dex/crit relation, WS crits, WS gorgets discussion

[Nagamaki]

I was talking about the argument of whether DEX/crit items will affect crit rate on WS that are marked "crit rate varies with tp" or if TP is the only variable, that's why I picked Jin.

There's plenty of ways to test whether WS without crit in the descriptor can crit, that's a different topic.

[Vhess]

Looks like it's dependent upon your DEX - mob's AGI or VIT. If it's VIT it's easy to check, but is there an easy way to check for a mob's AGI?

I'm not sure of a way to determine the mob's AGI, but for testing purposes would it not be possible to use Absorb AGI to assist calculations? The amount of AGI drained would be easy to tell of course, and if you were able to, for instance, get 10% crit rate before using the spell and 18% crit rate after the spell, you would be able to at least determine if AGI is the x-factor and if so where the relative threshold might be.

You'd want someone otehr than the tester to do the absorbing, otherwise the +AGI gained by the caster would throw off the calculations.

[Lhexh]

One other thing pops into my mind which I'm confused about. I've read a lot on both sides about whether or not it's possible to crit on a WS that doesn't have crit in it's descriptions and I don't think its been proven either way whether its possible or not. If in fact you cannot crit on a normal WS how can we say with any certainty that DEX even affects crit WS and that TP isnt the only affect whatsoever.

If you weren't able to crit on a normal WS, stacking SA with WS wouldn't work. Any WS that SA affects can potentially crit.

I'm not sure that it's as easily applied as that. For the longest time I was fairly certain any WS could crit, I mean why couldn't it. It didn't make any sense to my that wouldn't be possible. After reading a few threads on KI and referenced here as well ( can't seem to find them at the moment but looking around since I'm bored as hell at work ) I started to question it myself.

The difference in damage between Metatron Torment sneak attacked as WAR/THF vs using it as WAR/NIN is fairly sizable and ever since Salvage came out I've been noticing just how consistently mediocre it is. When I have berserk up I do consistently decent damage and with berserk down I do consistently crappy damage. With a base 10% crit rate I should at least see occasional spikes in damage since its a 1 hit WS but it just never happens. Maybe I have shit luck and just never roll a crit with berserk down, or maybe crit's on a normal WS just behave slightly differently than a forced crit from SA. I'm really not sure either way but it's fairly time consuming and tricky to test.

[Ferien]

I was talking about the argument of whether DEX/crit items will affect crit rate on WS that are marked "crit rate varies with tp" or if TP is the only variable, that's why I picked Jin.

There's plenty of ways to test whether WS without crit in the descriptor can crit, that's a different topic.

I didn't know there was still an argument about non-crit WSes being able to crit. Also, you can use the same baseline with crit WSes as for non-crit. They're all affected by Sneak Attack, as they're all standard physical WSes.

Proving that Crit WSes use your standard melee rate as a baseline would be a pain in the ass, as you'd have to restrict your wses to a single hit, either by using single-hit crit WSes (Power Slash, Vorpal Scythe, Vorpal Thrust, Keen Edge), or by fighting mobs that you will one-shot. You'd then have to pool all WSes together and take a guess at how often you critted on them, and then do the test all over again with an item or weapon that increases crits.

I think it's generally accepted though that these WSes use your standard melee base as their baseline, so it's safe to assume that adding DEX would work...

[Lhexh]

Actually just found one of the threads here outlining the argument and it brought up an interesting argument.

http://bluegartr.com/forum/viewtopic...ical&start=390

The person arguing against a crit being possible on WS tested Viper Bite as RDM/THF on a mob where the first hit killed the mob every time and claimed he experienced 0 crit on WS despite a 25% crit rate on normal melees. Now the counterargument was the damage ranges on Guillotine supporting the idea that it does indeed crit on WS. It sort of brings up the idea of 3 possibilities

1) First hit ( or first 2 hits when dual wielding ) are calculated differently and cannot crit on a normal WS but the remaining hits afterwards can crit.

2) The pdif variations on each hit can account for that large of a range in damage to make one think crits are occuring.

3) The test itself was flawed.

I guess this has kind of gotten off track but there are really two different scenarios that play out.

1) Do the DEX tiers affect crit WS like Rampage/Jin/etc or is the TP the only modifier. In this case we need to test either on a mob where first hit kills or a mob weak enough that all hits land but still have capped attack/crit rate on. I think the second test is probably better since we can't be sure if first hit isn't calculated differently than subsequent hits.

2) If a normal WS can crit, do the DEX tiers affect them as well. Some weapons can easily cap fstr because of a low weapon rank, and if they don't have STR modifiers ( quite a few of the relic WS have some pretty shitty modifiers ) then it might be viable to stack DEX in other slots to try to improve damage on the single hit WS with more crits.

[suji]

1) First hit ( or first 2 hits when dual wielding ) are calculated differently and cannot crit on a normal WS but the remaining hits afterwards can crit.

2) The pdif variations on each hit can account for that large of a range in damage to make one think crits are occuring.

3) The test itself was flawed.

I guess this has kind of gotten off track but there are really two different scenarios that play out.

Some alla thieves have tested this a bit with Dancing Edge. It was really clear that the first hit of Dancing Edge cannot crit without SA or TA. In the results, there were 7 out of 99 attempts where the first hit missed and one of the later hits killed the mob instead (1tp return). None of those 7 crit either, but 7 is a monumentally shitty sample size. It's just that it takes a long ass time to kill 700 rabbits in Ronfaure, so that number hasn't been refined at all.

The pdif variation doesn't really matter in this case because it's easy to calculate the min/max range for what a normal hit vs a crit would be.

[Keno]

What you all seem to be missing during this little orgasmfest for DEX builds and crit rate is that Haste/Attack/Acc setup while eating meat will still out preform/out parse these DEX builds. Yes you might have a 30% crit rate but the guy next to you is still hitting harder and faster. These DEX builds just aren't practical unfortunately.

What you seem to be missing is that with DEX merits and the right pieces of gear (Commander's Cape, Thunder Ring/Rajas Ring, Dex+2 earrings on each ear, spike necklace [if you have to]) you can possibly break the crit tier while still KEEPING YOUR ENTIRE HASTE SET ON. So you're choosing between a haste/acc/attck/str build on TP with a haste/acc/crit build. The crit build will easily outperform if the cap is low enough so that you dont have to mod a ton of gear. And yes, you are sacrificing acc and attck in a lot of slots for marginal DEX, but you get slight acc returns from it so it's not a huge waste, AND all of them put together increase crits by 10%+...

[Lordwafik]

What you all seem to be missing during this little orgasmfest for DEX builds and crit rate is that Haste/Attack/Acc setup while eating meat will still out preform/out parse these DEX builds. Yes you might have a 30% crit rate but the guy next to you is still hitting harder and faster. These DEX builds just aren't practical unfortunately.

What you seem to be missing is that with DEX merits and the right pieces of gear (Commander's Cape, Thunder Ring/Rajas Ring, Dex+2 earrings on each ear, spike necklace [if you have to]) you can possibly break the crit tier while still KEEPING YOUR ENTIRE HASTE SET ON. So you're choosing between a haste/acc/attck/str build on TP with a haste/acc/crit build. The crit build will easily outperform if the cap is low enough so that you dont have to mod a ton of gear. And yes, you are sacrificing acc and attck in a lot of slots for marginal DEX, but you get slight acc returns from it so it's not a huge waste, AND all of them put together increase crits by 10%+...

Please tell me how you suggest to keep all those gears AND dex?

Your examples,

Commanders cape - replaces a big attack boost, +15 attack STR 2.
Thunder/Rajas replace snipers, big acc boost.
Dex+2 earrings, you mean those rare ones that are hardly synthed? You'd replace Brutal earring with a dex earring? Suppa with a dex earring?
Spike necklace over Peacock Charm?

Already you lost a big chunk off melee build with those few items.

[Lhexh]

Placing Dex on earrings is about the stupidest thing you can do for most jobs since you are removing Brutal Earring and Suppa for 5 dex? Most people that eat meat are usually at a borderline on the acc boundary while keeping haste and sacrificing any acc at that point usually results in pretty noticeable drop in accuracy.

I think over the next few days a bit more info will come out from people that will help narrow what the exact dex values you need to make this viable. I also have a feeling it will only be viable for certain jobs at certain camps for specific races and will also favor sushi builds over meat builds, not only because of the increasing return on crits with a lower attack but also because of the bonus 6 dex to base the build on since most jobs cant sacrifice slots for dex outside of ranged, rings, back, neck without sacrificing haste substantially.

As of right now, I found some time to at least verify to myself that first hit on a WS cannot crit. I performed 10 SA+Shadowstiches as WAR/THF on Death Wasp in Rolanberry and using only 5 TP returns I had a range of 112-123. So far i've done 24 Shadowstiches without SA and have ranged 65-91 damage on 5 TP returns so I havent seen a crit on this WS yet. I plan to get the sample size up over 50-60 WS before I quit just to be sure.

[Nagamaki]

Brutal Earring for any melee, and Suppa for DW jobs are stuff you wouldn't want to remove, pretty much ever.

The whole issue with this crit tier is that unless you're already close to it, you won't be able to add enough DEX without big sacrifices. There's situations where going for the tier will be good, but there's also a lot of situations where it would be horribly counterproductive.

Its still useful to find where that tier is on various mob types, but expecting that to become the holy grail of damage dealing is retarded.

I'm bored to death of killing those retarded Too weak things, but here's the results with a full range of DEX. I actually could try even lower DEX with Alky's and stuff but huh... not now.


69 DEX (Elv MNK/THF base)
9.3%

79 DEX (Osode)
9.5%

82 DEX (Osode, Pixie)
15.8%

84 DEX (Byakko's)
16.3%

87 DEX (Byakko's, Pixie)
18.3%

89 DEX (Byakko's, Rajas)
21.2%

100 DEX (Byakko's, Rajas, Thunder, Squid)
23.7%

105 DEX (Byakko's, Rajas, Osode, Squid)
23.9%

119 DEX (Spike, Pixie, Osode, Rajas, Thunder, Warwolf, Byakko's, Melee Gaiters)
24.9%


Since my sample size is small in each test (300 swings) those rates do have a fair margin of error, but at least it shows a fairly clear curve, from what seems like a crit floor to a crit cap.

EDIT:

These are from my other set of tests, just to clarify the crit+% and max dex stuff

Thunder Staff
+15% Crit +4 DEX

136 DEX + 19% rate

300ish hits = 41.3%

-------------

Destroyers
+6% Crit

132 DEX + 10% rate

300ish hits = 30.2%

-------------

Faith Baghnakhs

132 DEX +4% rate

300ish hits = 26.1%

[suji]

79 DEX (Osode)
9.5%

82 DEX (Osode, Pixie)
15.8%

Since my sample size is small in each test (300 swings) those rates do have a fair margin of error, but at least it shows a fairly clear curve, from what seems like a crit floor to a crit cap.

Except that gap. Was anything weird about those particular tests? That's a pretty big jump for only 3 DEX.

Edit: Plotted those numbers just to see it more visually

http://img366.imageshack.us/img366/8641/plotvn1.png

[Nagamaki]

That particular gap was definitely a little suprising, I'm thinking its probably just statistical variation due to the small sample though. It did coincide with the switch to lightningsday and some lightning weather out in Lufaise. I don't recall any sort of past proof that those would affect crit rate, but it doesn't seem impossible. The 82 DEX set was the last I did, everything else was before lightningsday, and I didn't notice any weather.

[safer]

this might sound stupid, but, is it anyway possible to sneak in another point of dex for your build? earlier on in zirk/ravenlockes tests , putting aside maats cap, was that 1 build was below 120 dex and the other was above it

79->82 jumped in crit.. (increase of 6%~)
119->120 could potentially be another tier (making 24%->30%~ change of 6%)
40 dex could possibly be the first tier to break... maybe try taking a low level job and see if the crit hit rate doesnt go past 4%~

would be neat to see if these numbers played out like that,

also wouldnt it be better if tests were done without +crit gear / merits? since if there is a cap of 30% then it'd be hard to tell when you went over or not...

[Ferien]

I know you're bored of fighting the TW mobs...but that's a pretty significant difference between 79 and 82 DEX, and should be looked at more in depth, at 1 point increments. What mobs were you fighting?

[Tordall]

Yay Gnuplot!

You can have Gnuplot include statistical error, which might make it a bit easier to see what variation there is.

gnuplot > plot [xminmax][ymin:ymax] "datafilename" using ($xdatacolumn)$ydatacolumn)$Uncertainty) with errorbars

To me that looks like a phase transition. I'll see if I can find a nice formula with paramaters to fit for you.

--------------

The 'back half' of the curve looks like it could be fit well with A*(1-Exp(B*(DexDif)))

Where A is the max crit rate, B is the rate at which the crit rate approaches the max crit rate, and DexDif is the difference between your Dex (or which ever stat lowers crit hit rates) and the mob.

I'ld guess that the crit-hit rate is fixed (that is unaffected by dex) as long as the dex is lower than whatever the stat is that defends against crit hits.


--------------


My suggestion is that crit rate (without modifiers) is something like:

If Attacker's Crit up stat (Dex it appears) is higher than the Defender's Crit down stat (?????agility????):

Crit rate = A * (1 - Exp( B * (Defender's crit down stat - Attacker's Crit up stat)))

If the Attacker's Crit down stat is lower thant he defender's crit down stat then it appears to be a fixed number.

A appears to be around 20, B appears to be around 3, I'm running windows atm so I can't easily check.

+%crit from weapons, merits, and other gear is applied after the dex vs. ?agi? calculation.

[Bagel]

I'll laugh if this ends up being the new Suppa.

[Septimus]

I'll laugh if this ends up being the new Suppa.

Yay! I'm ahead of the curve!

[Tordall]

I'm just realizing that this could explain a lot about a few things I've wondered about with regaurds to ninja gear:

Haubergeon+1 > Koga Chainmail for total damage output during TP build

U/u being a viable combination of katanas.

Why Byakko's Hiadate rocks so hard.

Why P. Body doesn't seem to do anything for crit hit rates relative to haubergeon.

[Keno]

What you all seem to be missing during this little orgasmfest for DEX builds and crit rate is that Haste/Attack/Acc setup while eating meat will still out preform/out parse these DEX builds. Yes you might have a 30% crit rate but the guy next to you is still hitting harder and faster. These DEX builds just aren't practical unfortunately.

What you seem to be missing is that with DEX merits and the right pieces of gear (Commander's Cape, Thunder Ring/Rajas Ring, Dex+2 earrings on each ear, spike necklace [if you have to]) you can possibly break the crit tier while still KEEPING YOUR ENTIRE HASTE SET ON. So you're choosing between a haste/acc/attck/str build on TP with a haste/acc/crit build. The crit build will easily outperform if the cap is low enough so that you dont have to mod a ton of gear. And yes, you are sacrificing acc and attck in a lot of slots for marginal DEX, but you get slight acc returns from it so it's not a huge waste, AND all of them put together increase crits by 10%+...

Please tell me how you suggest to keep all those gears AND dex?

Your examples,

Commanders cape - replaces a big attack boost, +15 attack STR 2.
Thunder/Rajas replace snipers, big acc boost.
Dex+2 earrings, you mean those rare ones that are hardly synthed? You'd replace Brutal earring with a dex earring? Suppa with a dex earring?
Spike necklace over Peacock Charm?

Already you lost a big chunk off melee build with those few items.

Well when we get the exact DEX tier cap I'll let you know what the ideal build is and then we can test it. Speculation doesn't mean a whole lot.

EDIT (Dbl post): [Referring to the tests done above]
So with these results, one should draw that after 90 DEX you get diminishing crit rate% returns... And 90 DEX isn't a ton to shoot for, would be easy to build a proper build around that while keeping haste, brutal, and supp.

[Lhexh]

The 90 dex is based on a too weak mob in Lufaise Meadows and isn't reflective of merit party mobs. It seems from a few of the xp party parses a dex of at least 110 is required.