Throw. Cutting that from a shadow tanking rotation would be absolutely unthinkable. The frequency on Double Strike is actually not quite as much as it appears, since I only use it at >55 Force. My Project is delayed slightly, and therefore also my Telekinetic Throw, but not by much.

I'm really just hopeful that the developers will take note of the community's efforts in the area of theory crafting w.r.t. accuracy. The new Dread Guard gear is just silly, and they need to change it. It almost seems to stem from a misunderstanding of their own game, since assertions like "10% accuracy is beneficial to everyone, including tanks" are demonstrably false.
Computer Programmer. Theory Crafter. Streaming Dilettante on The Ebon Hawk.
Tam (shadow tank) Tov-ren (commando healer) Aveo (retired sentinel) Nimri (ruffian scoundrel)
Averith (marksman sniper) Alish (lightning sorcerer) Aresham (vengeance jugg) Effek (pyro pt)

Kitru's Avatar


Kitru
09.16.2012 , 09:03 AM | #4
Quote: Originally Posted by SamuelAU View Post
Nice article. I've never really understood the need for accuracy on my tanking gear.
There isn't any "need" for accuracy on tank gear. The devs know this, or at least should. The only reason I can imagine why the developers put it on tank gear is because they would feel weird putting a more useful DPS stat on there (since everyone wants to hit more often, even if it's got a truly terrible return rate compared to *absolutely everything else* and is the only stat that can actually be rendered completely redundant, right?). I can only assume that the developers are still using itemization tables and concepts that operate under the beta assumption that redudant accuracy provides increase arpen (which, honestly, would make accuracy somewhat useful for tanks and other casters since it would increase K/E damage and provide for a direct tank anti-stat).

Quote:
I'm also wondering whether you use harnessed shadows and telekentic throw at all, but that's not really relevant to the thread.
TK Throw doesn't list as "Telekinetic Throw" in parses. It lists as "Crushed". You'll also need to keep in mind that the 59 uses are listed as the individual ticks (of which there are 4 per TkT full channel). As such, he used TkT at least 15 times during the fight, which groks with the 22 Slow Times and ~24 Project uses (with Upheaval, you'll see 1.45 times as many Projects as you actually use).

Honestly, I would be curious to see the reasoning behind the Typist-Assassin's massive use of Double Strike though (which would also explain the very heavy use of Saber Strike). Personally, I don't really see much point in it. I net better damage and survivability out of largely ignoring it.
Walls of Text? I *love* Walls of Text!
My New Class Idea
Shadow Class Rep - Suggest/Review Questions Here
Quote: Originally Posted by Fende View Post
Listen to Kitru. Kitru knows all.

KeyboardNinja's Avatar


KeyboardNinja
09.16.2012 , 03:22 PM | #5
Quote: Originally Posted by Kitru View Post
Honestly, I would be curious to see the reasoning behind the Typist-Assassin's massive use of Double Strike though (which would also explain the very heavy use of Saber Strike). Personally, I don't really see much point in it. I net better damage and survivability out of largely ignoring it.
I'm still working through that reasoning for a subsequent post. The glib answer is that I've actually tried both rotations (the commonly-glorified low-thrash, and my non-strict PA rotation), and I find in parses that my rotation gives me more DPS (by about 12%), more threat (by the same scalar), and about the same amount of HPS as low-thrash. That's a glib answer because I could just be not executing correctly on low-thrash, but it's where this whole quest started for me.

Quick note: Saber Strike is a flurry of 3 hits, so I actually only used it 28 times. I generally see a much heavier use of Saber Strike when I try low-thrash, since I'm constantly saving up to allow Project on CD.

The longer answer is that I don't use Double Strike blindly. The truth is that on this particular fight, I was spending a lot of time making corrections to the raid mid-combat, so I actually derped my rotation quite a bit. Whenever I derp my rotation, it shows up as over-use of Double Strike (and poorly-timed use of Project). My priority queue cuts out Double Strike when below 55 Force, or below 40 Force if FP is active. If it's a frequently-hitting m/r boss (e.g. Foreman Crusher), I go down to 45 Force. If it's a force/tech boss (e.g. Soa), I almost drop Double Strike altogether (closer to 70 Force is the minimum).

I haven't finished working out all of the math here, but the preliminary results indicate that a lot of the early math which devalues Double Strike was actually failing to properly account for some of the indirect procs. That math also seemed to assume only 90% accuracy on the ability, which is 6% shy of what it actually has. The threat gen from Double Strike is actually quite good (145.7339 TPF, compared to just 95.3759 TPF for Project). So far, it really looks like the only thing you give up in using it (carefully) is a slightly delayed 3 stacks of HS (in practice, delayed by about 2 cooldowns per 3-stack).

Anyway, as I said, I haven't finished working through all the math yet, so I could be missing some things. So far though, the numbers I'm running seem to agree with my practical experience, which is that Double Strike *is* a force dump, but not as valueless as everyone seems to think.
Computer Programmer. Theory Crafter. Streaming Dilettante on The Ebon Hawk.
Tam (shadow tank) Tov-ren (commando healer) Aveo (retired sentinel) Nimri (ruffian scoundrel)
Averith (marksman sniper) Alish (lightning sorcerer) Aresham (vengeance jugg) Effek (pyro pt)

CitizenFry's Avatar


CitizenFry
09.16.2012 , 08:57 PM | #6
I wouldn't say it's actually harmful so much as it is valueless. But yeah, I see similar things on Blayze. Here's my most recent HM EC run, which jibes with basically every other parse I've ever looked at:
http://swtor.askmrrobot.com/combatlo...ayer/6#d=0,b=1

Zero accuracy rating, I miss about 9% of my High Impact Bolt across the whole raid, and one or two hits out of every Hammer Shot (which is just filler anyway). Nice stat.
Crafting on The Ebon Hawk. PST/mail: Blayze (rep) / K'ron (imp)
Armoring: Commando 26, Reflex 26; Barrel: Patron 26
Mod: Weighted 26+26B, Deflecting 26 Enhancement: Bulwark 26, Battle 26, Finesse 26
Ear: Black Hole Mender's MK-2 Relay

KeyboardNinja's Avatar


KeyboardNinja
09.16.2012 , 10:51 PM | #7
Quote: Originally Posted by CitizenFry View Post
I wouldn't say it's actually harmful so much as it is valueless. But yeah, I see similar things on Blayze. Here's my most recent HM EC run, which jibes with basically every other parse I've ever looked at:
http://swtor.askmrrobot.com/combatlo...ayer/6#d=0,b=1

Zero accuracy rating, I miss about 9% of my High Impact Bolt across the whole raid, and one or two hits out of every Hammer Shot (which is just filler anyway). Nice stat.
It's harmful in that it reallocates stat budget to something which is, as you say, valueless. In any case, the title was a quote from Edsger Dijkstra (a famous misquote, actually).

In any case, accuracy really is quite silly. A 9% miss rate on HiB is almost getting to be annoying, but still not to the point where I would consider it a problem. It's certainly not worth sacrificing survivability to fix.
Computer Programmer. Theory Crafter. Streaming Dilettante on The Ebon Hawk.
Tam (shadow tank) Tov-ren (commando healer) Aveo (retired sentinel) Nimri (ruffian scoundrel)
Averith (marksman sniper) Alish (lightning sorcerer) Aresham (vengeance jugg) Effek (pyro pt)

Dhariq's Avatar


Dhariq
09.17.2012 , 02:28 AM | #8
Good and interesting thread. I felt like chipping in but when thinking about it, it just made me tired. Thanks guys for showing with some math what everyone with a clue knows by heart. The only problems I see with it is:
a) how can Bioware have senior developers who does not understand their own game to begin with?
b) given that they do not understand; they have never, ever acted on constructive math heavy explanations, ever. I fear all this will just fly over their head unfortunately.

Kitru's Avatar


Kitru
09.17.2012 , 03:29 AM | #9
Quote: Originally Posted by Dhariq View Post
a) how can Bioware have senior developers who does not understand their own game to begin with?
b) given that they do not understand; they have never, ever acted on constructive math heavy explanations, ever. I fear all this will just fly over their head unfortunately.
First off, you're predicating the conclusion of your second point upon the first point being true when there are substantially more likely explanations. BW *knows* how the various stats operate. If they didn't, the tanks wouldn't be as balanced as they are now (they're within a tiny margin of incoming damage difference). The more important question to ask is why, if the developers *know* about the terrible contribution of accuracy, which is both more likely and more sensible, why they keep putting it on gear?

Secondly, the developers are substantially more likely to respond to math intensive explanations and suggestions than they are to anecdotal evidence or simple player protest. Math is something concrete that can be proven or disproven. Every developer in every game that I have ever played has responded to math when the math is appropriate and a legitimate suggestion is given. Since developers make most balance decisions based upon the accuracy and legitimacy of in-house modeling, using math is an excellent way to demonstrate with a high degree of certainty that whatever model the developers were using is flawed.

Now, the models the developers are using for tanks and mitigation are relatively spot on. If they weren't, the cross class mitigation balance between the tanks would be off, like they were when the game was first released before the various tank buffs that brought everything where they are now (a lot of the reason for the tweaks to the developer models is actually due to the *loads* of mitigation and tank comparison math that was done within the first couple months of the game's release). The only real problem that *could* be within the models is that the developers could be using a threat generation model that overinflates the value of accuracy for any number of reasons.

So either the problem is that the developers are using a fundamentally flawed model for tank damage and/or threat gen (and potentially for DPS threat gen as well which might indicate a reason why larger discrepancies between DPS classes exist in anecdotal and parse based evidence than the developer statements would suggest) or the developers are purposefully putting sub-standard stats on tank gear (which rewards gear customization but makes the baseline gear less useful, especially when gear only becomes useful when customized, as was true with Rak gear and as what looks like is going to be true with Dread gear). In the former, it's a problem that can be fixed with math; in the latter, it's pretty much WAI that can only really be addressed by demonstrating that the differentiation between baseline and properly customized gear is a larger discrepancy than should be allowed (which, honestly, would be easier if we could get a decent baseline for how large of an increase the developers *want* to have from customization as opposed to having to wonder while different tiers vacillate between marginal and massive improvements from customization).
Walls of Text? I *love* Walls of Text!
My New Class Idea
Shadow Class Rep - Suggest/Review Questions Here
Quote: Originally Posted by Fende View Post
Listen to Kitru. Kitru knows all.

Dhariq's Avatar


Dhariq
09.17.2012 , 04:10 AM | #10
Quote: Originally Posted by Kitru View Post
First off, you're predicating the conclusion of your second point upon the first point being true
Well I read that they agreed to that Rakata was badly itemized. This was changed for Campaign. Now a "senior" designer state facts which they already smacked themself as something bad. So has the mechanics of the game changed? No. So he is a senior designer who is clueless and wrong both according to Bioware and just about everyone here. So yes I base it on the first statement beeing correct.

Also I have never seen them care anymore about cases with math. I have never ever even seem them reply to them in anyway or discussed them. Oh, sorry I saw one thread once when one guy were stating some crap that were not correct. So I do not understand how you can type something like "substantially more likely" and try to push it like a fact. It is an oppinion based on nothing more than that you seriously hope that this should be the case as it should be because what they are doing is just horrible.

20 guys screaming OP NERF without anything but a screenshot of a wz summary in the pvp forum are alot more likely to get some knee jerk out of this company than some brilliant facts will ever do. Not unique for BW but still worthless. This is what I would post as facts since they are based on empirical evidence not on how the world should work.